The airflow field was measured and the cleaning test of rice was carried out on the self-developed-air-and-screen cleaning unit. The airflow field was studied by the neural network techniques. Three BP neural network models were developed, which were built between two wind factors and distribution of airflow field, distribution of airflow field and cleaning performance, two wind factors and cleaning performance, respectively. Predict test was carried out with the experimental data. Forecasting results proved the effectiveness of the neural network model.