Reconstructing wind pressure measurement data by neural network technique

Structural health monitoring (SHM) technique is increasingly used in civil engineering structures, from which the authentic environmental and structural conditions can be directly obtained. To get accurate structural condition assessment and damage detection, it is important to make sure the monitoring system is robust and the sensors are working properly. When sensor fault occurs, data cannot be correctly obtained at the broken sensor(s). In such situation, approaches are needed to help reconstruct the missing data. This paper presents an investigation on wind pressure monitoring of the Canton Tower of 600 m high during the Typhoon Kai-Tak, aiming to formulate a neural network (NN)-based method for reconstructing the measurement data at sensor points. After identifying the mean value of wind pressure during the typhoon, an NN model with one hidden layer and one output layer is formulated by using the measurement data. Data of 30-min time period are selected within the typhoon period and are used to train the model and test its capability. With the NN model, wind pressure data at a sensor location can be reproduced and predicted by using the measured wind pressure data from the adjacent sensor locations. The reconstructed data set has root mean square errors less than 20 Pa and correlation coefficients larger than 0.8 in comparison with the field monitoring data. Results show that the monitoring data can be satisfactorily reconstructed by using the NN-based technique.