基于改进D-S信息融合方法的发动机故障诊断

Fault Diagnosis of Engine Based on Improved Dempster-Shafer Information Fusion Method

为提高发动机故障诊断的精确度和准确性,将信息融合技术应用于发动机非电控单元的故障诊断。针对经典D-S理论方法所存在的对高度冲突证据融合失效的问题,提出一种基于D-S证据理论的改进信息融合方法:首先将不同类型的故障诊断模型给出的诊断结果组成证据组,计算证据组各证据之间冲突的程度;利用证据冲突系数对证据组进行分类,计算各高冲突证据的重要度,应用证据间距离函数来判定各证据的可信度;根据重要度和可信度来修正证据的权重,并通过Dempster组合规则实现故障诊断信息融合。通过理论分析和发动机实车试验研究验证了新方法的有效性,与经典D-S组合方法和改进合成规则公式法比较,新组合方法可提高故障诊断的准确率和确定度。

In order to enhance the accuracy of fault diagnosis for engine, information fusion technology was applied for engine mechanical units. An improved information fusion method based on D-S evidence theory was proposed. The evidence group was composed of fault diagnosis results obtained from different types of diagnosis model, the conflict degree between evidences in the evidence group was calculated, the evidence group was classified by evidence conflict coefficient, the importance of each highly conflict evidence was calculated, and the credibility of each evidence was determined with a distance function of evidence bodies. Then the weight value of each evidence was revised with its importance and credibility respectively. Finally, the Dempster combination rule was used to realize the information fusion. The effectiveness of the new approach was verified by theoretical analysis and experiment research. Comparing with the classical D-S combination method and the improved synthesis formula, the new combination method is more efficient in improving the accuracy and the certainty degree of engine fault diagnosis.