Comparative Analysis of Classical Machine Learning Models for Predictive Maintenance in the Brewing Industry

The article investigates the application of classical Machine Learning (ML) Algorithms for enhanced equipment for reliability and failure prognosis within complex industrial systems, specifically focusing on the brewing industry. The objective is to develop a Robust Machine Learning framework to anticipate equipment breakdowns and ensure optimal operational performance accurately. This methodology encompasses a dual approach, first leveraging comprehensive simulation data sets derived from a demanding manufacturing environment to identify operational irregularities and predict potential equipment failures. Second, undertake a rigorous comparative analysis of various supervised learning models, including Decision Trees, Random Forests, Logistic Regression, Support Vector Machines (SVM), and K-nearest neighbors. These models were implemented using Python and evaluated meticulously through metrics such as confusion matrices, classification reports, ROC curves, and stratified cross-validation. Results indicate that the Random Forest model demonstrates superior overall performance for binary classification in this context. This comparative assessment provides critical insights for selecting and implementing the most effective predictive maintenance strategies, aiming to optimize brewing operations. Future work will concentrate on refining the identified high-performing models and exploring class-specific performance metrics to mitigate further the costs associated with false positives and false negatives.