Multi-linear Regression Model to Estimate Temperature Signal in FAST Project Active Reflective Surface Health Monitoring System

The temperature measurement point malfunction in the large-scale structural health monitoring system could cause safety hazards. The paper extracts 9 temperature measurement points' data of the FAST project active reflective surface health monitoring system and analyzes the linear correlation of them. In order to establish a multiple linear regression model, the variables are grouped and the optimal subset is selected. The fault measuring point is estimated by fusing the normal measuring points' data. Focusing to the multiple collinear issue between variables, the ridge regression method is applied and the ridge parameter is set to 6. Using F-test and fit-degree, the significance and validity of models are tested. Based on different time data, the estimation accuracy is verified. The results show the multiple linear regression model has a higher degree of fit and accuracy than one variable model, considering its RMSE is 0.475 ℃. Besides, the ridge regression model is more stable, considering its RMSE is 0.435 ℃.