Physics-Based Precursor Wiring Diagnostics for Shielded-Twisted-Pair Cable

Abstract

The capability to locate and characterize precursor wiring faults, such as chafing or pinching potentially enables preventive maintenance well before hard failures occur, thus maximizing system functionality and safety while minimizing out-of-service time. Toward this goal, results are presented on the application of a deterministic Bayesian inference procedure well suited for detecting chafing and pinch faults through the use of a newly developed physics-based model for shielded-twisted-pair cable. This method is significantly faster than more traditional nondeterministic Bayesian methods, such as Markov chain Monte Carlo, and retains many of the desirable features inherent to the Bayesian approach. These include the ability to quantify estimation uncertainty and model evidence in probabilistic terms, which then enables the study and design of noise-tolerant fault detection algorithms capable of classifying different types of faults. The fault parameter estimation results from both laboratory and field measurements on a C17 jet engine are shown to demonstrate the achievable model fidelity and the overall viability of the approach.