Cavitation fault is one of the most likely fault categories to occur in piston pumps. The occurrence of cavitation faults not only increases the vibration and noise of the piston pump, but also causes damage to its components. Therefore, diagnosing cavitation faults of piston pumps is of great significance. Traditional cavitation fault diagnosis has drawbacks such as low diagnostic accuracy and poor interpretability of the model. To address these issues, this paper proposes an intelligent fault diagnosis method based on the rough set and radial basis function neural network (RS-RBFNN). Firstly, the mechanism of cavitation faults is analyzed. Meanwhile, a simulation model of the fluid domain of the piston pump is established to simulate different degrees of cavitation categories. Then, various characteristic signals of the piston pump are collected. Wavelet transform is employed for filtering and noise reduction. The rough set (RS) theory is utilized to construct the original decision table for cavitation faults. Through equal-interval discretization processing and attribute reduction based on the genetic algorithm, the effective knowledge and rules of the data are mined. Subsequently, an initial structure of the radial basis function neural network (RBFNN) is constructed according to the rough set conditional attribute set to achieve fault diagnosis for different cavitation degrees. The results show that the RS-RBFNN-based method can effectively diagnose cavitation faults, with an accuracy rate of 98.6%. It has significant advantages in diagnostic performance, which is of great importance for improving the reliability and economy of piston pumps.