基于证据推理规则的船舶柴油机磨损类型辨识研究  

Identification on Wear Mode for Marine Diesel Engine Based on Evidential Reasoning Rule

作　　者：徐晓健 严新平[1,2,3] 盛晨兴 袁成清[1,2,3] 

机构地区：[1]武汉理工大学能源与动力工程学院可靠性工程研究所,湖北武汉430063 [2]船舶动力工程技术交通行业重点实验室,湖北武汉430063 [3]国家水运安全工程技术研究中心,湖北武汉430063

出　　处：《摩擦学学报》2017年第6期814-822,共9页Tribology

基　　金：国家自然科学基金项目(51422507);NSFC-浙江两化融合联合基金项目(U170920062)资助~~

摘　　要：为了利用不确定、不完整信息对船舶柴油机磨损故障进行诊断,以磨粒的二维和三维特征作为诊断信息,提出了基于证据推理(ER)规则的船舶柴油机磨粒类型辨识模型,并以此作为识别柴油机磨损类型的依据.在船舶柴油机油底壳附近安装油液监测系统以采集油液样本,提取油液中磨粒的二维和三维特征.对每一磨粒特征进行k均值聚类确定其参考值,对磨粒样本与参考值的相似性进行似然归一化,获得了证据的置信度分布.考虑辨识证据的可靠性和重要性,利用ER融合规则对多条辨识证据进行融合,辨识磨粒类型,进而确定柴油机磨损类型.通过5折交叉验证以及不同模型对比分析表明:该模型能够利用不确定、不完整信息识别磨粒类型,确定柴油机磨损类型,并且模型结构简单,辨识准确性高,辨识结果客观、可靠.To diagnose wear faults of marine diesel engines with uncertain and incomplete information, an evidential reasoning（ER） rule-based model was proposed to identify wear particles of marine diesel engines by using two dimensional（2-D） and three dimensional（3-D） characteristics of particles, which was further used as the basis of identifying wear modes of diesel engines. An oil monitoring system was installed near an engine oil sump to collect oil samples, and the 2-D and 3-D characteristics of wear particles were extracted from the oil samples. Referential points of every characteristic were determined by k-means clustering, and the belief degree distribution of each piece of evidence was acquired by calculating samples＇ similarity distribution about referential points and normalizing the likelihoods. ERrule was used to combine multiple pieces of evidence,of which reliability and importance were considered.Five-fold cross-validation and various models comparison indicated that the ER rule-based model can well identify wear particles from marine diesel engines with uncertain and incomplete information,and then the wear mode of marine diesel engines can be determined with the particle type given by the model.The model proved to be simple,accurate,objective and reliable.

关 键 词：船舶柴油机 磨损类型辨识 证据推理规则 证据可靠性 证据重要性 

分 类 号：TH117.1[机械工程—机械设计及理论]