Abstract
Bayesian Network (BN) models are being successfully applied to improve fault diagnosis, which in turn can improve equipment uptime and customer service. Most of these BN models are essentially trained using quantitative data obtained from sensors. However, sensors may not be able to cover all faults and therefore such BN models would be incomplete. Furthermore, many systems have maintenance logs that can serve as qualitative data, potentially containing historic causation information in unstructured natural language replete with technical terms. The motivation of this paper is to leverage all of the data available to improve BN learning. Specifically, we propose a method for fusion-learning of BNs: for quantitative data obtained from sensors, metrology data and qualitative data from maintenance logs, corrective and preventive action reports, and then follow by fusing these two BNs. Furthermore, we propose a human-in-the-loop approach for expert knowledge elicitation of the BN structure aided by logged natural language data instead of relying exclusively on their anecdotal memory. The resulting fused BN model can be expected to provide improved diagnostics as it has a wider fault coverage than the individual BNs. We demonstrate the efficacy of our proposed method using real world data from uninterruptible power supply (UPS) fault diagnostics.
Highlights
The productivity and sustainability of manufacturing and service industries depend largely on prompt identification of fault root cause(s)
Interpretability—Clearly identifies relationship between variables; Root Cause Analysis—Can model the cause and effect relationship together with the causation path in case of indirect causation; Model Uncertainty—Since it is based on probability theory, it can readily handle uncertainty, which is inherent in fault diagnosis; Compact Representation—Directed acyclic graphs (DAG) are used to represent variables that influence each other along with causal direction
The contribution of this paper is twofold: (1) we propose a human-in-the-loop natural language processing approach for expert knowledge elicitation of the Bayesian Network (BN) structure aided by logged natural language data instead of relying on exclusively their anecdotal memory; (2) we propose a combined modeling algorithm for training BN from both qualitative and quantitative data sources that result in improved fault root cause identification when compared with BN trained using only one data source
Summary
The productivity and sustainability of manufacturing and service industries depend largely on prompt identification of fault root cause(s). Since these systems are complex and machine breakdowns are inevitable, fast and accurate root cause analysis (RCA). A BN model represents the factorization of the joint probability distribution among a set of random variables. It can be denoted as a set ( G, θ ) where G is the directed graph that depicts the dependencies between the variables and is made up of the vertex or node set V and the edge or arc set E. An indirect path between two nodes via another node such as X j → Xi → Xh , means that Xh is a descendant of X j , and X j is an ancestor of Xh .
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
