ADVANCED PREDICTIVE MAINTENANCE USING CATBOOST CLASSIFIER FOR FAULT DETECTION IN INDUSTRIAL EQUIPMENT

Abstract

In the realm of industrial automation, predictive maintenance has emerged as a pivotal strategy to ensure operational efficiency, minimize equipment downtime, and reduce maintenance costs. This study presents an advanced predictive maintenance framework aimed at the early detection of mechanical faults in industrial equipment using machine learning techniques. The primary objective is to classify the operational status of machinery as either Normal or Fault based on sensor data and operational features. The process begins with data preprocessing, including missing value treatment, label encoding, and visual analytics such as class distribution plots, heatmaps, and count plots. These steps provided key insights into feature relevance, data imbalance, and potential anomalies, effectively guiding the modelling process. Comparative analysis revealed that while both Ridge and Cat Boost classifiers perform effectively, the Categorical Boosting (Cat Boost) classifier offers distinct advantages in handling complex categorical data, enhancing model interpretability, and improving training efficiency—making it a strong candidate for deployment in real-time industrial environments. The existing system, which employs a Ridge Classifier, demonstrated strong classification capabilities with a confusion matrix indicating 70,376 true positives, 188,119 true negatives, 4,636 false negatives, and 3,033 false positives. Although overall accuracy was high, the number of false negatives raised concerns due to the critical nature of undetected faults in industrial settings. To address this limitation, a proposed system was developed using the Cat Boost Classifier— a gradient boosting algorithm optimized for categorical feature handling and high-performance learning. The Cat Boost model achieved comparable results, with 70,936 true positives, 188,119 true negatives, and identical false negatives (4,636) and false positives (3,033) to the Ridge Classifier, indicating robust predictive power and consistent classification reliability.