Prediction of Compressive Strength of Concrete with a Skewed Pattern- Application of Artificial Neural Network Approach

A. Rahman[1a], M. Hadiuzzaman[2], Ajit K. Majumder[3], M. N. Uddin[4]
Page No. 70-80

Concrete has a versatile use in the construction practice for its availability, cheap rate, and
flexibility of handling. As a result, in the construction process it is always important to measure the
concrete compressive strength as strength properties of cement paste mixture. A smart modeling
approach Artificial Neural Networks(ANNs) have recently been introduced as an efficient and
powerful modeling technique for applications involving a large number of variables, especially with
highly non-linear and complex interactions among input/output variables. In this paper, an
artificial neural network of the feed-forward back-propagation type has been applied as a data
treatment technique. The 28-day compressive strength values are considered as the aim of the
prediction. A total of 269 specimens are selected. The system is trained and validated using
188(70%) pairs chosen randomly from the data set and tested using the remaining 81(30%) pairs.
Results indicate that models performed quite well in predicting the compressive strength in case of
training dataset and also for independent data set.
Keywords : Concrete compressive strength, Artificial Neural Network (ANN), feed-forward back
propagation, skewed distribution.