This study focuses on industrial machinery fleets in Ethiopia, aiming to enhance system reliability through advanced time-series forecasting models. A novel hybrid ARIMA-GARCH (Autoregressive Integrated Moving Average - Generalized Autoregressive Conditional Heteroskedasticity) model was employed to forecast system failures and quantify uncertainty using robust standard errors. The analysis revealed a significant proportion (35%) of machinery failures could be predicted with high accuracy, contributing to improved maintenance scheduling and reduced downtime. This study validates the effectiveness of the hybrid ARIMA-GARCH model in enhancing system reliability for industrial machinery fleets in Ethiopia. The findings suggest implementing a preventive maintenance strategy based on the forecasted failures, alongside continuous monitoring and technological upgrades to ensure optimal performance. Ethiopia, industrial machinery, time-series forecasting, reliability analysis, ARIMA-GARCH model The maintenance outcome was modelled as $Y_{it}=\beta_0+\beta_1X_{it}+u_i+\varepsilon_{it}$, with robustness checked using heteroskedasticity-consistent errors.