Development of a Decision Support System for Automated Anomaly Detection : Shifting from Condition-Based to Predictive Maintenance

Sammanfattning

In today’s rapidly evolving industrial landscape, ensuring smooth operations while min-imizing downtime is essential for maintaining efficiency and competitiveness. Modern industrial systems generate vast amounts of data from sensors, event logs, and other monitoring tools, providing valuable insights into equipment performance and poten-tial failures. The growing adoption of data-driven strategies has enabled the shift from traditional, reactive maintenance to predictive maintenance (PdM), leveraging machine learning (ML), artificial intelligence (AI), and automation to anticipate and prevent breakdowns. This thesis focuses on enhancing maintenance decision-making through the development of an advanced DSS. The DSS is designed to classify detected issues into two categories: those that can be resolved by operators and those requiring inter-vention from specialized technicians or system engineers. By providing structured guid-ance and actionable recommendations, the system aims to improve response times, op-timize maintenance workflows, and reduce unplanned downtime. Results from the study demonstrate that the DSS can reduces maintenance delays by streamlining issue classification and resolution processes. Initial evaluations indicate improved operational efficiency, reduced technician workload, and better resource allocation. The accuracy of the DSS depends heavily on the quality and completeness of the event log data. Inconsistent or insufficient data can lead to misclassification of issues, affecting the system’s reliability. Furthermore, the system’s effectiveness may vary across different industrial environments, requiring customization to adapt to specific workflows and machine types. Future work can extend the system by integrating forecasting tech-niques. By utilizing historical data and frequency analysis, predictive models can be de-veloped to anticipate failures before they occur, further enhancing maintenance effi-ciency. The findings contribute to the broader field of industrial maintenance by demonstrating how AI-driven decision support can streamline troubleshooting pro-cesses and reduce unplanned downtime. Finally, while automation improves efficiency, operator training and user adoption are crucial to ensuring the DSS is used correctly and effectively in real-world scenarios.