A bibliometric literature review of integrated data and model based diagnosis approaches for the industry 4.0

The increasing presence of Cyber-Physical Systems and the Internet of Things has accelerated the digital transformation of industrial environments, commonly known as Industry 4.0. In this context, Artificial Intelligence techniques are increasingly used to support automatic diagnostic tasks. This paper presents a systematic literature review of hybrid diagnostic systems that combine Model-Based Diagnosis (MBD), which relies on physical models to detect abnormal behaviour, and Data-Based Diagnosis (DBD), which uses machine learning to identify faults from data. The review has two objectives: (i) to examine how MBD and DBD methods have been combined to improve diagnostic performance, and (ii) to identify integration opportunities through existing machine learning frameworks to support reusable and adaptive solutions. A bibliometric analysis was conducted following a simplified PRISMA 2020 methodology. From over 1300 records, 75 articles were selected and analysed. Most hybrid systems adopt a serial architecture where DBD classifiers analyze residuals from MBD for fault detection and isolation. The most common applications are found in smart manufacturing and energy systems, and as for the most used machine learning techniques. Challenges remain regarding real-time scalability, interpretability, and standardisation. This review provides a structured foundation for designing explainable, efficient, and reusable diagnostic solutions for Industry 4.0. Highlights The literature review of hybrid data-based diagnosis (DBD) and model-based diagnosis (MBD) systems in Industry 4.0 reveals key trends and promising future directions. Hybrid diagnostic systems facilitate online monitoring, improve explainability, and enable intelligent feature selection. The most widely used hybrid architecture for integrating MBD and DBD techniques is serial integration. Serial integration uses MBD for residual generation, and DBDs for fault isolation and detection. There is growing interest in integrating hybrid models with explainability and online learning. The most common applications are in the automotive and energy sectors, in predictive maintenance and smart monitoring. Current hybrid diagnostic systems lack real-time scalability and lack interpretability and reusability.