基于CNN-LSTM的机械手臂关节内部轴承故障诊断方法OA

To address the issues of strong subjectivity,significant lag,and low accuracy in traditional fault diagnosis methods for robotic joint bearings,a fault diagnosis method based on the fusion of a Convolutional Neural Network(CNN)and a Long Short-Term Memory(LSTM)network was proposed.Vibration signals were collected using piezoelectric acceleration sensors.After undergoing window segmentation,wavelet denoising,and standardized preprocessing,the signals were fed into a dual-branch parallel network.Local impact features and temporal dynamic features were extracted separately and then fused through a connection layer.Hyperparameters were optimized using a chaos algorithm,ultimately enabling fault classification.Experimental results showed that the system achieved an accuracy of 98.70％with a processing time of 0.012 s per sample.Its performance was superior compared to single models and unoptimized fusion models,and it demonstrated better real-time performance than the 2D-CNN method.This approach effectively enables high-precision,real-time fault diagnosis for robotic joint bearings,offering valuable reference for ensuring the continuity of industrial production.