Multiattribute Modeling for Oil Condition Assessment Considering Uncertainties

Lubricating oil carries the primary wear failure information of the critical tribological components and, therefore, severs for the condition-based maintenance of equipment. However, the oil condition assessment (OCA) presents low reliability due to the uncertainties originating from the variable working conditions and the redundant indicators. To address the uncertainties with multiple indicators, a knowledge-guided three-layer model is established for characterizing the multiattribute oil state. Furthermore, data dispersion is considered by assigning fuzzy probability among the attribute layer. The inconsistent decisions are solved by improved evidential reasoning (ER) embedding inference rules in the state layer. The effectiveness of the proposed approach is verified using the real-world lubricant oil monitoring data from vehicle engines.

Automated summary

This study establishes a knowledge-guided three-layer model to characterize the multiattribute oil state. Data dispersion is considered by assigning fuzzy probability among the attribute layer. Improved evidential reasoning is used to solve inconsistent decisions. The effectiveness of the proposed approach is verified using real-world monitoring data.