Knowledge Discovery and Diagnosis Using Temporal-Association-Rule-Mining-Based Approach for Threshing Cylinder Blockage

Accurately diagnosing blockages in a threshing cylinder is crucial for ensuring efficiency and quality combine harvester operations. However, terms of blockage diagnostic methods, the current state affairs characterized by model-based approaches that can be highly time-consuming difficult to implement, while data-driven lack interpretability. To address this situation, we propose temporal association rule mining (TARM)-based fault diagnosis method identifying discovering knowledge. This study performs field trials varying actual feed rate obtains datasets three classes (slight, moderate, severe). Firstly, symbolic aggregate approximation (SAX) employed reduce data dimensionality construct transaction set with sliding window. Next, cSpade used mine extract strong rules applying improved support, confidence, lift indicators. With established rules, comprehensively elucidate variation pattern each characteristic under several failure conditions effectively identify faults. The results demonstrate proposed distinguishes between levels faults, achieving an overall accuracy 0.94. And yields precisions 0.90, 0.92, 0.99 corresponding recalls 0.93, 0.98 slight, medium, severe respectively. Specifically, knowledge acquired from extracted explain operational characteristics when its cylinders are blocked. Furthermore, approach provide reasonable reliable reference future research on blockages.