摘要
针对传统故障诊断方法准确性不高、耗时长问题,研究通过多个EfficientNet模型对传感器数据进行预训练,并使用XGBoost作为元学习器,提出了一种基于Blending集成学习的多源信息液压系统多类故障诊断方法。实验结果表明,各个子分类器在训练次数达到300次后趋于收敛,准确率均达到95%左右。该方法具有较高的准确性和鲁棒性,为液压系统故障诊断提供了一种有效的解决方案。
In response to the issues of low accuracy and long time consumption in traditional fault diagnosis methods,a multi source information hydraulic system multi class fault diagnosis method based on Blending ensemble learning was proposed by pre training sensor data using multiple EfficientNet models and using XGBoost as a meta learner.The experimental results showed that each sub classifier tended to converge after training 300 times,with an accuracy rate of around 95%.This method has high accuracy and robustness,providing an effective solution for hydraulic system fault diagnosis.
作者
杜凯乐
朱为全
陈瑞宝
刘丽珊
Du Kaile;Zhu Weiquan;Chen Ruibao;Liu Lishan(CNOOC Energy Development Oil Production Service Branch,Tianjin 300456,CHN)
出处
《模具制造》
2024年第2期229-231,共3页
Die & Mould Manufacture
