Fault detection and diagnosis of brake play a critical role in ensuring vehicle safety. Failure to detect alarming defects in a timely manner can lead to component wear and eventual vehicle failure, posing a significant risk to lives and causing potential loss of time and resources. Emphasising the importance of proactive maintenance, timely fault detection can mitigate these risks and prove to be cost-effective. In recent years, the utilisation of machine learning techniques for vehicle fault diagnostics has gained traction. The main objective of this study is to develop a method for identifying issues specific to air brake components utilised by heavy vehicles. With an astounding 99.4% classification accuracy, the random forest (R.F.) technique blew away all of the competing algorithms in a thorough review of classification methods. Also, the black-box models were made more intelligible using the SHapley Additive Explanation (SHAP) approach. This revealed that 20 key factors significantly influenced decision-making processes. Leveraging this insight, a sequential application of the R.F. algorithm using these 20 features yielded comparable accuracy, showcasing the robustness of the proposed approach. Consequently, our proposed algorithm not only maintains high accuracy but also streamlines computational power and complexity, offering a promising solution for efficient and effective brake fault identification and diagnosis in heavy commercial vehicles.