Hybrid Knowledge-Based and Data-Driven Explainable AI Architectures for Intelligent Fault Diagnosis in Cyber-Physical Systems

Abstract

The proposed study will design a hybrid, knowledge-based and data-driven, explainable artificial intelligence (XAI) architecture for intelligent fault diagnosis in cyber-physical systems (CPS), aiming to enhance diagnostic accuracy, interpretability, and reliability in complex industrial settings. The approach presented, combining machine learning models with rule-based reasoning and explainability mechanisms, is proposed in a hybrid framework. The sensor data measured on CPS components is preprocessed and features extracted, then fault classification is performed on data-driven models developed with CNN, LSTM, and Random Forest. Model predictions are authenticated and improved by a knowledge-based module that uses expert rules, and explainability methods, such as SHAP, are used to understand the effects of input features on diagnostic results. The experimental data show that the offered hybrid XAI model outperforms traditional methods across all evaluation metrics. The model has an accuracy of 96.3%, which is higher by 94.1% and 91.2%, respectively, compared to LSTM, CNN, and Random Forest. On the same note, the hybrid method yields a precision of 95.4%, a recall of 94.8%, and an F1-score of 95.1%, which indicates better classification balance and lower false positives and false negatives. These statistical gains confirm that the combination of domain knowledge and data-driven learning increases predictive performance and strength. To sum up, the suggested hybrid architecture is an effective solution in that it satisfies the requirements of the two fields: accuracy and understanding in CPS fault diagnosis. The framework not only enhances the performance of diagnostic processes but also provides clear, operational guidance to system operators through explainable AI and knowledge-based reasoning. This both qualifies it as a good solution for safety-critical industrial systems and offers strong prospects for future development in intelligent maintenance systems.