Abnormal state identification for T beam based on novelty detection

The cracking state and abnormal positions are recognized during T beam model tests using the BP neural network based novelty detection technology. Neural network training sample data is generated by analyzing the static load test data, the neural network model based on novelty detection technology is established, the state of the T beam anomaly recognition and crack position recognition is accomplished. Stepwise partition method is used in crack position recognition, which includes narrowing the crack position as the first step, specifically analyzing the sensor data, and determination of the crack position. T beam neural network model is verified by the measured data. The results show that, the method can accurately identify state and effectively identify the location of the crack.