显微磨粒图像识别知识规则提取及应用

Knowledge Rules Extraction and Application of Micro Debris Image Recognition

针对新研制的多功能油液磨粒智能检测系统MIDCS中的磨粒图像识别问题,引入数据挖掘方法获取了磨粒图像识别的知识规则,实现对磨粒类别的智能识别。利用MIDCS系统获取实际航空发动机运行过程中由于滚动轴承磨损而产生的大量典型磨粒,基于图像分析方法提取16个磨粒特征参数,形成标准案例库;利用Weka软件的决策树算法自动提取知识规则,并对知识规则进行优化和简化;对所提取得到的知识规则进行验证和分析。结果表明,所提取的磨粒识别规则符合磨粒识别的统计规律,识别规则不仅简洁,而且具有很高的精度。基于Weka软件的规则提取方法大大提高了MIDCS系统的磨粒识别自动化和智能化水平,对于利用MIDCS进行航空发动机滚动轴承疲劳磨损故障诊断,具有重要的工程实用价值。

Aimed at the wear particle recognition problem of the new Multiple Intelligent Debris Classifying System（ MIDCS）, data mining method was introduced in order to obtain the knowledge rules of wear particle recognition, and theexpert system theory was used to realize the intelligent recognition of debris classes. A large number of typical debriscaused by rolling bearing wear in the actual aero-engine operational process was obtained by MIDCS, 16 debris character-istic parameters were extracted based on the image analysis method, and the standard case library was formed. The deci-sion tree algorithm of the Weka software was used for automatic extraction of the knowledge rules, and the knowledge ruleswere opti-mized and simplified. The extracted knowledge rules were verified and analyzed. The results show that the rulesagree well with the wear particles recognition statistical laws, the extracted rules is very brief and correct, the extractionmethod based on Weka software can be used in the debris class recognition of MIDCS well, and the automation and intelli-gent level of MIDCS debris class recognition are greatly improved. It is of significant engineering value for the aero-enginerolling bearing fatigue wear fault diagnosis by using MIDCS.