The research is aiming at the fault diagnosis of HAGC system of strip rolling mill. Taking the advantage of the accompany characteristics of the closed-loop control system, rolling force forecasting model is established based on neural networks. The comparison results between the forecasting results and the actual signal are taken as residual signals, which have eliminated the effects of abrupt input. Wavelet transform is used to acquire the components of high and low frequency, and fault numerical criterion is established through Lipschitz exponent. Wavelet decomposition results make fault feature clear and time-domain positioning accurately. By analyzing the varied fault features which correspond to varied fault reasons, the fault diagnosis of HAGC system is implemented successfully.

The research is aiming at the fault diagnosis of HAGC system of strip rolling mill. Taking the advantage of the accompany characteristics of the closed-loop control system, rolling force forecasting model is established based on neural networks. The comparison results between the forecasting results and the actual signal are taken as residual signals, which have eliminated the effects of abrupt input. Wavelet transform is used to acquire the components of high and low frequency, and fault numerical criterion is established through Lipschitz exponent. Wavelet decomposition results make fault feature clear and time-domain positioning accurately. By analyzing the varied fault features which correspond to varied fault reasons, the fault diagnosis of HAGC system is implemented successfully.