Strength Evaluation and Prediction of Cement Concrete by Deep Learning Classification Using Non-destructive Test Results

Concrete strength assessment through NDE has opened very widen prospects for the implementation of Non-destructive techniques and evaluation (NDT&E) approach through static and dynamic approach. As strength gain in concrete is complex phenomena, hence, NDT approaches are more suitable to acquire early-stage strength measurement to allow the structure to impose with various constructional loads and early removal of form work. With advent of wave propagation approach in many NDT approaches, the concrete strength monitoring is more feasible. Also, this allows to store, analyze and predict strength values obtained through various inspection carried out in different timeline of construction. In similar line the Deep learning has recently become popular in structural alternation due to its excellent feature learning capabilities. This paper aims studied the concrete compressive strength through NDE approaches (Rebound hammer and Ultrasonic pulse velocity meter) for lab sized concrete cube for 28 days. Later, the same experimental results are post processed using deep learning models like convolutional neural network (CNN), long short-term memory (LSTM) and CNN-LSTM hybrid model. The goodness of fit has measures through mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE) value for all the deep learning models. Through evaluation metrics values, it is seen that the proposed model shows better prediction compared to that of CNN and LSTM model. Overall, proposed research highlights the potential of hybrid deep learning model in accurately assessing concrete strength.