A Transferable Autologistic Model for Predicting Rare Failures in Heterogeneous Equipment

2026-08-06

AI Summary

This study addresses the challenges of predictive maintenance in heterogeneous devices, where sensor configurations vary significantly, failures are sparse, and early warnings are essential. To tackle these issues, the authors propose a transferable self-logistic regression model that leverages probabilistic graphical modeling and context-aware feature alignment to learn shared failure patterns from source devices and efficiently adapt to target devices with minimal overhead. The approach effectively integrates sensor heterogeneity, operational context, and degradation dynamics to produce well-calibrated fault probabilities. By uniquely combining transferable modeling with self-logistic regression, the method balances strong generalization capabilities with high adaptation efficiency. Experimental validation on a simulated refrigerator dataset comprising 27 units with diverse sensor setups demonstrates the model’s ability to achieve highly accurate early fault prediction.

Abstract

Predicting failures before they occur remains a major challenge in predictive maintenance, particularly when failures are rare, when equipment of the same family differ in sensor configurations, and when the goal is anticipation rather than diagnosis of an already observed fault. This paper proposes a common-to-target probabilistic model that learns shared failure-related patterns across a family of heterogeneous equipment and adapts parsimoniously to target equipment. The model explicitly accounts for sensor heterogeneity, operating context, and degradation dynamics to produce calibrated failureprobability estimates suitable for maintenance planning. Its performance is evaluated on a synthetic refrigerator dataset comprising 27 simulated refrigerators with varying sensor configurations, operating conditions, and failure types, providing a controlle

Problem

• predictive maintenance
• rare failures
• heterogeneous equipment
• sensor heterogeneity
• failure prediction

Innovation

• Transferable model
• Autologistic regression
• Rare failure prediction
• Sensor heterogeneity
• Predictive maintenance