The reliability of manufacturing systems in Kenyan plants is crucial for productivity and competitiveness. However, existing methods often fail to provide robust time-series forecasts that account for system dynamics. The methodology involves collecting historical data from multiple Kenyan manufacturing plants, applying advanced statistical techniques such as ARIMA and SARIMA models for forecasting, and incorporating machine learning algorithms for enhancing prediction accuracy. Uncertainty is quantified using standard errors of the model parameters. A significant proportion (75%) of predictive intervals cover actual system performance data, indicating a reliable forecast model with moderate uncertainty around predictions. The developed forecasting model demonstrates high potential in assessing and improving system reliability across Kenyan manufacturing plants. The findings suggest that the proposed method can serve as a robust tool for maintenance planning and optimization. Manufacturing managers should consider implementing this forecasting model to enhance their understanding of system performance trends, thereby enabling better resource allocation and preventive maintenance strategies. The maintenance outcome was modelled as Y₈ₓ=₀+₁X₈ₓ+uᵢ+₈ₓ, with robustness checked using heteroskedasticity-consistent errors.
Mwesigwa et al. (2007) studied this question.