Sensor fault diagnosis method based on fractal dimension

The development of automation technology has opened a new field - sensor fault diagnosis. Sensor is the source and foundation of information, once the sensor faults occur, which will lead to the degradation of automation system's performance, even disastrous consequences. There are various sensor fault diagnosis methods, while each method has its imperfection, so this paper has proposed to use fractal dimension for sensor fault diagnosis. Fractal theory is a powerful tool to deal with nature fragmented or complex phenomenon, which can describe the non-linear behaviour of the signal quantitatively. Firstly, this paper indicates the necessity and practical utility of sensor fault diagnosis and the development status of fractal theory. Then it describes the correlation dimension in detail which is the most suitable one to analyze sequenced dynamic signal, thereafter analyzing the type of the sensor fault. Finally, in order to verify the effectiveness of the method proposed, we set up an experimental platform. Five kinds of signals (no-fault signal and four kinds of fault signals) are gathered, whose correlation dimensions are calculated. The experimental results show that: various sensor faults can be diagnosed effectively by correlation dimension. At last, in order to verify the correctness of the correlation dimensions' results, the paper also gives the box counting dimension results, which further verifies the correctness of the results.