Aimed at the need of data processing and analysis in the performance experiment of hydraulic torque converter, we put forward a way of experiment data analysis using neural-fuzzy system. It can reduce manual participation, improve the data processing ability, accelerate computing velocity and realize data handling automation. We can judge product’s performance by this method. The use in experiment data indicates that it can express the relation of original data nicely. The identifying accuracy of data is good. It can satisfy the request of experiment analysis and the analytical conclusion is accurate too.