Analysis and prediction of failures in heavy machinery using logistic regression models

Abstract

This study analyzes the failure behavior of heavy machinery used in mining operations by a company that provides services to the extractive sector. These failures result in unplanned downtime, increasing costs and negatively impacting operational efficiency. The primary objective is to identify trends and patterns in failures by analyzing historical data recorded between January 2019 and July 2024, to optimize corrective maintenance management. The methodology involved a descriptive statistical analysis of failures reported in the SAP system, focusing on four critical subsystems: engine, brake, implement, and electrical system. Subsequently, a logistic regression model was applied to estimate the probability of failure occurrence, incorporating variables such as hourometer readings, temperature, and operational conditions. The results show that failure frequency increases with accumulated equipment usage, particularly in the electrical and implement systems. Usage thresholds were identified, beyond which the failure rate rises significantly. These findings contribute to improved planning of maintenance interventions, better management of critical spare parts, and reduction of downtime. In conclusion, the study provides a practical tool for maintenance decision-making by integrating statistical analysis with predictive modeling. It is recommended to incorporate these models into management platforms such as SAP to enhance the traceability, availability, and reliability of equipment in demanding mining environments.