CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
Cigos_carbonation.pptx
1. EMERGING TECHNOLOGIES
AND APPLICATIONS
FOR GREEN INFRASTRUCTURE
Using Artificial Neural Network Containing Two
Hidden Layers for Predicting Carbonation
Depth of Concrete
Van Quan Tran
University of Transport Technology
Hanoi, 28 October 2021
11. Context
Approach and
Method
Results and
Discussion
Conclusion
k
2
i i
2 i=1
k
2
i
i=1
v - v
R = 1-
v - v
(2)
k
2
i i
i=1
1
RMSE = v - v
k
(3)
N
i i
j=1
1
MAE = v - v
N
(4)
Machine learning algorithm: Artificial Neural Network
Model performance
12. Context
Approach and
Method
Results and
Discussion
Conclusion
Data analysis
Description of database
Unit count mean std min average max
Binder
kg/
m3 300.00 369.68 53.93 260.00 350.00 500.00
Fly ash % 300.00 21.44 20.56 0.00 17.50 70.00
Water/Binder - 300.00 0.45 0.09 0.28 0.45 0.63
[CO2] % 300.00 9.08 11.65 1.00 6.50 50.00
RH % 300.00 64.25 3.66 55.00 65.00 70.00
t^0.5 days 300.00 7.60 4.41 1.73 6.48 19.08
x mm 300.00 9.05 9.34 0.00 6.00 57.00
14. Context
Approach and
Method
Results and
Discussion
Conclusion
The ability of machine learning technique to predict the carbonation depth of
concrete was shown.
To save time and money for conducting experiments, an ANN model one of the
most powerful algorithms was developed
The carbonation depth of concrete has been predicted by ANN models with
network structure [6-12-8-1].
The results indicated that ANN models performed well for predicting the
carbonation depth of concrete with higher performance such as R2, RMSE and
MAE value to be equal to 0.9536, 1.7761, 1.0325, respectively
A performance comparison between ANN model and different machine learning
models such as tree model, random forest, support vector machine seems to be
interesting to perform in the future.