Explainable Hierarchical Self-Supervised Learning Framework for Intelligent Fault Discovery

Fault detection and diagnosis in complex industrial systems, such as slurry pumps used in mining, is impractical with lack of labeled data. However, most of the existing studies rely on labelled data and the black-box nature of machine learning or deep learning approaches. While some studies have focused on model-specific explanations, domain-specific explanations are still missing, which hinders actionable maintenance decisions. To address those challenges, this study focuses on an Explainable Hierarchical Self-Supervised Learning Framework for intelligent fault discovery and it has three stages: (1) a self-supervised module that generates high-confidence ground-truth labels by fusing vibration sensor and process data through autoencoders, Isolation Forest, and Gaussian Mixture Models with cross-modality consensus, (2) a supervised classification stage enhanced with SHAP-based explainability to ensure transparent and accurate fault detection, and (3) a rule-based reasoning layer that categorises fault severity (Tolerate, Moderate, Severe) and infers likely root causes using domain knowledge and shallow decision trees. An experimental study is evaluated on real-world slurry pump data, the framework successfully generated 19,369 healthy and 654 anomalous samples as a reliable ground truth. The supervised models achieved exceptional performance, with 99.95% accuracy and perfect recall (100%). The severity categorisation identified 58.6% cases as tolerable, 27.8% as severe, and 13.6% as moderate, while fault diagnosis revealed cavitation or impeller damage as the most prevalent issue (46%). This approach bridges data-driven learning and expert knowledge, enabling trustworthy, interpretable, and actionable fault diagnosis in safety-critical industrial environments.