.png){kind=icon}
.png){kind=icon}
Return
Information Integration-Based Factor Object Approach for Object Classification Judgment in the System Fault Evolution Process
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jii.2025.100967.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Industrial system fault information is a fusion of fault-related data, where objects represent core fault events and factors quantify their dynamic states. Factor variations directly reflect object characteristics, driving the System Fault Evolution Process (SFEP)—a complex progression of system functionality from normal operation to failure, shaped by temporal changes in object states and factor values. To assess how factors influence object classification during SFEP, this paper proposes the Object Classification Judgment Method Integration-Based Factor-Object (OCJM-IFO), a novel approach rooted in the Neighborhood Preserving Embedding (NPE) algorithm. OCJM-IFO addresses critical limitations of existing methods: it handles sparse data, avoids the curse of dimensionality, and reduces reliance on prior rules by dynamically fusing weights from labelled (intra-class) and unlabelled (inter-class) data via an optimal weight ratio coefficient. This fusion enables comprehensive evaluation of factor impacts. Experiments on electrical systems and MOSFET faults (each involving 6 factors and 100 objects) validate the method: it identifies sets of favorable, uncertain, and unfavorable factors, with results aligning closely with physical fault characteristics. The algorithm requires a data structure composed of time-series objects, supporting real-time dataset updates. Thus, it is particularly well-suited for intelligent real-time monitoring systems in industrial environments, offering universal applicability and easy data accessibility. The construction process of the OCJM-IFO dataset is presented. This study strengthens fault information integration in industrial systems, providing a robust tool for fault diagnosis and preventive maintenance, with proven engineering applicability in enhancing system reliability.
Journal
IF:
11.6
Papers:
893
Citations:
4.4K