Load Identification Method for Spindle Rotor System of Rolling Mill Based on Fusion Information

Aiming at the limitations of single type response signal in load identification, this paper proposes a fusion information method that combines vibration signals and motor current signals to identify the load type of the spindle rotor system in rolling mill. That is, for different types of load excitation, a qualitative load identification method for the spindle rotor system of rolling mill based on Bayes estimation is proposed. Specifically, after converting the vibration and motor current response information into the frequency domain, the “singular value decomposition—wavelet packet analysis” method is used for signal preprocessing. Bayes estimation method in feature level fusion is used to calculate the probability of the two types of signal samples after preprocessing, and then the load type of the spindle rotor system in rolling mill is identified. According to the load characteristics and test conditions, a test plan was developed and loading tests were conducted on a dedicated load identification bedstand. The results show that under different types of loads, after singular value decomposition and wavelet packet energy feature extraction of the two signals, the fusion information method based on naive Bayes can successfully identify the load types of the spindle rotor system, and the identification accuracy is significantly improved compared to that based on a single type of signal. Therefore, the method proposed is feasible. Due to reliable and experimentally validated theoretical results, this method can be applied to load identification and fault diagnosis of spindle rotor system of rolling mill in engineering.