Multi-fault identification of iron oxide gas sensor based on CNN-wavelelet-based network

Because of its stability, hydrogen semiconductor sensor has become one of the most important means widely used in hydrogen monitoring. Sensor fault identification is of great significance. Traditional fault diagnosis methods, such as machine learning, are complicated in operation and network structure. In this paper, an improved method based on traditional CNN network is proposed for fault detection of hydrogen iron oxide sensors. Different from the previous method, it is not necessary to filter the data first, and the data will be converted into two-dimensional gray image as the input of the neural network. The excitation function of CNN network is optimized and combined with the wavelet analysis method, the wavelet basis function is used as the excitation function to detect and identify the fault. At the same time, different from the previous fault diagnosis methods, this paper has achieved good results in multi-fault detection and identification, and improved the accuracy of multi-fault diagnosis while avoiding the complex network structure.