The failure of automotive belts in vehicle transmission systems can result in severe mechanical disruptions. However, conventional time-based maintenance strategies lack precision, often leading to either premature replacements or unexpected failures. We propose a condition-based predictive maintenance strategy that utilizes the real-time monitoring of belt wear to enhance the accuracy of lifespan estimation. The primary objective of this study is to overcome the limitations of time-based maintenance by developing a data-driven predictive model that accurately estimates the remaining lifespan of automotive belts on the basis of thickness-wear progression. A nine-month experimental study was conducted, during which an automotive belt’s thickness wear was continuously monitored using a high-precision displacement sensor. The collected data was processed using MATLAB, where curve fitting was performed, leading to the derivation of an eighth-order polynomial equation by the least squares method. This mathematical model serves as the foundation for predictive analysis, enabling accurate estimations of belt-wear progression and failure timelines. By leveraging this predictive model, maintenance planning can be significantly improved, reducing the risk of unexpected failures while optimizing replacement schedules and lowering operational costs. Furthermore, in this study, we present a novel condition-based maintenance framework that is compatible with Internet of Things (IoT) applications, facilitating real-time diagnostics and smart predictive maintenance in automotive engineering.