Sensor Location Selection and Response Prediction Based on Sparse Regularization and Linear Regression

The unknown vibration has a significant negative effect on the engineering structure health. Therefore, it is necessary to predict the vibration response of the structure accurately. However, the existing vibration response prediction models have some problems, such as too much freedom, difficult sensor optimization and low prediction accuracy. To overcome these shortcomings, we propose a method based on Independently Interpretable Lasso and linear regression (IILasso-LR) to optimize sensor layout for response prediction. IILasso-LR more aggressively induces the sparsity of the active variables and reduces the correlations among them. Hence, we can independently interpret the effects of the selected sensor on the response prediction. In addition, the optimized sensor is used for response prediction, which greatly improves the efficiency of prediction. Experiments of finite element model is used to validate the effectiveness and accuracy of IILasso-LR. Effects of sensor location, number of sensors, excitation type and noise level are studied in detail. The results show that the IILasso-LR could predict vibration response effectively and satisfy industrial requirements.