Predictive Maintenance Planning Using a Hybrid ARIMA-ANN Model

Abstract

Predicting machine faults is crucial for maintaining operational efficiency in industrial settings, minimizing unplanned downtime, and ensuring customer satisfaction. Fault prediction helps identify faults and create maintenance schedules. Maintenance planning involves strategically scheduling activities to ensure the continuous operational efficiency of systems. This study focuses on reducing unplanned downtime in a food company by developing a predictive maintenance plan through machine fault prediction. Artificial Neural Networks (ANNs) are excellent in handling non-linear models, while the ARIMA model is adequate for linear models. However, real-world data often contains linear and non-linear elements, requiring hybrid models for improved accuracy. This study employs ARIMA, ANNs, and a Hybrid ARIMA-ANN model. The dataset is individually modelled using each approach. Using a 3-month machine fault dataset, predictive values for machine fault times are generated and statistically evaluated using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE). The findings indicate that the hybrid model outperforms both ARIMA and ANN models. The food company can significantly reduce unplanned downtime and ensure operational efficiency using a hybrid model. Predictive maintenance planning can help the food company save costs and maintain a competitive edge in the market.