Deep learning prediction and evaluation of stability for piezoresistive behavior of the cement sensors

Cement-based sensors are exposed to continuous dynamic loading and/or damage, degrading their sensing stability, nevertheless, predictions of long-term sensing stability have rarely been reported. Therefore, this study presents a deep-learning analysis combining experimental data and a LSTM model to predict the stability of long-term piezoresistivity. Related experiments are conducted, and the test results are used as training data. The simulations indicate that the parameters of the LSTM model have a notable effect on the predicted long-term piezoresistive sensing performances of the composites. By comparing the predictions with the experimental results, the validity of the proposed deep-learning approach is evaluated, and the following conclusions can be drawn from this study.