Operational Pattern Mining in Industrial Maintenance Using the SPADE Algorithm
Abstract
Operational pattern mining plays a vital role in industrial maintenance by identifying recurring sequences of machine behavior, faults, and maintenance activities to enhance system reliability. The increasing complexity of industrial systems and the large volume of sensor and maintenance log data create challenges for effective pattern discovery. Existing methods often rely on statistical analysis or manual expert rules, which are unable to efficiently capture hidden sequential dependencies and temporal relationships in large datasets, leading to incomplete or inaccurate predictive insights. This study proposes the SPADE-Based Sequential Pattern Mining Framework (SSPMF), a datadriven approach that leverages the SPADE algorithm to automatically extract frequent operational and fault sequences from historical maintenance and sensor data. The framework transforms raw data into structured sequences, applies SPADE for pattern discovery, and enables predictive maintenance decision support. The proposed method helps maintenance teams predict faults and schedule preventive actions in advance. Experimental results demonstrate that SSPMF effectively uncovers significant patterns, improving fault prediction accuracy and reducing unplanned downtime.