Graph Structural Residuals: A Learning Approach to Diagnosis

TL;DR

This paper introduces a data-driven diagnosis framework that combines graph structure learning with traditional model-based methods, enabling dynamic system analysis without explicit modeling.

Contribution

It redefines system representation and introduces two self-supervised graph learning models to improve diagnosis capabilities.

Findings

Effective diagnosis on coupled oscillators

Seamless integration of graph learning with diagnosis

Potential to reduce modeling labor

Abstract

Traditional model-based diagnosis relies on constructing explicit system models, a process that can be laborious and expertise-demanding. In this paper, we propose a novel framework that combines concepts of model-based diagnosis with deep graph structure learning. This data-driven approach leverages data to learn the system's underlying structure and provide dynamic observations, represented by two distinct graph adjacency matrices. Our work facilitates a seamless integration of graph structure learning with model-based diagnosis by making three main contributions: (i) redefining the constructs of system representation, observations, and faults (ii) introducing two distinct versions of a self-supervised graph structure learning model architecture and (iii) demonstrating the potential of our data-driven diagnostic method through experiments on a system of coupled oscillators.