Abstract: In order to reduce the occurrence of safety accidents in agricultural machinery, a sensitivity analysis method of agricultural machinery safety evaluation based on BP neural network was proposed. Through analysis of the man-machine-environment system of agricultural machinery, various safety evaluation index systems were determined whereby the safety level of agricultural machinery was defined, the index parameters were sampled and normalized, and the safety evaluation sensitivity mapping model based on BP neural network was established. BP neural network mathematical model and Tchaban algorithm were calculated and analyzed, and the predicted value and sensitivity coefficient of agricultural machinery safety test were obtained. The results showed that an agricultural machinery safety evaluation system with 10 secondary indexes had been constructed. The structure of BP neural network was“9-10-1”. The distribution of the safety risk value of agricultural machinery was 0.70≤Y＜0.85, and the prediction results were in good agreement with the actual situation. According to the sensitivity analysis of Tchaban algorithm, the sensitivity coefficients of large safety indicators were2.42% for maintenance personnel, 2.28% for electrical system, and 2.05% for work space. The presented sensitivity analysis method provides design guidance for modular and systematic scheme design optimization of agricultural machinery products and subsequent product iterative design.