A signal processing denoising method based on the Dung Beetle Optimizer (DBO) and Variational Mode Decomposition (VMD) is proposed to address the issues of complex structure, variable operating conditions, frequent failures, and noise and irrelevant information in monitoring data of diesel engines for certain types of equipment. The DBO is used to optimize the decomposition layer number and penalty parameter of VMD. The decomposed modes are then reconstructed based on kurtosis values, and the Interpretable Random Forest (IRF) model is employed for fault diagnosis of the diesel engine. Furthermore, the proposed method is compared with six fault diagnosis models, including EEMD-IRF and PSO-VMD-SVM, through experiments. The results show that the DBO-VMD-IRF method achieves an average accuracy of 99.18%, demonstrating significant advantages in diagnostic performance, accuracy, and reliability. This proves that the proposed method can provide reliable technical support for fault diagnosis of armored equipment diesel engines.