基于信息熵与判断矩阵的D-S证据理论改进方法在故障诊断中的应用

Application of an Improved Dempster-Shafer Evidence Theory Based on Information Entropy and Evaluation Matrix to Fault Diagnosis

针对凭借单一的证据很难进行科学故障诊断的实际情况,提出基于Dempster-Shafer证据理论进行多源信息融合的故障诊断方法.由于D-S证据理论合成公式存在不足,针对各证据的重要性不同,提出了一种将信息熵与判断矩阵一致性相结合对各证据进行加权赋值,并对其进行加权调整的方法,应用D-S证据理论合成公式对加权调整后的证据进行决策级融合.此种方法将主观因素与客观因素相结合.实例分析结果表明:融合结果提高了置信度,降低了不确定性,显著提高了故障诊断的科学性,具有理论与应用价值.

It is difficult to obtain scientific fault diagnosis based on single evidence. A multi-information fusion approach based on Dempster-Shafer evidence theory was put forward; however, the combination rules of D-S evidence theory is unreasonable to solve some combination. A new type of method was proposed to weight evidences combined information entropy with consistence of evaluation matrix in terms of different importance of evidences. Evidences were weighted and well adjusted. Evidences were fused based on combination formula of D-S evidence theory. Subjective and objective factors were well- considered during the process. Fusion results show that the new method has better performance in dealing with combinations, improves reliability, and decreases uncertainty of evidences. Diagnostic results were improved evidently. This new method had theoretical and practical application values.