Applications of fault diagnosis in rotating machinery by using time series analysis with neural network

The common diagnosis method of time series analysis is an autoregressive (AR) method, which is a kind of math model that can be established by time difference and vibration amplitude. As the AR model utilized the math method for fitting the variable, the AR coefficients represent the signal features and can be used to determine fault types. This study proposed the difference values of AR coefficients, which indicated that the AR coefficients of ideal signal for normal machine are deducted from faulty machines. It is convention that the relationship between the difference values of AR coefficients and fault types as trained by using back-propagation neural network (BPNN). The new fault diagnosis method by using the difference of AR coefficients with BPNN was proposed in this study. The diagnosis results were obtained and compared with the three methods, which include the difference of AR coefficients with BPNN, the AR coefficients with BPNN and the distance of AR coefficients method for 23 samples. And the diagnosis results obtained by using the difference of AR coefficients with BPNN were superior to AR coefficients with BPNN and distance of AR coefficients methods.