Research on Collision Detection Method Based on Feature Extraction

Research the collision detection problem during equipment operation. Collect data on collisions occurring at different positions during normal operation and operation of the equipment, process the signals using wavelet transform, extract feature indicators such as kurtosis for analysis, and finally use Support Vector Machine (SVM) classification method to distinguish collisions and their specific locations. The research results show that using the One-Versus-Rest (OVR) classification method with a lifting table as the experimental object, the accuracy of the data reaches 96.11%, the classification time is 0.021, and it has high detection and discrimination rates. Moreover, this method can effectively detect collisions during equipment operation based on the collected data and determine the specific location of the collision.