Modeling And Optimization Of Contact Angle Of Rice Leaf Surfaces Based On ANN And RSM

China is a major producer of rice,and spraying chemical pesticides is one of the most effective ways to increase yields during rice production.However,the utilization rate of pesticides in China is relatively low,which causes serious ecological and economic losses.Therefore,scientists have done a lot of research on pesticide utilization of rice.The wettability of rice leaf surfaces directly affects the spread,adhesion and penetration of pesticide droplets on the surface of rice leaf,thus affecting the utilization rate of pesticide.Therefore,the utilization rate of chemical pesticides can be improved by improving the wettability of rice leaf surface.In this paper,the contact angle of rice leaf surface was modeled and optimized to improve the wetness of rice leaf surface,reduce the loss of pesticides,and improve the utilization rate of chemical pesticides.In this paper,by comparing the results of ANN and RSM,a new method for modeling the contact angle of rice leaves was proposed to minimize the contact angle of rice leaves and improve the wettability of rice leaves.The experimental design method of Box-Behnken(BBD)with three factors and three levels was used to study and analyze the influence of temperature,relative humidity and chemical pesticide concentration on the surface contact angle of rice leaves.Through ANOVA analysis of variance and interference design analysis of BBD experimental design results,it was concluded that chemical pesticide concentration was the most important factor influencing the contact angle of rice leaf surface,followed by relative humidity,and temperature had the least influence on the contact angle of rice leaf surface.The interaction between temperature and relative humidity,temperature and concentration of chemical pesticides,and relative humidity and concentration of chemical pesticides had no significant effect on the surface contact angle of rice leaves.Then the optimal design analysis was carried out,and the optimal input parameters were obtained.When the temperature was 25.96℃,the relative humidity was60%,the concentration of chemical pesticides was 131.90mg/L,and the minimum contact angle of rice leaves was 103.565°.The artificial neural network was used to train 116 groups of experimental data,and through the regression curve and residual analysis of the experimental results,it was concluded that the artificial neural network model had high accuracy,could well predict the contact angle on the surface of rice leaves,and had strong generalization.Then the sensitivity analysis method of artificial neural network was used to determine the weight values of three external factors affecting the contact angle of rice leaf surface.The results showed that the chemical pesticide concentration had the greatest influence on the contact angle of rice leaf surface,and the influence weight value was 55.61%.The second is relative humidity,the influence weight value is 31.15%.The least significant is temperature,which has an influence weight of 13.24%.Finally,by analyzing the predicted values and evaluation parameters of RSM and ANN models,the reliability of RSM and ANN models in modeling and predicting contact angle on the surface of rice leaves was compared.The experimental results show that both modeling methods can predict the contact angle of rice leaf surface very well,but the fitting performance of ANN model is better than RSM model,and the prediction accuracy is higher than RSM model.