The diesel fuel system is critical to the functioning of diesel engines, and its failure can result in significant economic losses and operational risks. This paper presents a fault diagnosis method that integrates wavelet analysis with a neural network for managing the health of diesel engine fuel systems. First, the oil pressure waveform of a diesel engine’s high-pressure oil pipe is collected using an external clamp-on pressure sensor. Wavelet threshold denoising is then applied to eliminate noise from the oil pressure signal. Next, characteristic parameters of the oil pressure waveform, such as waveform width, amplitude, and maximum pressure, are extracted, and wavelet packet band analysis is employed to decompose and analyse the signal. These extracted parameters serve as input vectors for a neural network. A combined self-organizing mapping back propagation (SOM-BP) neural network model is developed to diagnose fuel system faults. Experimental results demonstrate that this model significantly enhances diagnostic accuracy and can effectively identify various fault types.