A Fault Diagnosis Method for Planetary Gearboxes Based on IFMD

Abstract

The vibration signal from the planetary gearbox exhibits nonlinear and impulsive characteristics amidst strong noise, impeding the effective extraction of fault information and compromising the accuracy of fault diagnosis. To address this challenge, a fault diagnosis method rooted in feature mode decomposition (FMD) is proposed. Initially, the critical parameters (modal number n and filter length L) of FMD are optimized using an improved genetic algorithm (IGA), and the refined FMD is employed to decompose the vibration signals from the planetary gearbox. Subsequently, a convolutional neural network integrated with the support vector machine model (CNN-SVM) is established, leveraging the convolutional neural network for feature extraction. Ultimately, SVM iteratively optimized by the particle swarm optimization (PSO) algorithm, serves as the classification technique. Simulation and experiment results demonstrate the effectiveness of this method in extracting and identifying fault information within planetary gearboxes.