基于神经网络的多种油样分析技术融合诊断

Neural-Network-Based Fusion Diagnosis Using Various Oil Analysis Techniques

针对4种最常用的油样分析技术(铁谱、光谱、颗粒计数及理化指标分析)的信息融合问题,依据基于规则的专家系统方法,建立了各油样分析技术的子专家诊断系统;依据专家经验建立各子专家系统的诊断结果与故障论域中各故障模式的关系,得到了用于神经网络学习的训练样本.在此基础上,通过对神经网络进行训练,并将待分析油样的各子专家系统诊断结论输入训练成功的网络,即得到融合诊断结果.实例分析表明所建立的分析方法便捷有效.

With a view to the data fusion of four oil analysis techniques, i.e., ferrographic analysis,spectrometric analysis, particle counting analysis, and oil quality testing, the subexpert system for each oil analysis technique was established in connection with the expert system (ES) methods based on knowledge rules.The relationship between each subES diagnosis result and the final fault pattern was established according to expert experiences, and the trained samples were obtained and used in the neural network learning. Thus the fusion diagnosis results were obtained by successful training of the neural network and inputting of each subES diagnosis result into the neural network.The analysis of a typical oil sample indicated that the neuralnetworkbased fusion diagnosis method was applicable to the analysis of oil samples and hence to the fault pattern identification with good feasibility and convenience.