SlideShare una empresa de Scribd logo
1 de 72
Descargar para leer sin conexión
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents




           Scientific Uncertainty Analysis in Pressure Metrology
                          EURAMET.M.P-K13 as an Illustrative Case Study


                                                  Vishal Ramnath
                                         R&D Metrologist: Pressure & Vacuum
                                                      vramnath@nmisa.org

                                                     Pressure & Vacuum Laboratory


                                                                 April 19, 2011



                                                                                    1 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents




Table of contents I
       Review of Methods for Pressure Generation/Measurement
          Mechanism of Pressure Generation
          Mathematical Model of Pressure Balance
          Mechanism to Solve Formulation of Mathematical Equation
          Solution for Roots of Multivariable Equation
          Defining an EoS for the Working Fluid
          Solution of the Roots (Pressure) of the Nonlinear Equation
       Calculation of Generated Pressure Uncertainty
          Specific Issues Encountered in Uncertainty Analysis
       Methodology Used in Cross-Floating Analysis
         Derivation of Basis Equation used in Cross-Floating I.
         Estimation of the Straight Line Uncertainties Corresponding to u (A0 )
         and u (λ)
       Discussion
          Specific Comments on Intercomparison Results
          Recommendations Going Foward
                                                                                  2 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure I.
                In general there are different possible approaches in defining
                pressure, of which two predominate:




                                                                                3 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure I.
                In general there are different possible approaches in defining
                pressure, of which two predominate:
                Thermodynamic pressure:

                                                          1       ∂ ln Z            ∂F
                                                P=                             =−            (1)
                                                          β        ∂V      β        ∂V   T




                                                                                                   3 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure I.
                In general there are different possible approaches in defining
                pressure, of which two predominate:
                Thermodynamic pressure:

                                                          1       ∂ ln Z                   ∂F
                                                P=                                  =−                 (1)
                                                          β        ∂V           β          ∂V   T

                defined in terms of the Maxwell thermodynamic equations:
                                              ∂T                     ∂P             ∂S          ∂P
                                              ∂V S        =−         ∂S V   ,       ∂V T   =    ∂T V
                                              ∂S                     ∂V             ∂T          ∂V     (2)
                                              ∂P T        =−         ∂T P   ,       ∂P S   =    ∂S P




                                                                                                             3 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure I.
                In general there are different possible approaches in defining
                pressure, of which two predominate:
                Thermodynamic pressure:

                                                          1       ∂ ln Z                   ∂F
                                                P=                                  =−                                        (1)
                                                          β        ∂V           β          ∂V      T

                defined in terms of the Maxwell thermodynamic equations:
                                              ∂T                     ∂P             ∂S             ∂P
                                              ∂V S        =−         ∂S V   ,       ∂V T   =       ∂T V
                                              ∂S                     ∂V             ∂T             ∂V                         (2)
                                              ∂P T        =−         ∂T P   ,       ∂P S   =       ∂S P

                Recall: Partition function Z (T , V , N ) = r exp(−β Er ), r microstate,
                                                                          1
                Er energy for r microstate, temperature paramter β = kT , k
                Boltzmann fundamental constant, Helmholtz free energy
                                                                                               1          ∂ S (E ,V ,N )
                F (T , V , N ) = −kT ln Z (T , V , N ), temperature                            T
                                                                                                   =           ∂E
                                                                                                                           V ,N

                                                                                                                                    3 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure II.
                Mechanical pressure: The full situation is actually more complicated
                since recourse to Hamiltonian theory is required to precisely define
                “force” in terms of the generalized displacements and generalized
                momenta:
                                                n                                ∂H              ∂H
                                  H=            α=1
                                                             ˙
                                                          pα qα − L,     ˙
                                                                         p α = − ∂ qα ,   ˙
                                                                                          qα =   ∂ pα   (3)




                                                                                                              4 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure II.
                Mechanical pressure: The full situation is actually more complicated
                since recourse to Hamiltonian theory is required to precisely define
                “force” in terms of the generalized displacements and generalized
                momenta:
                                                n                                ∂H              ∂H
                                  H=            α=1
                                                             ˙
                                                          pα qα − L,     ˙
                                                                         p α = − ∂ qα ,   ˙
                                                                                          qα =   ∂ pα   (3)


                This conceptual framework is necessary since ideally for
                measurement equivalence
                                                       Pthermodynamic = Pmechanical                     (4)




                                                                                                              4 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure II.
                Mechanical pressure: The full situation is actually more complicated
                since recourse to Hamiltonian theory is required to precisely define
                “force” in terms of the generalized displacements and generalized
                momenta:
                                                n                                ∂H              ∂H
                                  H=            α=1
                                                             ˙
                                                          pα qα − L,     ˙
                                                                         p α = − ∂ qα ,   ˙
                                                                                          qα =   ∂ pα   (3)


                This conceptual framework is necessary since ideally for
                measurement equivalence
                                                       Pthermodynamic = Pmechanical                     (4)


                Utilizing the concepts of strain and stress the mechanical pressure
                for a fluid is defined in terms of the shear stress tensor as
                                      1                                              2
                                 ¯
                                 p = − (τxx + τyy + τzz ) = p −                    λ+ µ           ·V    (5)
                                      3                                              3
                                                                                                              4 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure III.
                Mechanical pressure definition: Without going into the details a
                working definition for pressure is to apply Stoke’s hypothesis
                    2
                λ + 3 µ = 0 and couple it with a defining equation of state so that

                                         ∂F
                             P=               ⇔ P = NkT i .e. PV = nRT (ideal gas EoS )   (6)
                                         ∂ An




                                                                                                5 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure III.
                Mechanical pressure definition: Without going into the details a
                working definition for pressure is to apply Stoke’s hypothesis
                    2
                λ + 3 µ = 0 and couple it with a defining equation of state so that

                                         ∂F
                             P=               ⇔ P = NkT i .e. PV = nRT (ideal gas EoS )   (6)
                                         ∂ An


                The above definition for pressure is in accordance to the BIPM
                choice of using mechanical pressure in a NMI scale realization

       On a Fundamental Scientific SI System of Units
       In line with international best practise there is a move to define the SI
       in terms of fundamental scientific constants, and although pressure is a
       derived unit from the seven base SI units viz. P = F developments such
                                                            A
       as the Watt balance and Avogadro projects to redefine the kilogram will
       logically also impact on the definition of pressure.
                                                                                                5 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure IV.
                In practical terms pressure is then essentially ”p = F ” and presently
                                                                     A
                the most accurate technique of generating a pressure is in terms of
                a pressure balance




          1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a

       vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions
       and direction and orientation e.g. strain
                                                                                                     6 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure IV.
                In practical terms pressure is then essentially ”p = F ” and presently
                                                                      A
                the most accurate technique of generating a pressure is in terms of
                a pressure balance
                A pressure balance is essentially a precision manufactured piston
                that is able to slide in a cylinder where the bottom of the piston is
                exposed to a fluid under pressure and the top of the piston is stacked
                with mass pieces which exert a downwards weight due to gravity




          1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a

       vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions
       and direction and orientation e.g. strain
                                                                                                     6 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure IV.
                In practical terms pressure is then essentially ”p = F ” and presently
                                                                      A
                the most accurate technique of generating a pressure is in terms of
                a pressure balance
                A pressure balance is essentially a precision manufactured piston
                that is able to slide in a cylinder where the bottom of the piston is
                exposed to a fluid under pressure and the top of the piston is stacked
                with mass pieces which exert a downwards weight due to gravity
                From the full definition of “pressure” the quantity p is in fact a tensor1
                but for most practical purposes we essentially consider it a scalar
                e.g. in hydrostatic hydraulic pressure we assume that the pressure
                is equal in all directions (true for Newtonian fluids)




          1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a

       vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions
       and direction and orientation e.g. strain
                                                                                                     6 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Generation of Reference Pressure IV.
                In practical terms pressure is then essentially ”p = F ” and presently
                                                                      A
                the most accurate technique of generating a pressure is in terms of
                a pressure balance
                A pressure balance is essentially a precision manufactured piston
                that is able to slide in a cylinder where the bottom of the piston is
                exposed to a fluid under pressure and the top of the piston is stacked
                with mass pieces which exert a downwards weight due to gravity
                From the full definition of “pressure” the quantity p is in fact a tensor1
                but for most practical purposes we essentially consider it a scalar
                e.g. in hydrostatic hydraulic pressure we assume that the pressure
                is equal in all directions (true for Newtonian fluids)
                For a piston-cylinder unit (PCU) there is then a fluid exerted
                pressure at the “bottom” and a mechanical loading exerted pressure
                at the “top”
          1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a

       vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions
       and direction and orientation e.g. strain
                                                                                                     6 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Illustrative Examples2 of Pressure Balances




             (a) Ruska model 2465A- (b) Ruska model 2485- (c) Desgranges & Huot
             U-754 pneumatic PB     930D hydraulic PB     model DH 5503S hydraulic
                                                          PB

                Figure: Low range PB (a), medium range PB (b) and high range PB (c)




         2 images courtesy of National Institute of Metrology (Thailand) website

       http://www.nimt.or.th/
                                                                                      7 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism of Pressure Generation




Schematic Illustration of the Pressure Balance




             Figure: Geometry and force analysis of a piston-cylinder pressure balance


                                                                                         8 / 30
Pressure & Vacuum Laboratory TAF 2011       Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Mathematical Model of Pressure Balance
       By an analysis of the physics of the problem one may derive the following
       mathematical model:

                                         n                       ρa
                                         i =1   mi 1 −           ρi      − Vs (ρf − ρa ) + H (ρf − ρa )S g + σ C
        p−Π             =
                                                                                         S
                                                                                                                                          (7)
                S       =       A0 (1 + λP )[1 + α(t − t0 )]                                                                              (8)
       In the above equations the nomenclature used is:


                          p       = absolute pressure ( true pressure )                                                                   (9)
                          Π =             ambient pressure ( top pressure )                                                             (10)
                          P       = p − Π = applied pressure ( gauge pressure)                                                          (11)
       In the case of the EURAMET.M.P-K13 key intercomparison Π is known since this the measured atmospheric pressure and in essence the above

       equations are to be solved to determine the unknown applied pressure P (and then by implication the absolute pressure p).
                                                                                                                                                 9 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece




                                                                          10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)




                                                                                         10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece




                                                                                         10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure




                                                                                               10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure




                                                                                               10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)




                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration




                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface




                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface
                S [m2 ] – ”effective area” of piston-cylinder




                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface
                S [m2 ] – ”effective area” of piston-cylinder
                A0 [m2 ] – zero-pressure area of PCU LS




                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface
                S [m2 ] – ”effective area” of piston-cylinder
                A0 [m2 ] – zero-pressure area of PCU LS
                λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa)



                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface
                S [m2 ] – ”effective area” of piston-cylinder
                A0 [m2 ] – zero-pressure area of PCU LS
                λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa)
                α [K−1 ] – thermal expansion coefficient term (α = αp + αc )


                                                                                                 10 / 30
Pressure & Vacuum Laboratory TAF 2011     Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mathematical Model of Pressure Balance




Explanation of Nomenclature Used in Mathematical Model
                mi [kg] – mass of the i th weight piece
                ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density)
                ρi [kg m−3 ] – density of the i th weight piece
                Vs [m3 ] – volume of the part of the piston which is submerged into the fluid
                which applies the pressure
                ρf [kg m−3 ] – density of the fluid which applies the pressure
                H [m] – height correction term to account for possible variation of laboratory
                standard (LS) and transfer standard (TS)
                g [m s−2 ] – local gravitational acceleration
                σ [N m−1 ] – surface tension term to account for surface tension effects
                between liquid (oil) and gas (air) interface
                S [m2 ] – ”effective area” of piston-cylinder
                A0 [m2 ] – zero-pressure area of PCU LS
                λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa)
                α [K−1 ] – thermal expansion coefficient term (α = αp + αc )
                C [m] – circumference of piston exposed to oil-air interface (typically set as
                C ≈ 4π A0 )                                                                      10 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Formulation of Mathematical Model as Nonlinear Root
Search I.
       Substituting the above terms we have
                                                                                n
        A0                                                                                      ρa
           (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]                            =          mi   1−        − Vs (ρf − ρa )
        g                                                                                       ρi
                                                                               i =1

                                                                               + H (ρf − ρa )A0 [1 + λ(p − Π)]

                                                                               × [1 + α(t − 20)]
                                                                                    σ                        −1
                                                                               +      (4π A0 [1 + α(t − 20)]) 2
                                                                                    g
                                                                                                              (12)

       which can be rewritten as a multi-variable equation f (X ) = 0 for which
       the roots (unknown pressure p) must be solved for.
                                                                                                                11 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Formulation of Mathematical Model as Nonlinear Root
Search II.
       The form of f (X ) is then:

                                            n                       ρa
                   f (X )       =           i =1   mi 1 −   − Vs (ρf − ρa )
                                                                    ρi
                                        + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                     −1                                 (13)
                                        + σ (4π A0 [1 + α(t − 20)]) 2
                                          g
                                             A0
                                        −     g
                                                  (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]
       where the input state variable X is


                              X         = [ m1 , m2 , . . . , mn ,                                      (14)
                                              ρ1 , ρ2 , . . . , ρn ,                                    (15)
                                                                                                    T
                                              A0 , λ, αp , αc , t , Vs , H , Π, σ, g , tair , RHair ]   (16)


                                                                                                               12 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Formulation of Mathematical Model as Nonlinear Root
Search III.

       The density of the air may be calculated using the CIPM-2007 formula3
       which expresses the air density in terms of
                ta – atmospheric temperature (measured with in-house thermometer
                traceable to water triple point (WTP) cell)
                pa – atmospheric pressure (measured with in-house barometer
                traceable to absolute medium gas pressures with Ruska LS)
                ha – atmospheric relative humidity expressed as a pure number
                such that 0 ha 1 (traceable to in-house hybrid humidity
                generator: validated in humidity regional inter-comparison)



          3 Metrologia 45 2008 149–155 ”Revised Formula for the Density of Moist Air

       (CIPM-2007)”, Picard A et al
                                                                                       13 / 30
Pressure & Vacuum Laboratory TAF 2011       Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Details of CIPM-2007 Formula for Moist Air Density
       Briefly the underlying formula is of form

                                                           pMa                                     Mv
                                                 ρa =          1 − xv                     1−                                          (17)
                                                           ZRT                                     Ma
       where
                ρa [kg m−3 ] – air density

                p [Pa] – atmospheric pressure (total pressure with implicit assumption of Dalton’s partial pressure law)

                t [◦ C] – air temperature

                T [K] – thermodynamic absolute temperature

                xv [−] – mole fraction of water vapour

                Ma [kg mol−1 ] – molar mass of dry air

                Mv [kg mol−1 ] – molar mass of water vapour

                Z [−] – compressibility factor

                R [J mol−1 K−1 ] – universal gas constant (using CODATA 2006 values4 cited in: Reviews of Modern Physics, Vol 80, April–June
                2008, Mohr P J et al)




            4 http://www.nist.gov/pml/div684/fcdc/upload/rmp2006-2.pdf
                                                                                                                                               14 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Note on Mass Input Values



                In the state variable we assume n mass pieces in total




                                                                           15 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Note on Mass Input Values



                In the state variable we assume n mass pieces in total
                This choice is incorporated by assigning the mass values to
                correspond to a weight set of N weights used (where n > N) and
                then appending the piston mass mpiston and bell mass mbell viz.

                    [m1 , m2 , . . . , mn−2 , mn−1 , mn ]T = [m1 , m2 , . . . , mN , mpiston , mbell ]T
                                                                                                     (18)




                                                                                                            15 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Mechanism to Solve Formulation of Mathematical Equation




Varying Meanings of Mass
       On the Differences Between True Mass, Conventional Mass and Apparent
       Mass
       Conventional mass: air density of 1.2 kg m−3 and mass density of 8000 kg m−3
       Detailsa Objective is to determine the true mass of a weight piece from calibration
       data:


                                                                                               ρa
                                                                                          1−   ρs
                                                            true mass : m          = ms        ρa                   (19)
                                                                                          1−   ρm
                                                                            (c )                     (s )
                                                           (c )            ρa                       ρa
           calibration cert data : ms                               1−      (c )
                                                                                   = ms      1−                     (20)
                                                                           ρs                        ρs
                                                                                                    (c )
                                                                                                  ρa
                                                                                             1−    (c )        1−   ρa
                                                                                      (c )        ρs                ρs
                                                                           ⇔m      = ms             (s)
                                                                                                           ×         (21)
                                                                                                                    ρa
                                                                                             1−   ρa           1−   ρm
                                                                                                  ρs

         a The Pressure Balance: Theory and Practice, Dadson R S et al, 1982, National Physical

       Laboratory / HMSO                                                                                                    16 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Solution for Roots of Multivariable Equation




Solution of Roots of f (X ) = 0
                 In the mathematical model of the EURAMET.M.P-K13 key
                 comparison there will in general be a variable number of inputs since
                 a varying number of mass pieces n are used to generate a
                 sequence of (applied) pressures for P ∈ [50, . . . , 500] MPa
                 It can be shown that for n mass pieces used in
                     X = [m1 , . . . , mn , ρ1 , . . . , ρn , A0 , λ, αp , αc , t , Vs , H , pa , ta , ha , σ, g ]T
                 there will be a total of (2n + 12) state variables incorporated into the
                 measurand model to solve for the unknown generated applied
                 pressure
                 Consider the terms in f (X ) which are known and unknown
                                                     n                        ρa
                         f (X )        =             i =1   mi 1 −  − Vs (ρf − ρa )
                                                                              ρi
                                                + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                             −1                                (22)
                                                + σ (4π A0 [1 + α(t − 20)]) 2
                                                  g
                                                     A0
                                                −     g
                                                          (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]
                                                                                                                      17 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Defining an EoS for the Working Fluid




EoS for a Pressure Balance Working Fluid I.
                During the intercomparison both the transfer standard and the NMISA
                Ruska high pressure PCU used di(2-ethylhexgl) sebecate i.e. ‘DHS’ as the
                working fluid formulated as
                    ρf /[kg m−3 ]              =       [912.7 + 0.752(10−6 p) − 1.65 × 10−3 (10−6 p)2
                                                       + 1.5 × 10−6 (10−6 p)3 ] × [1 − 7.8 × 10−4 (tf − 20)]
                                                                                                          (23)
                Alternative forms that may be considered are linear extrapolations from a
                known reference state                  ρ                            f0
                                                                              1+β(tf −tf 0 )
                                                                      ρf =         p −p
                                                                                                          (24)
                                                                              1 − E f0
                                                                                       f
                where tf 0 is the reference oil temperature at which the oil’s properties are
                known (say 20 ◦ C), Ef is the bulk modulas fluid elasticity, pf 0 is the referance
                oil fluid density pressure, tf is the current oil temperature and p is the current
                oil pressure
                For pure ideal gases the EoS is
                                                                                  ρRT
                                                                             p=                           (25)
                                                                                   M                             18 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Defining an EoS for the Working Fluid




EoS for a Pressure Balance Working Fluid II.


       On the Necessary Distinction Between Absolute and Applied Pressure
       The is a natural distinction between the absolute pressure p and the
       applied pressure P and equations-of-state are fundamentally defined in
       terms of the absolute pressure in order to be unambigous. Consider the
       following hypothetical cases:
            • case (i) {P (i ) = 200 kPa} ∪ {Π(i ) = 100 kPa} ⇒ p(i ) = 300 kPa
            • case (ii) {P (ii ) = 200 kPa} ∪ {Π(ii ) = 80 kPa} ⇒ p(ii ) = 280 kPa
       As a result even if the applied pressure is the same this does not neces-
       sary imply that the fluid density will be the same




                                                                                     19 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Solution of the Roots (Pressure) of the Nonlinear Equation




Numerical Solution for Roots of a Nonlinear Equation
                 From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T

                                                    n                        ρa
                        f (X )        =             i =1   mi 1 −  − Vs (ρf − ρa )
                                                                             ρi
                                               + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                            −1                  (26)
                                               + σ (4π A0 [1 + α(t − 20)]) 2
                                                 g
                                                    A0
                                               −     g
                                                         (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]




                                                                                                       20 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Solution of the Roots (Pressure) of the Nonlinear Equation




Numerical Solution for Roots of a Nonlinear Equation
                 From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T

                                                    n                        ρa
                        f (X )        =             i =1   mi 1 −  − Vs (ρf − ρa )
                                                                             ρi
                                               + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                            −1                  (26)
                                               + σ (4π A0 [1 + α(t − 20)]) 2
                                                 g
                                                    A0
                                               −     g
                                                         (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]

                 the original nonlinear equation may be transformed into an equation
                 in a single unknown Y by setting all the all components such that




                                                                                                       20 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Solution of the Roots (Pressure) of the Nonlinear Equation




Numerical Solution for Roots of a Nonlinear Equation
                 From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T

                                                    n                        ρa
                        f (X )        =             i =1   mi 1 −  − Vs (ρf − ρa )
                                                                             ρi
                                               + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                            −1                  (26)
                                               + σ (4π A0 [1 + α(t − 20)]) 2
                                                 g
                                                    A0
                                               −     g
                                                         (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]

                 the original nonlinear equation may be transformed into an equation
                 in a single unknown Y by setting all the all components such that
                 xi = p∀i ∈ [1, . . . , 2n + 12] as parameters i.e. by solving




                                                                                                       20 / 30
Pressure & Vacuum Laboratory TAF 2011        Navigate to Table of Contents

  Review of Methods for Pressure Generation/Measurement
      Solution of the Roots (Pressure) of the Nonlinear Equation




Numerical Solution for Roots of a Nonlinear Equation
                 From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T

                                                    n                        ρa
                        f (X )        =             i =1   mi 1 −  − Vs (ρf − ρa )
                                                                             ρi
                                               + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)]
                                                                            −1                  (26)
                                               + σ (4π A0 [1 + α(t − 20)]) 2
                                                 g
                                                    A0
                                               −     g
                                                         (p − Π)[1 + λ(p − Π)][1 + α(t − 20)]

                 the original nonlinear equation may be transformed into an equation
                 in a single unknown Y by setting all the all components such that
                 xi = p∀i ∈ [1, . . . , 2n + 12] as parameters i.e. by solving

                                                                       f (Y ; X ) = 0           (27)

                 The solution Y is of course the absolute pressure


                                                                                                       20 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Calculation of Generated Pressure Uncertainty




Estimation of Generated Pressure Uncertainty

                For the EURAMET.M.P-K13 key comparison V&V in terms of the
                uncertainty analysis by the research metrologist was performed
                within the framework of the GUM




         5 Cox M G and Harris P M, Software Support for Metrology Best Practice Guide No. 6 –

       Uncertainty Evaluation, Technical Report, National Physical Laboratory (United Kingdom),
       2004
                                                                                                  21 / 30
Pressure & Vacuum Laboratory TAF 2011      Navigate to Table of Contents

  Calculation of Generated Pressure Uncertainty




Estimation of Generated Pressure Uncertainty

                For the EURAMET.M.P-K13 key comparison V&V in terms of the
                uncertainty analysis by the research metrologist was performed
                within the framework of the GUM
                The underlying matrix based equation for the multi-variable input X
                was
                                             2
                                         ∂f
                                u 2 (y )       = ( x f )V x ( x )T               (28)
                                         ∂y
                following the methodology5 of Cox and Harris.
                In our initial approach I assumed no correlation effects in the
                components of the input variable X for simplicity


         5 Cox M G and Harris P M, Software Support for Metrology Best Practice Guide No. 6 –

       Uncertainty Evaluation, Technical Report, National Physical Laboratory (United Kingdom),
       2004
                                                                                                  21 / 30
Pressure & Vacuum Laboratory TAF 2011       Navigate to Table of Contents

  Calculation of Generated Pressure Uncertainty
      Specific Issues Encountered in Uncertainty Analysis




The Validity of the Assumption of
Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12
                 In general the only possible significant non-zero correlation effect
                 that may occur is for the case Cov(A0 , λ)




           6 consultation of the statistical literature such as Metrologia journal articles and IMEKO

       conferences states that it is poor practice to simply assume unity correlation in the absence
       of information: the recommended option is to assume zero correlation
                                                                                                        22 / 30
Pressure & Vacuum Laboratory TAF 2011       Navigate to Table of Contents

  Calculation of Generated Pressure Uncertainty
      Specific Issues Encountered in Uncertainty Analysis




The Validity of the Assumption of
Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12
                 In general the only possible significant non-zero correlation effect
                 that may occur is for the case Cov(A0 , λ)
                 Correlation effects do occur in the case of the weights mass/density
                 but in general this has a very minor impact on final uncertainties as
                 u (A0 ) and particularly u (λ) will dominate uncertainty contributions at
                 elevated pressures




           6 consultation of the statistical literature such as Metrologia journal articles and IMEKO

       conferences states that it is poor practice to simply assume unity correlation in the absence
       of information: the recommended option is to assume zero correlation
                                                                                                        22 / 30
Pressure & Vacuum Laboratory TAF 2011       Navigate to Table of Contents

  Calculation of Generated Pressure Uncertainty
      Specific Issues Encountered in Uncertainty Analysis




The Validity of the Assumption of
Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12
                 In general the only possible significant non-zero correlation effect
                 that may occur is for the case Cov(A0 , λ)
                 Correlation effects do occur in the case of the weights mass/density
                 but in general this has a very minor impact on final uncertainties as
                 u (A0 ) and particularly u (λ) will dominate uncertainty contributions at
                 elevated pressures
                                                                               (LS )
                 Correlation is certain to occur in the present instance as A0 and
                 λ(LS ) are in fact obtained from an experimental cross-floating
                 againsts another precision piston-cylinder standard however in the
                 present instance no accessible information on Cov(A0 , λ) is
                 available as these data values are sourced from an external
                 certificate that does not provide any information from which one may
                 reasonably infer a magnnitude of correlation6
           6 consultation of the statistical literature such as Metrologia journal articles and IMEKO

       conferences states that it is poor practice to simply assume unity correlation in the absence
       of information: the recommended option is to assume zero correlation
                                                                                                        22 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating
       Make the assumption of linear elasticity theory in the PDE’s formulation
       of the strain/stress constituent behaviour:
                 Assume a piston-cylinder A with known characteristics i.e.
                        (A)
                 DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with
                                                                              (B )
                 unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be
                 determined




                                                                                         23 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating
       Make the assumption of linear elasticity theory in the PDE’s formulation
       of the strain/stress constituent behaviour:
                 Assume a piston-cylinder A with known characteristics i.e.
                        (A)
                 DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with
                                                                                     (B )
                 unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be
                 determined
                 It follows that
                                                                       m
                                                1                                           ρa
                             PB        =           ×              g           mB j   1−            − Vs B (ρf − ρa )g
                                                SB                                          ρB j
                                                                      j =1

                                                                      (B )
                                                + σB           4π A0           + (HB (ρf − ρa )g ) SB                   (29)




                                                                                                                               23 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating
       Make the assumption of linear elasticity theory in the PDE’s formulation
       of the strain/stress constituent behaviour:
                 Assume a piston-cylinder A with known characteristics i.e.
                        (A)
                 DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with
                                                                                     (B )
                 unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be
                 determined
                 It follows that
                                                                       m
                                                1                                           ρa
                             PB        =           ×              g           mB j   1−            − Vs B (ρf − ρa )g
                                                SB                                          ρB j
                                                                      j =1

                                                                      (B )
                                                + σB           4π A0           + (HB (ρf − ρa )g ) SB                   (29)




                                                                                                                               23 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating II.
                 Noting that HB = 0 since the height correction was already
                 accounted for in PCU A when working out the applied pressure




                                                                                24 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating II.
                 Noting that HB = 0 since the height correction was already
                 accounted for in PCU A when working out the applied pressure

                                                                m
                                           1                                        ρa                                  (B )
                 ⇒ SB             =           × g                     mB j    1−            − VB (ρf − ρa )g + σB   4π A0
                                           PA                                      ρm B j
                                                               j =1

                                                                                                                       (30)




                                                                                                                        24 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating II.
                 Noting that HB = 0 since the height correction was already
                 accounted for in PCU A when working out the applied pressure

                                                                m
                                           1                                          ρa                                     (B )
                 ⇒ SB             =           × g                     mB j     1−             − VB (ρf − ρa )g + σB     4π A0
                                           PA                                        ρm B j
                                                               j =1

                                                                                                                             (30)


                 As a result there is now a data point k with {P , SB } where
                                                                              (B )
                                                                SB = A0 (1 + λ(B ) P )                                (31)




                                                                                                                             24 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating II.
                 Noting that HB = 0 since the height correction was already
                 accounted for in PCU A when working out the applied pressure

                                                                m
                                           1                                          ρa                                     (B )
                 ⇒ SB             =           × g                     mB j     1−             − VB (ρf − ρa )g + σB     4π A0
                                           PA                                        ρm B j
                                                               j =1

                                                                                                                             (30)


                 As a result there is now a data point k with {P , SB } where
                                                                              (B )
                                                                SB = A0 (1 + λ(B ) P )                                (31)

                 for k ∈ [1, . . . , 10] corresponding to pressures [50, . . . , 500] MPa
                 from which a straight line may be plotted in an analogous form to


                                                                                                                             24 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Derivation of Basis Equation used in Cross-Floating I.




Basis Equation for Cross-Floating II.
                 Noting that HB = 0 since the height correction was already
                 accounted for in PCU A when working out the applied pressure

                                                                m
                                           1                                          ρa                                     (B )
                 ⇒ SB             =           × g                     mB j     1−             − VB (ρf − ρa )g + σB     4π A0
                                           PA                                        ρm B j
                                                               j =1

                                                                                                                             (30)


                 As a result there is now a data point k with {P , SB } where
                                                                              (B )
                                                                SB = A0 (1 + λ(B ) P )                                (31)

                 for k ∈ [1, . . . , 10] corresponding to pressures [50, . . . , 500] MPa
                 from which a straight line may be plotted in an analogous form to

                                   y = ax + b ⇐ D = {Dk = (xk , yk )|k ∈ [1, . . . , 10]}                             (32)
                                                                                                                             24 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ)




Estimation of Zero-Pressure and Distortion Coefficient
Uncertainties I.

                 Standard linear regression analysis in spreadsheet environments
                 are not suitable for the provision of adequate parameter
                 uncertainties




                                                                                         25 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ)




Estimation of Zero-Pressure and Distortion Coefficient
Uncertainties I.

                 Standard linear regression analysis in spreadsheet environments
                 are not suitable for the provision of adequate parameter
                 uncertainties
                 In general optimization routines are necessary for a χ2 merit
                 function determination and recently a practical implementation has
                 been proposed in the scientific literature by scientists at the PTB
                 (Germany): A Least Squares Algorithm for Fitting Data Points with
                 Mutually Correlated Coordinates to a Straight Line, Meas. Sci.
                 Technol. 22 (2011) doi: 10.1088/0957-0233/22/3/035101, Krystek M
                 and Anton M




                                                                                         25 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ)




Estimation of Zero-Pressure and Distortion Coefficient
Uncertainties I.

                 Standard linear regression analysis in spreadsheet environments
                 are not suitable for the provision of adequate parameter
                 uncertainties
                 In general optimization routines are necessary for a χ2 merit
                 function determination and recently a practical implementation has
                 been proposed in the scientific literature by scientists at the PTB
                 (Germany): A Least Squares Algorithm for Fitting Data Points with
                 Mutually Correlated Coordinates to a Straight Line, Meas. Sci.
                 Technol. 22 (2011) doi: 10.1088/0957-0233/22/3/035101, Krystek M
                 and Anton M




                                                                                         25 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ)




Estimation of Zero-Pressure and Distortion Coefficient
Uncertainties II.
                 The method fits y = ax + b and provides estimates for u (a) and
                 u (b) noting that for a pressure balance the defining basis equation is

                                                                  A(P ) = A0 (1 + λP )   (33)




                                                                                                26 / 30
Pressure & Vacuum Laboratory TAF 2011         Navigate to Table of Contents

  Methodology Used in Cross-Floating Analysis
      Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ)




Estimation of Zero-Pressure and Distortion Coefficient
Uncertainties II.
                 The method fits y = ax + b and provides estimates for u (a) and
                 u (b) noting that for a pressure balance the defining basis equation is

                                                                  A(P ) = A0 (1 + λP )                (33)


                 It then follows that in a pressure cross-floating experimental
                 measurement that

                                                       u 2 (a)         = u 2 (A0 )                    (34)
                                                          2                       2 2
                                                       u (b )          = λ u (A0 ) +     A2 u 2 (λ)
                                                                                          0           (35)
                                                                          1
                                                 ⇒ u 2 (λ)             =     u 2 (b) − λ2 u 2 (a)     (36)
                                                                         A20


                                                                                                             26 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Discussion
      Specific Comments on Intercomparison Results




Specific Comments on Intercomparison Participation
                The participation of the technical metrologists in
                EURAMET.M.P-K13 can be considered a learning experience as
                there is a significant difference is the level and complexity of key
                comparisons when compared to industrial level measurements




                                                                                      27 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Discussion
      Specific Comments on Intercomparison Results




Specific Comments on Intercomparison Participation
                The participation of the technical metrologists in
                EURAMET.M.P-K13 can be considered a learning experience as
                there is a significant difference is the level and complexity of key
                comparisons when compared to industrial level measurements
                It is recognized that in terms of data processing, analysis, and
                post-processing of results as a NMI there is always room for
                improvement and that there are certain inherent fundamental
                limitations in spreadsheet utilization, both in terms of what can be
                mathematically/statistically achieved and also in terms of validation
                and verification




                                                                                        27 / 30
Pressure & Vacuum Laboratory TAF 2011    Navigate to Table of Contents

  Discussion
      Specific Comments on Intercomparison Results




Specific Comments on Intercomparison Participation
                The participation of the technical metrologists in
                EURAMET.M.P-K13 can be considered a learning experience as
                there is a significant difference is the level and complexity of key
                comparisons when compared to industrial level measurements
                It is recognized that in terms of data processing, analysis, and
                post-processing of results as a NMI there is always room for
                improvement and that there are certain inherent fundamental
                limitations in spreadsheet utilization, both in terms of what can be
                mathematically/statistically achieved and also in terms of validation
                and verification
                In terms of internal comparison of processed results there is good
                general agreement in generated pressures and fair/moderate
                agreement in computed effective area and distortion coefficient of
                applicable transfer standard


                                                                                        27 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Recommendations and Future Options Going Forward I.


                There is from the research metrologist’s perspective a need to more
                fully examine the ratio method of analysis in cross-floating as this
                may present a more numerically stable implementation than a more
                direct P vs. A(P ) regression analysis however this must be weighed
                up against the implication of directly introducing the LS’s distortion
                coefficient into the uncertainty analysis




                                                                                         28 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Recommendations and Future Options Going Forward I.


                There is from the research metrologist’s perspective a need to more
                fully examine the ratio method of analysis in cross-floating as this
                may present a more numerically stable implementation than a more
                direct P vs. A(P ) regression analysis however this must be weighed
                up against the implication of directly introducing the LS’s distortion
                coefficient into the uncertainty analysis
                There is a need to more fully investigate the actual correlation
                between the parameter estimates for A0 and λ as opposed to a
                simplistic unity correlation that merely provides an upper bound




                                                                                         28 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Recommendations and Future Options Going Forward II.
                FEM numerical simulations should be performed to benchmark the
                existing experimental/statistical distortion coefficient however the
                challenge posed is not actually in terms of the FEM analysis but
                rather in terms of the materials properties of the piston-cylinder:
                tungsten-carbide is prone to some variation in the Young’s modulas
                of elasticty and Poisson’s ratio – a possibility is to submit a research
                proposal for an electrical capacitance / materials characterization
                experimental study of the distortion coefficent on the low and
                medium pressure Ruska piston-cylinder laboratory standards




                                                                                           29 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Recommendations and Future Options Going Forward II.
                FEM numerical simulations should be performed to benchmark the
                existing experimental/statistical distortion coefficient however the
                challenge posed is not actually in terms of the FEM analysis but
                rather in terms of the materials properties of the piston-cylinder:
                tungsten-carbide is prone to some variation in the Young’s modulas
                of elasticty and Poisson’s ratio – a possibility is to submit a research
                proposal for an electrical capacitance / materials characterization
                experimental study of the distortion coefficent on the low and
                medium pressure Ruska piston-cylinder laboratory standards
                It may be beneficial to consider alternative statistical formulations to
                bootstrapping such as Jackknifing




                                                                                           29 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Recommendations and Future Options Going Forward II.
                FEM numerical simulations should be performed to benchmark the
                existing experimental/statistical distortion coefficient however the
                challenge posed is not actually in terms of the FEM analysis but
                rather in terms of the materials properties of the piston-cylinder:
                tungsten-carbide is prone to some variation in the Young’s modulas
                of elasticty and Poisson’s ratio – a possibility is to submit a research
                proposal for an electrical capacitance / materials characterization
                experimental study of the distortion coefficent on the low and
                medium pressure Ruska piston-cylinder laboratory standards
                It may be beneficial to consider alternative statistical formulations to
                bootstrapping such as Jackknifing
                Certain aspects of the Pressure & Vacuum Laboratory R&D on the
                Ruska high pressure characterization may be able to (in partnership
                with another NMI e.g. PTB / CMI) make a modest contribution on
                the Avogadro project that requires precision pressure
                measurements at 7 MPa
                                                                                           29 / 30
Pressure & Vacuum Laboratory TAF 2011   Navigate to Table of Contents

  Discussion
      Recommendations Going Foward




Acknowledgements

       This work was performed with financial support of the South African
       Department of Trade and Industry




                                                                            30 / 30

Más contenido relacionado

Destacado

Presentation 1 4 ginanger
Presentation 1 4 ginangerPresentation 1 4 ginanger
Presentation 1 4 ginangerGigin Ginanjar
 
Vacuum Class -Gas flow - gigin 1
Vacuum Class -Gas flow - gigin 1Vacuum Class -Gas flow - gigin 1
Vacuum Class -Gas flow - gigin 1Gigin Ginanjar
 
Demand Pressure: The Balance Between Candidate Supply & Hiring Demand
Demand Pressure: The Balance Between Candidate Supply & Hiring DemandDemand Pressure: The Balance Between Candidate Supply & Hiring Demand
Demand Pressure: The Balance Between Candidate Supply & Hiring DemandWANTED Technologies
 
push button and pressure balance cartridge -Jenny
push button and pressure balance cartridge -Jennypush button and pressure balance cartridge -Jenny
push button and pressure balance cartridge -JennyJieli Tan
 
Pressure balance calibration in kriss
Pressure balance calibration in krissPressure balance calibration in kriss
Pressure balance calibration in krissGigin Ginanjar
 
Establishment of pressure standard using pressure balance
Establishment of pressure standard using pressure balanceEstablishment of pressure standard using pressure balance
Establishment of pressure standard using pressure balanceGigin Ginanjar
 
Vacuum Class - Getter pump
Vacuum Class - Getter pumpVacuum Class - Getter pump
Vacuum Class - Getter pumpGigin Ginanjar
 
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...Review of Methodology and Rationale of Monte Carlo Simulation - Application t...
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...vramnath
 
Body Fluid And Electrolyte Balance
Body Fluid And Electrolyte BalanceBody Fluid And Electrolyte Balance
Body Fluid And Electrolyte Balancemvraveendrambbs
 
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...Lawrence kok
 

Destacado (12)

Presentation 1 4 ginanger
Presentation 1 4 ginangerPresentation 1 4 ginanger
Presentation 1 4 ginanger
 
Vacuum Class -Gas flow - gigin 1
Vacuum Class -Gas flow - gigin 1Vacuum Class -Gas flow - gigin 1
Vacuum Class -Gas flow - gigin 1
 
Demand Pressure: The Balance Between Candidate Supply & Hiring Demand
Demand Pressure: The Balance Between Candidate Supply & Hiring DemandDemand Pressure: The Balance Between Candidate Supply & Hiring Demand
Demand Pressure: The Balance Between Candidate Supply & Hiring Demand
 
push button and pressure balance cartridge -Jenny
push button and pressure balance cartridge -Jennypush button and pressure balance cartridge -Jenny
push button and pressure balance cartridge -Jenny
 
Pressure balance calibration in kriss
Pressure balance calibration in krissPressure balance calibration in kriss
Pressure balance calibration in kriss
 
Establishment of pressure standard using pressure balance
Establishment of pressure standard using pressure balanceEstablishment of pressure standard using pressure balance
Establishment of pressure standard using pressure balance
 
Vacuum Class - Getter pump
Vacuum Class - Getter pumpVacuum Class - Getter pump
Vacuum Class - Getter pump
 
Renal physiology bpums
Renal physiology bpumsRenal physiology bpums
Renal physiology bpums
 
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...Review of Methodology and Rationale of Monte Carlo Simulation - Application t...
Review of Methodology and Rationale of Monte Carlo Simulation - Application t...
 
Pressure Measurement Part I
Pressure Measurement   Part IPressure Measurement   Part I
Pressure Measurement Part I
 
Body Fluid And Electrolyte Balance
Body Fluid And Electrolyte BalanceBody Fluid And Electrolyte Balance
Body Fluid And Electrolyte Balance
 
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...
IB Chemistry, IB Biology on Uncertainty calculation, error analysis and stand...
 

Similar a Scientific Uncertainty Analysis in Pressure

The joule thomson experiment
The joule thomson experimentThe joule thomson experiment
The joule thomson experimentMidoOoz
 
thermodynamicsrelations-161231100305.pdf
thermodynamicsrelations-161231100305.pdfthermodynamicsrelations-161231100305.pdf
thermodynamicsrelations-161231100305.pdfMarcia184919
 
Thermodynamics relations
Thermodynamics relationsThermodynamics relations
Thermodynamics relationsnaphis ahamad
 
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficient
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficientThermodynamic relations, Clausius Clapreyon equation , joule thomson coefficient
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficientPEC University Chandigarh
 
publication_2_21613_432.pptxbvgvgghcghfcc
publication_2_21613_432.pptxbvgvgghcghfccpublication_2_21613_432.pptxbvgvgghcghfcc
publication_2_21613_432.pptxbvgvgghcghfccssuser8900291
 
P&w tables of compressible flow functions
P&w tables of compressible flow functionsP&w tables of compressible flow functions
P&w tables of compressible flow functionsJulio Banks
 
Thermodynamicsworkandheatdrkatte
ThermodynamicsworkandheatdrkatteThermodynamicsworkandheatdrkatte
Thermodynamicsworkandheatdrkattedrkatte
 
007 isometric process
007 isometric process007 isometric process
007 isometric processphysics101
 
006 isobaric process
006 isobaric process006 isobaric process
006 isobaric processphysics101
 
Redlich kwong, 1949
Redlich kwong, 1949Redlich kwong, 1949
Redlich kwong, 1949Vero Miranda
 

Similar a Scientific Uncertainty Analysis in Pressure (13)

Sol8
Sol8Sol8
Sol8
 
The joule thomson experiment
The joule thomson experimentThe joule thomson experiment
The joule thomson experiment
 
thermodynamicsrelations-161231100305.pdf
thermodynamicsrelations-161231100305.pdfthermodynamicsrelations-161231100305.pdf
thermodynamicsrelations-161231100305.pdf
 
Thermodynamics relations
Thermodynamics relationsThermodynamics relations
Thermodynamics relations
 
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficient
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficientThermodynamic relations, Clausius Clapreyon equation , joule thomson coefficient
Thermodynamic relations, Clausius Clapreyon equation , joule thomson coefficient
 
publication_2_21613_432.pptxbvgvgghcghfcc
publication_2_21613_432.pptxbvgvgghcghfccpublication_2_21613_432.pptxbvgvgghcghfcc
publication_2_21613_432.pptxbvgvgghcghfcc
 
P&w tables of compressible flow functions
P&w tables of compressible flow functionsP&w tables of compressible flow functions
P&w tables of compressible flow functions
 
Thermodynamicsworkandheatdrkatte
ThermodynamicsworkandheatdrkatteThermodynamicsworkandheatdrkatte
Thermodynamicsworkandheatdrkatte
 
Thermodynamics
ThermodynamicsThermodynamics
Thermodynamics
 
007 isometric process
007 isometric process007 isometric process
007 isometric process
 
Statistical Physics Assignment Help
Statistical Physics Assignment Help Statistical Physics Assignment Help
Statistical Physics Assignment Help
 
006 isobaric process
006 isobaric process006 isobaric process
006 isobaric process
 
Redlich kwong, 1949
Redlich kwong, 1949Redlich kwong, 1949
Redlich kwong, 1949
 

Scientific Uncertainty Analysis in Pressure

  • 1. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Scientific Uncertainty Analysis in Pressure Metrology EURAMET.M.P-K13 as an Illustrative Case Study Vishal Ramnath R&D Metrologist: Pressure & Vacuum vramnath@nmisa.org Pressure & Vacuum Laboratory April 19, 2011 1 / 30
  • 2. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Table of contents I Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Mathematical Model of Pressure Balance Mechanism to Solve Formulation of Mathematical Equation Solution for Roots of Multivariable Equation Defining an EoS for the Working Fluid Solution of the Roots (Pressure) of the Nonlinear Equation Calculation of Generated Pressure Uncertainty Specific Issues Encountered in Uncertainty Analysis Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Discussion Specific Comments on Intercomparison Results Recommendations Going Foward 2 / 30
  • 3. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure I. In general there are different possible approaches in defining pressure, of which two predominate: 3 / 30
  • 4. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure I. In general there are different possible approaches in defining pressure, of which two predominate: Thermodynamic pressure: 1 ∂ ln Z ∂F P= =− (1) β ∂V β ∂V T 3 / 30
  • 5. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure I. In general there are different possible approaches in defining pressure, of which two predominate: Thermodynamic pressure: 1 ∂ ln Z ∂F P= =− (1) β ∂V β ∂V T defined in terms of the Maxwell thermodynamic equations: ∂T ∂P ∂S ∂P ∂V S =− ∂S V , ∂V T = ∂T V ∂S ∂V ∂T ∂V (2) ∂P T =− ∂T P , ∂P S = ∂S P 3 / 30
  • 6. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure I. In general there are different possible approaches in defining pressure, of which two predominate: Thermodynamic pressure: 1 ∂ ln Z ∂F P= =− (1) β ∂V β ∂V T defined in terms of the Maxwell thermodynamic equations: ∂T ∂P ∂S ∂P ∂V S =− ∂S V , ∂V T = ∂T V ∂S ∂V ∂T ∂V (2) ∂P T =− ∂T P , ∂P S = ∂S P Recall: Partition function Z (T , V , N ) = r exp(−β Er ), r microstate, 1 Er energy for r microstate, temperature paramter β = kT , k Boltzmann fundamental constant, Helmholtz free energy 1 ∂ S (E ,V ,N ) F (T , V , N ) = −kT ln Z (T , V , N ), temperature T = ∂E V ,N 3 / 30
  • 7. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure II. Mechanical pressure: The full situation is actually more complicated since recourse to Hamiltonian theory is required to precisely define “force” in terms of the generalized displacements and generalized momenta: n ∂H ∂H H= α=1 ˙ pα qα − L, ˙ p α = − ∂ qα , ˙ qα = ∂ pα (3) 4 / 30
  • 8. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure II. Mechanical pressure: The full situation is actually more complicated since recourse to Hamiltonian theory is required to precisely define “force” in terms of the generalized displacements and generalized momenta: n ∂H ∂H H= α=1 ˙ pα qα − L, ˙ p α = − ∂ qα , ˙ qα = ∂ pα (3) This conceptual framework is necessary since ideally for measurement equivalence Pthermodynamic = Pmechanical (4) 4 / 30
  • 9. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure II. Mechanical pressure: The full situation is actually more complicated since recourse to Hamiltonian theory is required to precisely define “force” in terms of the generalized displacements and generalized momenta: n ∂H ∂H H= α=1 ˙ pα qα − L, ˙ p α = − ∂ qα , ˙ qα = ∂ pα (3) This conceptual framework is necessary since ideally for measurement equivalence Pthermodynamic = Pmechanical (4) Utilizing the concepts of strain and stress the mechanical pressure for a fluid is defined in terms of the shear stress tensor as 1 2 ¯ p = − (τxx + τyy + τzz ) = p − λ+ µ ·V (5) 3 3 4 / 30
  • 10. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure III. Mechanical pressure definition: Without going into the details a working definition for pressure is to apply Stoke’s hypothesis 2 λ + 3 µ = 0 and couple it with a defining equation of state so that ∂F P= ⇔ P = NkT i .e. PV = nRT (ideal gas EoS ) (6) ∂ An 5 / 30
  • 11. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure III. Mechanical pressure definition: Without going into the details a working definition for pressure is to apply Stoke’s hypothesis 2 λ + 3 µ = 0 and couple it with a defining equation of state so that ∂F P= ⇔ P = NkT i .e. PV = nRT (ideal gas EoS ) (6) ∂ An The above definition for pressure is in accordance to the BIPM choice of using mechanical pressure in a NMI scale realization On a Fundamental Scientific SI System of Units In line with international best practise there is a move to define the SI in terms of fundamental scientific constants, and although pressure is a derived unit from the seven base SI units viz. P = F developments such A as the Watt balance and Avogadro projects to redefine the kilogram will logically also impact on the definition of pressure. 5 / 30
  • 12. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure IV. In practical terms pressure is then essentially ”p = F ” and presently A the most accurate technique of generating a pressure is in terms of a pressure balance 1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions and direction and orientation e.g. strain 6 / 30
  • 13. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure IV. In practical terms pressure is then essentially ”p = F ” and presently A the most accurate technique of generating a pressure is in terms of a pressure balance A pressure balance is essentially a precision manufactured piston that is able to slide in a cylinder where the bottom of the piston is exposed to a fluid under pressure and the top of the piston is stacked with mass pieces which exert a downwards weight due to gravity 1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions and direction and orientation e.g. strain 6 / 30
  • 14. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure IV. In practical terms pressure is then essentially ”p = F ” and presently A the most accurate technique of generating a pressure is in terms of a pressure balance A pressure balance is essentially a precision manufactured piston that is able to slide in a cylinder where the bottom of the piston is exposed to a fluid under pressure and the top of the piston is stacked with mass pieces which exert a downwards weight due to gravity From the full definition of “pressure” the quantity p is in fact a tensor1 but for most practical purposes we essentially consider it a scalar e.g. in hydrostatic hydraulic pressure we assume that the pressure is equal in all directions (true for Newtonian fluids) 1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions and direction and orientation e.g. strain 6 / 30
  • 15. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Generation of Reference Pressure IV. In practical terms pressure is then essentially ”p = F ” and presently A the most accurate technique of generating a pressure is in terms of a pressure balance A pressure balance is essentially a precision manufactured piston that is able to slide in a cylinder where the bottom of the piston is exposed to a fluid under pressure and the top of the piston is stacked with mass pieces which exert a downwards weight due to gravity From the full definition of “pressure” the quantity p is in fact a tensor1 but for most practical purposes we essentially consider it a scalar e.g. in hydrostatic hydraulic pressure we assume that the pressure is equal in all directions (true for Newtonian fluids) For a piston-cylinder unit (PCU) there is then a fluid exerted pressure at the “bottom” and a mechanical loading exerted pressure at the “top” 1 a scalar has zero dimensions and only a magnitude e.g. quantity of matter [moles], a vector has one dimension and a direction e.g. velocity of wind, a matrix has two dimensions and direction and orientation e.g. strain 6 / 30
  • 16. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Illustrative Examples2 of Pressure Balances (a) Ruska model 2465A- (b) Ruska model 2485- (c) Desgranges & Huot U-754 pneumatic PB 930D hydraulic PB model DH 5503S hydraulic PB Figure: Low range PB (a), medium range PB (b) and high range PB (c) 2 images courtesy of National Institute of Metrology (Thailand) website http://www.nimt.or.th/ 7 / 30
  • 17. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism of Pressure Generation Schematic Illustration of the Pressure Balance Figure: Geometry and force analysis of a piston-cylinder pressure balance 8 / 30
  • 18. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Mathematical Model of Pressure Balance By an analysis of the physics of the problem one may derive the following mathematical model: n ρa i =1 mi 1 − ρi − Vs (ρf − ρa ) + H (ρf − ρa )S g + σ C p−Π = S (7) S = A0 (1 + λP )[1 + α(t − t0 )] (8) In the above equations the nomenclature used is: p = absolute pressure ( true pressure ) (9) Π = ambient pressure ( top pressure ) (10) P = p − Π = applied pressure ( gauge pressure) (11) In the case of the EURAMET.M.P-K13 key intercomparison Π is known since this the measured atmospheric pressure and in essence the above equations are to be solved to determine the unknown applied pressure P (and then by implication the absolute pressure p). 9 / 30
  • 19. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece 10 / 30
  • 20. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) 10 / 30
  • 21. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece 10 / 30
  • 22. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure 10 / 30
  • 23. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure 10 / 30
  • 24. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) 10 / 30
  • 25. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration 10 / 30
  • 26. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface 10 / 30
  • 27. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface S [m2 ] – ”effective area” of piston-cylinder 10 / 30
  • 28. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface S [m2 ] – ”effective area” of piston-cylinder A0 [m2 ] – zero-pressure area of PCU LS 10 / 30
  • 29. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface S [m2 ] – ”effective area” of piston-cylinder A0 [m2 ] – zero-pressure area of PCU LS λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa) 10 / 30
  • 30. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface S [m2 ] – ”effective area” of piston-cylinder A0 [m2 ] – zero-pressure area of PCU LS λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa) α [K−1 ] – thermal expansion coefficient term (α = αp + αc ) 10 / 30
  • 31. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mathematical Model of Pressure Balance Explanation of Nomenclature Used in Mathematical Model mi [kg] – mass of the i th weight piece ρa [kg m−3 ] – density of the ambient fluid above the PCU (air density) ρi [kg m−3 ] – density of the i th weight piece Vs [m3 ] – volume of the part of the piston which is submerged into the fluid which applies the pressure ρf [kg m−3 ] – density of the fluid which applies the pressure H [m] – height correction term to account for possible variation of laboratory standard (LS) and transfer standard (TS) g [m s−2 ] – local gravitational acceleration σ [N m−1 ] – surface tension term to account for surface tension effects between liquid (oil) and gas (air) interface S [m2 ] – ”effective area” of piston-cylinder A0 [m2 ] – zero-pressure area of PCU LS λ [Pa−1 ] – distortion coefficient of PCU (typically measured in ppm/MPa) α [K−1 ] – thermal expansion coefficient term (α = αp + αc ) C [m] – circumference of piston exposed to oil-air interface (typically set as C ≈ 4π A0 ) 10 / 30
  • 32. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Formulation of Mathematical Model as Nonlinear Root Search I. Substituting the above terms we have n A0 ρa (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] = mi 1− − Vs (ρf − ρa ) g ρi i =1 + H (ρf − ρa )A0 [1 + λ(p − Π)] × [1 + α(t − 20)] σ −1 + (4π A0 [1 + α(t − 20)]) 2 g (12) which can be rewritten as a multi-variable equation f (X ) = 0 for which the roots (unknown pressure p) must be solved for. 11 / 30
  • 33. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Formulation of Mathematical Model as Nonlinear Root Search II. The form of f (X ) is then: n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (13) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] where the input state variable X is X = [ m1 , m2 , . . . , mn , (14) ρ1 , ρ2 , . . . , ρn , (15) T A0 , λ, αp , αc , t , Vs , H , Π, σ, g , tair , RHair ] (16) 12 / 30
  • 34. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Formulation of Mathematical Model as Nonlinear Root Search III. The density of the air may be calculated using the CIPM-2007 formula3 which expresses the air density in terms of ta – atmospheric temperature (measured with in-house thermometer traceable to water triple point (WTP) cell) pa – atmospheric pressure (measured with in-house barometer traceable to absolute medium gas pressures with Ruska LS) ha – atmospheric relative humidity expressed as a pure number such that 0 ha 1 (traceable to in-house hybrid humidity generator: validated in humidity regional inter-comparison) 3 Metrologia 45 2008 149–155 ”Revised Formula for the Density of Moist Air (CIPM-2007)”, Picard A et al 13 / 30
  • 35. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Details of CIPM-2007 Formula for Moist Air Density Briefly the underlying formula is of form pMa Mv ρa = 1 − xv 1− (17) ZRT Ma where ρa [kg m−3 ] – air density p [Pa] – atmospheric pressure (total pressure with implicit assumption of Dalton’s partial pressure law) t [◦ C] – air temperature T [K] – thermodynamic absolute temperature xv [−] – mole fraction of water vapour Ma [kg mol−1 ] – molar mass of dry air Mv [kg mol−1 ] – molar mass of water vapour Z [−] – compressibility factor R [J mol−1 K−1 ] – universal gas constant (using CODATA 2006 values4 cited in: Reviews of Modern Physics, Vol 80, April–June 2008, Mohr P J et al) 4 http://www.nist.gov/pml/div684/fcdc/upload/rmp2006-2.pdf 14 / 30
  • 36. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Note on Mass Input Values In the state variable we assume n mass pieces in total 15 / 30
  • 37. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Note on Mass Input Values In the state variable we assume n mass pieces in total This choice is incorporated by assigning the mass values to correspond to a weight set of N weights used (where n > N) and then appending the piston mass mpiston and bell mass mbell viz. [m1 , m2 , . . . , mn−2 , mn−1 , mn ]T = [m1 , m2 , . . . , mN , mpiston , mbell ]T (18) 15 / 30
  • 38. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Mechanism to Solve Formulation of Mathematical Equation Varying Meanings of Mass On the Differences Between True Mass, Conventional Mass and Apparent Mass Conventional mass: air density of 1.2 kg m−3 and mass density of 8000 kg m−3 Detailsa Objective is to determine the true mass of a weight piece from calibration data: ρa 1− ρs true mass : m = ms ρa (19) 1− ρm (c ) (s ) (c ) ρa ρa calibration cert data : ms 1− (c ) = ms 1− (20) ρs ρs (c ) ρa 1− (c ) 1− ρa (c ) ρs ρs ⇔m = ms (s) × (21) ρa 1− ρa 1− ρm ρs a The Pressure Balance: Theory and Practice, Dadson R S et al, 1982, National Physical Laboratory / HMSO 16 / 30
  • 39. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Solution for Roots of Multivariable Equation Solution of Roots of f (X ) = 0 In the mathematical model of the EURAMET.M.P-K13 key comparison there will in general be a variable number of inputs since a varying number of mass pieces n are used to generate a sequence of (applied) pressures for P ∈ [50, . . . , 500] MPa It can be shown that for n mass pieces used in X = [m1 , . . . , mn , ρ1 , . . . , ρn , A0 , λ, αp , αc , t , Vs , H , pa , ta , ha , σ, g ]T there will be a total of (2n + 12) state variables incorporated into the measurand model to solve for the unknown generated applied pressure Consider the terms in f (X ) which are known and unknown n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (22) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] 17 / 30
  • 40. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Defining an EoS for the Working Fluid EoS for a Pressure Balance Working Fluid I. During the intercomparison both the transfer standard and the NMISA Ruska high pressure PCU used di(2-ethylhexgl) sebecate i.e. ‘DHS’ as the working fluid formulated as ρf /[kg m−3 ] = [912.7 + 0.752(10−6 p) − 1.65 × 10−3 (10−6 p)2 + 1.5 × 10−6 (10−6 p)3 ] × [1 − 7.8 × 10−4 (tf − 20)] (23) Alternative forms that may be considered are linear extrapolations from a known reference state ρ f0 1+β(tf −tf 0 ) ρf = p −p (24) 1 − E f0 f where tf 0 is the reference oil temperature at which the oil’s properties are known (say 20 ◦ C), Ef is the bulk modulas fluid elasticity, pf 0 is the referance oil fluid density pressure, tf is the current oil temperature and p is the current oil pressure For pure ideal gases the EoS is ρRT p= (25) M 18 / 30
  • 41. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Defining an EoS for the Working Fluid EoS for a Pressure Balance Working Fluid II. On the Necessary Distinction Between Absolute and Applied Pressure The is a natural distinction between the absolute pressure p and the applied pressure P and equations-of-state are fundamentally defined in terms of the absolute pressure in order to be unambigous. Consider the following hypothetical cases: • case (i) {P (i ) = 200 kPa} ∪ {Π(i ) = 100 kPa} ⇒ p(i ) = 300 kPa • case (ii) {P (ii ) = 200 kPa} ∪ {Π(ii ) = 80 kPa} ⇒ p(ii ) = 280 kPa As a result even if the applied pressure is the same this does not neces- sary imply that the fluid density will be the same 19 / 30
  • 42. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Solution of the Roots (Pressure) of the Nonlinear Equation Numerical Solution for Roots of a Nonlinear Equation From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (26) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] 20 / 30
  • 43. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Solution of the Roots (Pressure) of the Nonlinear Equation Numerical Solution for Roots of a Nonlinear Equation From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (26) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] the original nonlinear equation may be transformed into an equation in a single unknown Y by setting all the all components such that 20 / 30
  • 44. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Solution of the Roots (Pressure) of the Nonlinear Equation Numerical Solution for Roots of a Nonlinear Equation From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (26) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] the original nonlinear equation may be transformed into an equation in a single unknown Y by setting all the all components such that xi = p∀i ∈ [1, . . . , 2n + 12] as parameters i.e. by solving 20 / 30
  • 45. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Review of Methods for Pressure Generation/Measurement Solution of the Roots (Pressure) of the Nonlinear Equation Numerical Solution for Roots of a Nonlinear Equation From the defining equation f (X ) with X = [x1 , . . . , x2n+12 ]T n ρa f (X ) = i =1 mi 1 − − Vs (ρf − ρa ) ρi + H (ρf − ρa )A0 [1 + λ(p − Π)][1 + α(t − 20)] −1 (26) + σ (4π A0 [1 + α(t − 20)]) 2 g A0 − g (p − Π)[1 + λ(p − Π)][1 + α(t − 20)] the original nonlinear equation may be transformed into an equation in a single unknown Y by setting all the all components such that xi = p∀i ∈ [1, . . . , 2n + 12] as parameters i.e. by solving f (Y ; X ) = 0 (27) The solution Y is of course the absolute pressure 20 / 30
  • 46. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Calculation of Generated Pressure Uncertainty Estimation of Generated Pressure Uncertainty For the EURAMET.M.P-K13 key comparison V&V in terms of the uncertainty analysis by the research metrologist was performed within the framework of the GUM 5 Cox M G and Harris P M, Software Support for Metrology Best Practice Guide No. 6 – Uncertainty Evaluation, Technical Report, National Physical Laboratory (United Kingdom), 2004 21 / 30
  • 47. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Calculation of Generated Pressure Uncertainty Estimation of Generated Pressure Uncertainty For the EURAMET.M.P-K13 key comparison V&V in terms of the uncertainty analysis by the research metrologist was performed within the framework of the GUM The underlying matrix based equation for the multi-variable input X was 2 ∂f u 2 (y ) = ( x f )V x ( x )T (28) ∂y following the methodology5 of Cox and Harris. In our initial approach I assumed no correlation effects in the components of the input variable X for simplicity 5 Cox M G and Harris P M, Software Support for Metrology Best Practice Guide No. 6 – Uncertainty Evaluation, Technical Report, National Physical Laboratory (United Kingdom), 2004 21 / 30
  • 48. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Calculation of Generated Pressure Uncertainty Specific Issues Encountered in Uncertainty Analysis The Validity of the Assumption of Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12 In general the only possible significant non-zero correlation effect that may occur is for the case Cov(A0 , λ) 6 consultation of the statistical literature such as Metrologia journal articles and IMEKO conferences states that it is poor practice to simply assume unity correlation in the absence of information: the recommended option is to assume zero correlation 22 / 30
  • 49. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Calculation of Generated Pressure Uncertainty Specific Issues Encountered in Uncertainty Analysis The Validity of the Assumption of Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12 In general the only possible significant non-zero correlation effect that may occur is for the case Cov(A0 , λ) Correlation effects do occur in the case of the weights mass/density but in general this has a very minor impact on final uncertainties as u (A0 ) and particularly u (λ) will dominate uncertainty contributions at elevated pressures 6 consultation of the statistical literature such as Metrologia journal articles and IMEKO conferences states that it is poor practice to simply assume unity correlation in the absence of information: the recommended option is to assume zero correlation 22 / 30
  • 50. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Calculation of Generated Pressure Uncertainty Specific Issues Encountered in Uncertainty Analysis The Validity of the Assumption of Cov(xi , xj ) = 0∀i = j : 1 i , j 2n + 12 In general the only possible significant non-zero correlation effect that may occur is for the case Cov(A0 , λ) Correlation effects do occur in the case of the weights mass/density but in general this has a very minor impact on final uncertainties as u (A0 ) and particularly u (λ) will dominate uncertainty contributions at elevated pressures (LS ) Correlation is certain to occur in the present instance as A0 and λ(LS ) are in fact obtained from an experimental cross-floating againsts another precision piston-cylinder standard however in the present instance no accessible information on Cov(A0 , λ) is available as these data values are sourced from an external certificate that does not provide any information from which one may reasonably infer a magnnitude of correlation6 6 consultation of the statistical literature such as Metrologia journal articles and IMEKO conferences states that it is poor practice to simply assume unity correlation in the absence of information: the recommended option is to assume zero correlation 22 / 30
  • 51. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating Make the assumption of linear elasticity theory in the PDE’s formulation of the strain/stress constituent behaviour: Assume a piston-cylinder A with known characteristics i.e. (A) DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with (B ) unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be determined 23 / 30
  • 52. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating Make the assumption of linear elasticity theory in the PDE’s formulation of the strain/stress constituent behaviour: Assume a piston-cylinder A with known characteristics i.e. (A) DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with (B ) unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be determined It follows that m 1 ρa PB = × g mB j 1− − Vs B (ρf − ρa )g SB ρB j j =1 (B ) + σB 4π A0 + (HB (ρf − ρa )g ) SB (29) 23 / 30
  • 53. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating Make the assumption of linear elasticity theory in the PDE’s formulation of the strain/stress constituent behaviour: Assume a piston-cylinder A with known characteristics i.e. (A) DA = {A0 , λ(A) } is cross-floated with another piston-cylinder B with (B ) unknown characteristics i.e. DB = {A0 , λ(B ) } which are to be determined It follows that m 1 ρa PB = × g mB j 1− − Vs B (ρf − ρa )g SB ρB j j =1 (B ) + σB 4π A0 + (HB (ρf − ρa )g ) SB (29) 23 / 30
  • 54. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating II. Noting that HB = 0 since the height correction was already accounted for in PCU A when working out the applied pressure 24 / 30
  • 55. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating II. Noting that HB = 0 since the height correction was already accounted for in PCU A when working out the applied pressure m 1 ρa (B ) ⇒ SB = × g mB j 1− − VB (ρf − ρa )g + σB 4π A0 PA ρm B j j =1 (30) 24 / 30
  • 56. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating II. Noting that HB = 0 since the height correction was already accounted for in PCU A when working out the applied pressure m 1 ρa (B ) ⇒ SB = × g mB j 1− − VB (ρf − ρa )g + σB 4π A0 PA ρm B j j =1 (30) As a result there is now a data point k with {P , SB } where (B ) SB = A0 (1 + λ(B ) P ) (31) 24 / 30
  • 57. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating II. Noting that HB = 0 since the height correction was already accounted for in PCU A when working out the applied pressure m 1 ρa (B ) ⇒ SB = × g mB j 1− − VB (ρf − ρa )g + σB 4π A0 PA ρm B j j =1 (30) As a result there is now a data point k with {P , SB } where (B ) SB = A0 (1 + λ(B ) P ) (31) for k ∈ [1, . . . , 10] corresponding to pressures [50, . . . , 500] MPa from which a straight line may be plotted in an analogous form to 24 / 30
  • 58. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Derivation of Basis Equation used in Cross-Floating I. Basis Equation for Cross-Floating II. Noting that HB = 0 since the height correction was already accounted for in PCU A when working out the applied pressure m 1 ρa (B ) ⇒ SB = × g mB j 1− − VB (ρf − ρa )g + σB 4π A0 PA ρm B j j =1 (30) As a result there is now a data point k with {P , SB } where (B ) SB = A0 (1 + λ(B ) P ) (31) for k ∈ [1, . . . , 10] corresponding to pressures [50, . . . , 500] MPa from which a straight line may be plotted in an analogous form to y = ax + b ⇐ D = {Dk = (xk , yk )|k ∈ [1, . . . , 10]} (32) 24 / 30
  • 59. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Estimation of Zero-Pressure and Distortion Coefficient Uncertainties I. Standard linear regression analysis in spreadsheet environments are not suitable for the provision of adequate parameter uncertainties 25 / 30
  • 60. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Estimation of Zero-Pressure and Distortion Coefficient Uncertainties I. Standard linear regression analysis in spreadsheet environments are not suitable for the provision of adequate parameter uncertainties In general optimization routines are necessary for a χ2 merit function determination and recently a practical implementation has been proposed in the scientific literature by scientists at the PTB (Germany): A Least Squares Algorithm for Fitting Data Points with Mutually Correlated Coordinates to a Straight Line, Meas. Sci. Technol. 22 (2011) doi: 10.1088/0957-0233/22/3/035101, Krystek M and Anton M 25 / 30
  • 61. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Estimation of Zero-Pressure and Distortion Coefficient Uncertainties I. Standard linear regression analysis in spreadsheet environments are not suitable for the provision of adequate parameter uncertainties In general optimization routines are necessary for a χ2 merit function determination and recently a practical implementation has been proposed in the scientific literature by scientists at the PTB (Germany): A Least Squares Algorithm for Fitting Data Points with Mutually Correlated Coordinates to a Straight Line, Meas. Sci. Technol. 22 (2011) doi: 10.1088/0957-0233/22/3/035101, Krystek M and Anton M 25 / 30
  • 62. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Estimation of Zero-Pressure and Distortion Coefficient Uncertainties II. The method fits y = ax + b and provides estimates for u (a) and u (b) noting that for a pressure balance the defining basis equation is A(P ) = A0 (1 + λP ) (33) 26 / 30
  • 63. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Methodology Used in Cross-Floating Analysis Estimation of the Straight Line Uncertainties Corresponding to u (A0 ) and u (λ) Estimation of Zero-Pressure and Distortion Coefficient Uncertainties II. The method fits y = ax + b and provides estimates for u (a) and u (b) noting that for a pressure balance the defining basis equation is A(P ) = A0 (1 + λP ) (33) It then follows that in a pressure cross-floating experimental measurement that u 2 (a) = u 2 (A0 ) (34) 2 2 2 u (b ) = λ u (A0 ) + A2 u 2 (λ) 0 (35) 1 ⇒ u 2 (λ) = u 2 (b) − λ2 u 2 (a) (36) A20 26 / 30
  • 64. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Specific Comments on Intercomparison Results Specific Comments on Intercomparison Participation The participation of the technical metrologists in EURAMET.M.P-K13 can be considered a learning experience as there is a significant difference is the level and complexity of key comparisons when compared to industrial level measurements 27 / 30
  • 65. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Specific Comments on Intercomparison Results Specific Comments on Intercomparison Participation The participation of the technical metrologists in EURAMET.M.P-K13 can be considered a learning experience as there is a significant difference is the level and complexity of key comparisons when compared to industrial level measurements It is recognized that in terms of data processing, analysis, and post-processing of results as a NMI there is always room for improvement and that there are certain inherent fundamental limitations in spreadsheet utilization, both in terms of what can be mathematically/statistically achieved and also in terms of validation and verification 27 / 30
  • 66. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Specific Comments on Intercomparison Results Specific Comments on Intercomparison Participation The participation of the technical metrologists in EURAMET.M.P-K13 can be considered a learning experience as there is a significant difference is the level and complexity of key comparisons when compared to industrial level measurements It is recognized that in terms of data processing, analysis, and post-processing of results as a NMI there is always room for improvement and that there are certain inherent fundamental limitations in spreadsheet utilization, both in terms of what can be mathematically/statistically achieved and also in terms of validation and verification In terms of internal comparison of processed results there is good general agreement in generated pressures and fair/moderate agreement in computed effective area and distortion coefficient of applicable transfer standard 27 / 30
  • 67. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Recommendations and Future Options Going Forward I. There is from the research metrologist’s perspective a need to more fully examine the ratio method of analysis in cross-floating as this may present a more numerically stable implementation than a more direct P vs. A(P ) regression analysis however this must be weighed up against the implication of directly introducing the LS’s distortion coefficient into the uncertainty analysis 28 / 30
  • 68. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Recommendations and Future Options Going Forward I. There is from the research metrologist’s perspective a need to more fully examine the ratio method of analysis in cross-floating as this may present a more numerically stable implementation than a more direct P vs. A(P ) regression analysis however this must be weighed up against the implication of directly introducing the LS’s distortion coefficient into the uncertainty analysis There is a need to more fully investigate the actual correlation between the parameter estimates for A0 and λ as opposed to a simplistic unity correlation that merely provides an upper bound 28 / 30
  • 69. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Recommendations and Future Options Going Forward II. FEM numerical simulations should be performed to benchmark the existing experimental/statistical distortion coefficient however the challenge posed is not actually in terms of the FEM analysis but rather in terms of the materials properties of the piston-cylinder: tungsten-carbide is prone to some variation in the Young’s modulas of elasticty and Poisson’s ratio – a possibility is to submit a research proposal for an electrical capacitance / materials characterization experimental study of the distortion coefficent on the low and medium pressure Ruska piston-cylinder laboratory standards 29 / 30
  • 70. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Recommendations and Future Options Going Forward II. FEM numerical simulations should be performed to benchmark the existing experimental/statistical distortion coefficient however the challenge posed is not actually in terms of the FEM analysis but rather in terms of the materials properties of the piston-cylinder: tungsten-carbide is prone to some variation in the Young’s modulas of elasticty and Poisson’s ratio – a possibility is to submit a research proposal for an electrical capacitance / materials characterization experimental study of the distortion coefficent on the low and medium pressure Ruska piston-cylinder laboratory standards It may be beneficial to consider alternative statistical formulations to bootstrapping such as Jackknifing 29 / 30
  • 71. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Recommendations and Future Options Going Forward II. FEM numerical simulations should be performed to benchmark the existing experimental/statistical distortion coefficient however the challenge posed is not actually in terms of the FEM analysis but rather in terms of the materials properties of the piston-cylinder: tungsten-carbide is prone to some variation in the Young’s modulas of elasticty and Poisson’s ratio – a possibility is to submit a research proposal for an electrical capacitance / materials characterization experimental study of the distortion coefficent on the low and medium pressure Ruska piston-cylinder laboratory standards It may be beneficial to consider alternative statistical formulations to bootstrapping such as Jackknifing Certain aspects of the Pressure & Vacuum Laboratory R&D on the Ruska high pressure characterization may be able to (in partnership with another NMI e.g. PTB / CMI) make a modest contribution on the Avogadro project that requires precision pressure measurements at 7 MPa 29 / 30
  • 72. Pressure & Vacuum Laboratory TAF 2011 Navigate to Table of Contents Discussion Recommendations Going Foward Acknowledgements This work was performed with financial support of the South African Department of Trade and Industry 30 / 30