Applying Radial Basis Function Networks to Fault Diagnosis of Motorized Spindle

Abstract

In a motorized spindle, due to the complexity of the system and nonlinear relationship between features and types of faults, it is difficult and inefficient to use traditional methods or physical models for the fault diagnosis. This paper focuses on the research on applying Radial Basis Function (RBF) Networks for fault detection and classification in the motorized spindle. As a data driven model with high efficiency, RBF networks has the advantage solving the nonlinear problems and dealing with the contradictory samples in the training process. In this research, the data, including rotating speed, temperature, and acceleration signals with three axes (X, Y and Z), are collected from a dynamic balancing platform to evaluate the working condition and detect the potential faults of the motorized spindle.