The internal leakage of the hydraulic cylinder is unavoidable in the failure of construction machinery. This failure will reduce the working efficiency of the hydraulic system, and serious internal leakage can also cause safety accidents. A real-time measurement system for hydraulic cylinder leakage was constructed,a method of simulating the small internal leakage of hydraulic cylinders was proposed,and an experimental model using pressure strain gauge to convert the flow change into a strain signal was used to analyze the experimental data.A strain-flow mathematical model was established, and neural network learning and training was used to predict internal leakage.Finally, the actual small internal leakage was compared with the predicted value of the neural network.The experimental results show that the neural network has high-precision and high-efficiency predictive capabilities, which provides reference for the small leakage monitoring of hydraulic systems.