A Novel Error Compensation Framework for MEMS Sensor Array Applied to Seabed Terrain Deformation Monitoring

Micro-electromechanical systems (MEMS) sensor array plays a significant role in in situ and long-term monitoring of seabed terrain deformation. However, multisource errors in the monitoring process are difficult to avoid, such as natural disaster, communication issues, and calculation models. A novel error compensation framework based on a MEMS sensor array is presented and validated in this article. An elaborate laboratory test is carried out, and two datasets with an interval of 5 mm between 0 and 150 mm are collected using two MEMS sensor arrays. The correlation coefficient (CC) is introduced and applied for constructing optimal model input features. The window size, sliding step, differential coefficient, differential step, and lag length are determined by CC analysis and 12 model input features are identified. Seven models are comprehensively compared with two datasets, including 1-D convolutional neural network (CNN)-bidirectional long short-term memory (BiLSTM), BiLSTM, 1D-CNN, gated recurrent unit (GRU), long short-term memory (LSTM), linear regression (LR), and support vector regression (SVR). The proposed hybrid 1D-CNN-BiLSTM can achieve the lowest average RMSE (2.9677 mm; 2.3319 mm) and the highest average <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}^{{2}}$ </tex-math></inline-formula> (0.9269; 0.9356) on two datasets with k-fold cross validation. Also, it outperforms other models in the compensation experiment, with RMSE reduction of up to 34.46% and 31.00%, respectively. The results demonstrate that the proposed error compensation framework can effectively compensate deformation monitoring errors and provide practical guidelines for seabed terrain monitoring and deformation error compensation.