多传感信息融合在液压故障诊断中的应用

Multi-sensor Information Fusion in Hydraulic System Failure Diagnosis

针对恶劣工作环境下多传感信息融合识别效果差和D-S证据理论中证据难获取的问题,在组建有效的传感器网络的基础上,结合改进的JDL模型并根据数据融合分级处理思想,数据层采用自适应加权最小平方估计法对数据进行清洗和特征提取,特征层通过多并行PSO-Hopfield网络的联想记忆功能进行局部诊断,决策层根据修正的D-S证据理论进行时空域融合,并且每级和最终诊断之间都有直接数据通信和反馈,使得知识库信息能为数据挖掘进行知识发现作必要的数据储备。通过仿真结果可知:该数据融合系统容错性强、能综合利用传感器信息并准确定位故障。

A modified multi-sensor information fision method for hydraulic fault diagnosing system is proposed in this paper. Combining with the improved JDL data fusion model and the hierarchical processing idea, it can solve some difficult fault diagnosis problems of hydraulic system. The adaptive weighted least squares estimation method is used to clean the data and extract the feature in data layer. The multi-parallel PSO （Particle swarm optimization）-Hopfield neural network is applied in feature level for local diagnosis. When the time-airspace integration, there is a direct data communication and feedback between each level based on modified D-S （Dempster-Shafer） evidence theory in decision-making level. The final diagnosis has a direct data communication and feedback between each level, and it can makes the information of each level based on data mining as soon as possible. Experimental results show that the method in conflicted evidence has high correct rate and can avoid index explosion and fixed the fault exactly.