Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Block 25 Control Charts 13
1. Senior/Graduate
HMA Course
Quality control/quality assurance
Control Charts
Construction QC/QA Control Charts 1
2. Control
Charts
Construction QC/QA Control Charts 2
3. Variation
- Chance Causes
- Assignable Causes
Construction QC/QA Control Charts 3
4. Chance Causes
1. Everything varies
2. Individuals are unpredictable
3. Groups from a constant
system tend to be predictable
Construction QC/QA Control Charts 4 8-31
5. Example: Chance Causes
1. People live to different ages
2. No one knows how long he or
she will live
3. Insurance companies can
predict the percentage of people
who will live to certain ages
Construction QC/QA Control Charts 5
7. Benefits of Control Charts
- Early Detection of Trouble
- Decrease Variability
- Establish Process Capability
- Reduce Price Adjustment
Costs
- Decrease Inspection
Frequency
Construction QC/QA Control Charts 7
8. Benefits of Control Charts
• Basis for Altering Specification Limits
• Permanent Record of Quality
• Provide a Basis for Acceptance
• Instill Quality Awareness
Construction QC/QA Control Charts 8
9. Control Chart Analysis
AASHTO
“To Develop a
Quality Control/Quality
Assurance
Plan for Hot Mix Asphalt”
Construction QC/QA Control Charts 9
10. Upper control limit
Lower control limit
1 2 3 4 5 6 7
Lot Number
Construction QC/QA Control Charts 10
11. Statistical Control Chart
Upper Control Limit
Data Points
Target
Value
Lower Control Limit
Construction QC/QA Control Charts 11
12. Five lines on a statistical process
control chart
• Target value
• Warning upper control limit
• Warning lower control limit
• Action upper control limit
• Action lower control limit
Construction QC/QA Control Charts 12
13. Upper Control Limit
Mean
Lower Control Limit
1 2 3 4 5 6 7
Lot Number
Construction QC/QA Control Charts 13
14. Statistical Control Charts
Upper and Lower Warning
Control Limits
UWCL = X + { 2 (s) / (n)1/2 }
LWCL = X + { 2 (s) / (n)1/2 }
Construction QC/QA Control Charts 14
15. Statistical Control Charts
Upper and Lower Action
Control Limits
UWCL = X + { 3 (s) / (n)1/2 }
LWCL = X + { 3 (s) / (n)1/2 }
Construction QC/QA Control Charts 15
16. Example Problem
• Given:
– Target asphalt binder content: 5.7 %
– Standard deviation – 0.25 %
– The rolling average is based on five data
points
Construction QC/QA Control Charts 16
17. Example Problem
warning limits
UWCL = X + { 2 (s) / (n)1/2 }
UWCL = 5.7 + { 2 (.25) / (5)1/2 }
UWCL = 5.9
Construction QC/QA Control Charts 17
18. Example Problem
warning limits
LWCL = X - { 2 (s) / (n)1/2 }
LWCL = 5.7 - { 2 (.25) / (5)1/2 }
LWCL = 5.5
Construction QC/QA Control Charts 18
19. Example Problem
warning limits
UACL = X + { 3(s) / (n)1/2 }
UACL = 5.7 + { 3 (.25) / (5)1/2 }
UACL = 6.0
Construction QC/QA Control Charts 19
20. Example Problem
warning limits
LACL = X - { 3 (s) / (n)1/2 }
LACL = 5.7 - { 3 (.25) / (5)1/2 }
UWAL = 5.4
Construction QC/QA Control Charts 20
21. Look for Assignable Cause If:
• One point is outside of control limits
• Eight consecutive points are on one side of
the target value
Construction QC/QA Control Charts 21
46 Training Module III - QC/QA Concepts Of the many process control procedures that can be used, one of the most important is the use of control charts, particularly statistical control charts. Control charts provide a means of verifying that a process is in control. It is important to understand that control charts do not keep a process under control. Control charts simply provide a visual mechanism to identify when a contractor or producer should look for possible problems with the process.
47 Training Module III - QC/QA Concepts Variation of construction materials exists in all projects. In fact, the lack of variation should be looked upon with suspicion. The purpose of a control chart is not to eliminate variability, but to distinguish between inherent or chance causes of variability and a system of assignable causes. Assignable causes are factors that can be eliminated, thereby reducing overall variability.
48 Training Module III - QC/QA Concepts Duplicate measurements will not always be identical. Every process has some chance causes that cannot be eliminated, but may be reduced by changing the process. This variability prevents individual results from being accurately predicted. However, groups of results from a constant process or system do tend to be predictable.
As an example, it is not possible to predict how long an individual will live. However, insurance companies have actuarial tables that predict with relatively high accuracy what percentage of the population will live to various ages. The same thing can be done with a construction materials process, provided the process is in control. Chance causes are something a contractor/supplier must learn to live with. They cannot be eliminated, but their effects may be reduced.
However, assignable causes can be eliminated IF they can be identified, e.g., a hole in a hot bin screen, a fines feeder clogged or out of calibration, a weight scale needing recalibration, etc. Examples of assignable causes might be when the gradation goes out of specification due to a hole in a screen or because the cold feed conveyor setting is incorrectly adjusted.
50 Training Module III - QC/QA Concepts Control charts have been used for years in the manufacturing industry and have been successfully employed in construction materials applications. Some of the more important benefits of the use of control charts include early detection of trouble or identifying a change in the construction process and instilling an awareness of quality in the contractor's personnel.
Continuation of the list of benefits of control charts.
The ASSHTO Proposed Standard Practice “To Develop a Quality Control/Quality Assurance Plan for Hot Mix Asphalt” appendix A-II covers statistical control charts and was used as the primary guide in this part of the block.
51 Training Module III - QC/QA Concepts Physically, a control chart can be viewed as a distribution turned sideways with the vertical axis being the test results and the horizontal axis being successive test numbers. Typically, the charts are based on individual values, means, or ranges. The data can be assumed to fall within + 3 standard deviations of the mean or target when the process is in control. Statistical quality control charts for average or means rely on the fact that, for a normal distribution, essentially all of the values fall within + three standard deviations from the mean. A statistical quality control chart can be viewed as a normal distribution curve on its side. For a normal curve, only about 0.27% (1 out 370) of the measurements should fall outside of the 3 standard deviations. Therefore, control limits (indicating that an investigation for an assignable cause should be conducted) as set at + 3 standard deviations.
43 Training Module VI - Superpave Quality Control (QC) Plan: Part 1 23 The Run Chart is a start at process control. However, Statistical Control Charts are better because they can be used to distinguish chance causes from assignable causes.
Target value – the value of the property from the declared JMF. Warning upper control limit – target value plus two standard deviations divided by the square root of the number of samples in the moving average Warning lower control limit - target value minus two standard deviations divided by the square root of the number of samples in the moving average Action upper control limit – target value plus three standard deviations divided by the square root of the number of samples in the moving average Action lower control limit - target value plus three standard deviations divided by the square root of the number of samples in the moving average
55 Training Module III - QC/QA Concepts Statistical control charts are always based on the mean and range (or ) of a subgroup of size n > 1. There are several reasons for this. One of the main reasons is that the distribution of sample means tends to be normally distributed. Therefore, even if the underlying population from which the samples are taken is not normal, the distribution of sample means will be approximately normal. This allows the use of 3 x control limits to identify when the process is out of control. Secondly, the use of n > 1 is necessary to allow the calculation of ranges for the R charts.
To compute the upper warning control limit (UWCL) follow these steps.
To compute the lower warning control limit (LWCL) follow these steps.
To compute the upper action control limit (UACL) follow these steps.
To compute the lower action control limit (LACL) follow these steps.
57 Training Module III - QC/QA Concepts Statistical control charts are used to determine when an assignable cause is acting to change the process. For example, action is taken (defined as attempting to identify the assignable cause of the change in the mean or variability) when one point falls outside the control limits. Another cause for action is the occurrence of 8 points in a row on either side of the target mean (i.e., the grand mean for X-bar charts or the average range for range charts).
The next question to be asked is : How does one interpret the control charts to identify whether the process is out-of-control or in-control? Remember statistical control charts are used to determine when an assignable cause is acting to change the production/construction process.
A control chart for the sample mean () is shown on the visual aid to the right.
A control chart for the Range (R) is shown on the visual aid to the right.
What does one mean by “lack of control?” A process can be out of control (or lack of control) in one of three ways: ! The process mean changes while the process standard deviation remains constant. ! The process standard deviation changes while the process mean remains constant. ! Both the process mean and standard deviation change.
The process standard deviation can remain constant while the process mean can change or shift in several ways. These include: A sustained sudden shift in the mean A trend in the mean An irregular shift in the mean A sustained sudden shift in the mean. This could be indicative of a situation where the aggregate supplier for an HMA mix is changed during the project or one of the cold-feed bin feeders is out of calibration or has malfunctioned. Another example might be a sudden shift in the percent compaction of the HMA mat caused by an increase in the moisture content of the aggregate stockpiles that was not accounted for during the production process.
A trend in mean. This could be indicative of a progressive change brought on by wear of the paver and/or selected components of the HMA production facility.
An irregular shift in mean. This could be indicative of the operator making continuous, but unnecessary adjustments to the process settings. For example, continuously changing the thickness control handle, or crank, on a paver or improper loading and unloading of the HMA transport vehicles resulting in random segregation.
In addition, the range (or standard deviation) can change in several ways while the mean remains constant. This is related to an increase in R and includes: ! A sudden change in the range ! A gradual change in the range A sudden change in range. This could be indicative of a situation where the aggregate source or supplier for an HMA mix is changed during the project or where the operator for the breakdown roller is changed (and this person has not been trained properly for the compaction operation).
A gradual change in range. This could be indicative of a progressive change, such as machine wear, as noted in the above example for the mean.
A process also could exhibit irregular shifts in both the mean and range.
There are different rules for interpreting control charts. Two rules that are used most commonly are: One point outside the UCL or LCL and Eight consecutive points on one side of the target value. Either one or both of the above cases should trigger a cause for action by the producer (supplier) or contractor. Many producers/contractors also use warning limits in addition to the action limits (UCL and LCL). The warning limits generally are plus or minus two standard deviations from the target value. A number of other criteria are presented and discussed in the list of references.