Structural health monitoring using artificial neural network and statistical process control in a changing environment

A methodology for structural health monitoring using Artificial Neural Network (ANN) and Statistical Process Control (SPC) in a changing environment is presented in the paper. In many cases, the changes in the measured features due to the environmental factors can be much larger than those caused by structural damage. In real structures, the variations of measured features due to changing environmental factors must be taken consideration. This paper demonstrates a novel solution to the problem via the system model parameterised by environmental variables with Artificial Neural Network. First, the ANN model was established by using of the response measurements and environmental parameters from a normal structure. A technique of Principal Components Analysis (PCA) was adopted to pre-process the input data, which removes the correlation of the input variables. Next, damage-sensitive features or structural responses were predicted by ANN model with subsequent monitoring data. A Monte-Carlo simulation was used to construct the thresholds for the predicted values of the system ANN model. Final, the measured features or responses were monitored to the thresholds. The occurrence of measured data outside the thresholds was considered to damage or anomalous structural state. The effectiveness of the combined method based on ANN and SPC was validated with data from Fengman Dam in China. The results demonstrate that the proposed method can effectively monitor the civil structure in a changing environment.