Automated monitoring of deep excavation-induced horizontal displacements using sparse sensor measurements: A recurrent neural network approach

With the increasing complexity of excavation projects, the demand for automated monitoring systems has grown due to their ability to provide continuous, real-time data capture. However, traditional methods of using automatic inclinometers are cost-prohibitive due to the dense configuration of sensors required for accurate data collection. This study introduces a novel recurrent neural network (RNN) framework that leverages sparse sensor data to predict excavation-induced horizontal displacements in automated monitoring. The predictive performance of three RNN variants—Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM)—was assessed and compared against traditional machine learning models, including Backpropagation Neural Network (BPNN) and Support Vector Regression (SVR), using field data from two deep excavation sites in Hangzhou, China. The BiLSTM model achieved the best performance, with a mean absolute error (MAE) of 1.90 mm and a coefficient of determination (R2) of 0.89, outperforming both other RNN variants and traditional ML models. The model’s robustness was validated across different soil conditions and excavation strategies. The study also shows that optimized sensor deployment can reduce the number of sensors needed while maintaining high prediction accuracy, offering a cost-effective solution for real-time monitoring in construction projects.