The objective of this study is to develop a comprehensive framework that integrate probabilistic failure modeling techniques such as Bayesian network and fault tree to evaluate failure dependency, linguistic fuzzy assessment to incorporate the likelihood and impact of human errors on system failure. The proposed framework also seeks to facilitate better decision-making in maintenance strategies by offering insights into both technical and human-related failure mechanisms. This holistic approach will ultimately aim to improve system performance and resilience in the face of potential failures.
Maintenance is crucial for the efficient functioning of complex systems, as it addresses faults, failures, system degradation, and their consequences, including costs, reliability, availability, and performance. The study began by identifying critical components, followed by modeling the failure and repair rate distribution graphs. It utilized fault tree analysis for failure modeling, incorporating real-time data and the application of fuzzy set theory. The Fussel–Vesely measure was employed to identify critical elements from a reliability standpoint. Furthermore, a Bayesian network was used to analyze the probabilistic dependencies among failure causes.
The result obtained from fault tree failure modeling analysis revealed that the machine currently exhibits a failure rate of 0.64 per day, reliability of 0.53, and availability of 0.40. The result of the Bayesian network shows the failure probability of the machine is 0.44 per day and the comparison reveals that failure analysis for the system using the Bayesian network is more precise than using the fault tree, which decreases the error of failure probability from 49% to 44% and have 10.2% of reduction. The ability to more accurately represent the likelihood of failure not only aids in operational efficiency but also supports better decision-making processes in managing system reliability and availability.
This study makes a distinct contribution by filling a significant gap in the literature using a Bayesian network (BN) to model the failure behavior of air jet looms machine using integrated quantitative failure/repair data and qualitative expert insights on human factors. Current textile maintenance studies often rely on conventional techniques that overlook this crucial integration. This study provides the first evidence-based framework for BN resilience in capturing the intricate, human-influenced failure mechanisms inherent in the textile industry. This fills a gap in the literature and establishes a new standard for reliability analysis in the industry.
