基于改进优化算法的轧机滚动轴承深度学习故障诊断方法

Deep Learning-Based Fault Diagnosis Method for Rolling Bearing of Rolling Mill With Improved Optimization Algorithm

近年来,基于深度学习的智能故障诊断方法在理论研究和工程应用方面都取得了显著的效果,滚动轴承是轧机的核心部件之一,对轧机轴承进行精确的故障诊断能够有效保障轧机装备安全运行与生产效率.当前基于深度学习的故障诊断方法通常训练过程不稳定,模型难以收敛,造成工程应用中随机性强.本文提出基于改进优化算法的轧机滚动轴承深度故障诊断方法,在保证模型诊断精度的同时提升训练效率、模型输出结果的稳定性以及模型相对于参数变化所表现出的鲁棒性,并通过实验台获取滚动轴承的故障数据,使用该方法对数据进行诊断来证明方法的准确性.

In the recent years,the intelligent fault diagnosis methods based on deep learning have achieved remarkable results in both theoretical research and engineering applications.Rolling bearing is one of the core components in the rolling mill.Accurate fault diagnosis of rolling mill bearing can effectively ensure the safe operation and production efficiency of the rolling mill equipment.The current fault diagnosis methods based on deep learning usually have unstable training process and poor model convergence,resulting in significant randomness in engineering applications.This paper proposes a fault diagnosis method based on the improved deep learning optimization algorithm for rolling bearings in the rolling mill.This method not only guarantees the accuracy of fault diagnosis,but also improves the model training efficiency,stability of the output results and the model robustness against parameter selections.The faulty data of rolling bearings are obtained in the test bench.The effectiveness of the model is proved by validations on the real-world data.