Sensor Fault Diagnosis Based on Multi Generator Countermeasure Network

With the rapid development of integrated and intelligent technology, modern large-scale aviation is equipped with thousands of sensors, which can measure all kinds of aircraft status in real time to ensure the stable operation of important systems such as aircraft operation stability, avionics and power. Once the sensor fails, it will cause serious flight safety and economic losses and casualties. Therefore, effective and accurate sensor fault diagnosis methods are called research hotspots. Data driven fault diagnosis method has rapidly become the mainstream of current application and research. However, due to the complex working environment of aviation equipment and the lack of fault data with labels, the existing fault diagnosis methods are difficult to achieve high-precision fault diagnosis. In this paper, a sensor fault diagnosis method based on multi generator countermeasure network is proposed. By replacing the original single generator structure with multi generator structure, different generators are used to capture different high probability modes. When the discriminator is alternately trained, the corresponding class labels are assigned to different generators to encourage them to generate corresponding types of generated samples, which effectively improves the feature extraction ability and generalization ability of the model and solves the problem of pattern collapse. The experimental results show that the model can capture more high probability modes and generate high-quality samples, and has better generalization ability and higher diagnosis accuracy in the face of fault data under different working conditions.