基于可扩展置信规则库的工业设备故障诊断方法  

Fault Diagnosis Method for Industrial Equipment Based on Extended Belief Rule Base

作　　者：冯树成 贺维 马宁 周国辉[1] FENG Shucheng;HE Wei;MA Ning;ZHOU Guohui(College of Computer Science and Information Engineering,Harbin Normal University,Harbin 150025,China)

机构地区：[1]哈尔滨师范大学计算机科学与信息工程学院,黑龙江哈尔滨150025

出　　处：《火箭军工程大学学报》2025年第2期13-21,共9页Journal of Rocket Force University of Engineering

基　　金：火箭军工程大学重点实验室开放基金(AAIE-2023-0102)。

摘　　要：针对工业设备故障诊断中常面临的样本量大、参数繁杂等问题,提出了一种新型可扩展置信规则库模型(Extended Belief Rule Base,EBRB),该模型具有良好的可解释性和合理性,便于技术人员根据模型输出进行可靠的决策。首先,运用极致梯度提升方法对大量特征进行重要性评估,选取最具诊断价值的特征,有效降低问题维度;然后,在推理阶段,引入投影协方差矩阵自适应策略优化方法进行参数优化,从而提升模型性能;最后,在输出阶段,采用基于阈值的方式限制激活规则数量,并使用改进的证据推理引擎,以提高推理效率和准确性。结果表明:提出的EBRB在多个故障诊断任务中表现优异,在滚动体故障诊断中,准确率达98.15%,与K最邻近模型并列第一;在内圈故障诊断中,准确率为90.74%,优于所有其他对比模型;在外圈故障诊断中,准确率为88.89%,同样高于其他对比模型。实验结果验证了该模型在处理复杂工业设备故障诊断任务中的高准确性和鲁棒性,为工业设备故障诊断领域提供了一种新的有效解决方案。To address the problems of large sample sizes and complex parameters in industrial equipment fault diagnosis,a novel EBRB(Extended Belief Rule Base)model was proposed.This model had excellent interpretability and rationality,making it easy for technical personnel to make reliable decisions based on model outputs.First,the model employed the eXtreme Gradient Boosting method to assess the importance of numerous features and selected the most diagnostically-relevant ones,effectively reducing the dimensionality of the problem.Then,in the inference stage,the projection covariance matrix adaptive evolutionary strategy method was introduced to optimize the parameters,thereby improving the model’s performance.Finally,in the output stage,a threshold-based approach was adopted to limit the number of activated rules,and an improved evidential reasoning engine was utilized to enhance inference efficiency and accuracy.The experimental results indicate that the proposed EBRB model performs exceptionally well in multiple fault diagnosis tasks.In the fault diagnosis of rolling elements,it achieves an accuracy of 98.15%,tying for first place with the K-nearest neighbors model.In the fault diagnosis of the inner circles,the model attains an accuracy of 90.74%,surpassing all the other comparative models.In the fault diagnosis of the outer circles,the model achieves an accuracy of 88.89%,which exceeds that of other comparative models.The results verify the high accuracy and robustness of the model in complex industrial equipment fault diagnosis tasks,thereby providing a new effective solution in the field of industrial equipment fault diagnosis.

关 键 词：可扩展置信规则库 故障诊断 特征选择 参数优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]