Fault diagnosis of the fluid end faces two challenges. One is that the vibration signal used for diagnosis contains a lot of noise, and the other is that the vibration signal cannot fully reflect the fault state. In this paper, a model based on enhanced vibration-strain signals is proposed to improve the fault diagnosis accuracy of the drilling pump fluid end. First, the sparrow search algorithm (SSA) is employed to optimize maximum correlated kurtosis deconvolution (MCKD) to enhance the impact component of the vibration signal. Second, the enhanced vibration-strain signals are proposed to accurately describe the working state of the fluid end. Third, a one-dimensional convolutional neural network (CNN) with dual-signal feature fusion is established to diagnose faults of the fluid end. The results demonstrate that when utilizing the SSA-MCKD enhanced vibration-strain signal as input, the CNN model achieved an outstanding average diagnosis accuracy of 99.26% across various operating conditions.