Aiming at the problem that the vibrating screen surface is blocked due to the change of material feeding amount, which affects the screening performance, the angle of the guide chute for the variable amplitude screening mechanism as the variable, the variable amplitude antiblocking screening process under abnormal feeding of rice grains and stems (0.5 kg/s as an example) was simulated by RecurDynEDEM. BPNN was used to accurately predict the variable amplitude screening efficiency and impurity content for different chute angles, and optimized the parameters for the variable amplitude screen, to alleviate the accumulation and blockage on the screen surface and improve the screening performance. The results showed that under the abnormal feeding (0.5 kg/s), the screening efficiency increased steadily within 55% - 58% when the guide chute angle was 0°-10°, the impurity content fluctuated in a small range within 0.1%-0.16%, the screening efficiency increased linearly within 58%-75% when the chute angle was 15°-40°, the impurity content fluctuated and increased within 0.11%-0.19%, and the screening efficiency suddenly decreased linearly to 658% when the chute angle was 40°-45°, the impurity content increased sharply to 0.37%. BP neural network predicted that the variable amplitude screening efficiency was the highest, the impurity content was low, and the variable amplitude antiblocking screening performance was better, when the chute angle was 37°. The screening efficiency and impurity content were 74.30% and 020%. The R2 of BP neural network prediction model was 0999, the prediction error was concentrated at -0.000 41, and the prediction curves were highly fitted, which verified the reliability and accuracy. This provides a basis for the intelligent adjustment of variable amplitude antiblocking screening in the future.