Development of a Predictive Maintenance Management Model to Enhance the Reliability of Material Handling Equipment in the Cement Industry Using Machine Learning: The Case of Mbeya Cement Company

Equipment reliability remains a pressing concern in cement manufacturing, where unexpected breakdowns of material handling equipment lead to costly production interruptions. This study developed and evaluated a predictive maintenance management model designed to strengthen the dependability and operational continuity of materials handling equipment at Mbeya Cement Company in Tanzania. A mixed-methods approach was employed, combining qualitative insights from maintenance engineers with quantitative analysis of over 30,000 real-time sensor records capturing vibration, temperature, speed, load, and current through the plant’s Supervisory Control and Data Acquisition (SCADA) system. Among several machine learning classifiers tested, the Multilayer Perceptron (MLP) neural network delivered the strongest fault prediction capability, attaining an accuracy of 94.7%, an F1 score of 93.1%, a precision of 92.5%, and a recall of 93.8%. Vibration emerged as the single most informative diagnostic parameter (Information Gain: 0.682), followed by current, temperature, load, and speed. Strong positive correlations between load and vibration (r = 0.80), temperature and vibration (r = 0.74), and load and temperature (r = 0.73) confirmed known physical relationships and validated the sensor suite adopted for the model. These findings demonstrate that integrating sensor-driven condition monitoring with machine learning can substantially reduce unplanned downtime and maintenance expenditure, offering cement manufacturers a practical, data-driven framework for transitioning from reactive to predictive maintenance strategies.