When increasing the control system gain for a hot rolling mill unit

abnormal vibration fault may occur.For solving this problem

firstly vibration testing and test signal spectrum analysis of the relevant components were carried out

and the abnormal characteristics of vibration signals were obtained.Least squares method was used to identify the hot rolling mill abnormal vibration model on the basis of foregoing analysis.Least squares of the ARX fault models were built and model testing were carried out and model effectiveness was also verified and then the stability of the fault models were analyzed.The result shows that: the normal operating state of the rolling system is stable

but the pole of the system when abnormal vibration occurs is very close to the virtual axis.The most obvious difference between them is the difference of pure delay about the model structure

and the pure delay destroys the stability of the system

forming a self-excited vibration of the mill.At the same time

it also verifies the feasibility and superiority of least squares algorithm in solving large-scale and complex mechanical equipments fault diagnosis problems.