| Grain drying is one of the most important postharvest techniques in modern agricultural production because it can decrease grain loss by drying wet grains to specific safety moisture content(MC)levels for the purpose of preserving food from microbial spoilage.An effective automatic control strategy is an important means to achieve grain drying goals.From the control engineering point of view,grain drying is a complex process with the characteristics of long delay,multi-disturbance,strong nonlinearity and uncertainty parameters,for which it is not easy to build a precise mathematical model.Thus,it enables the classical traditional control to meet the challenges.Intelligent control does not rely on the controlled object model.It has the advantages of strong robustness and good real-time performance.It is suitable for the control of complex nonlinear systems and is an effective control method for grain drying.Usually,Artificial neural network or fuzzy logic and other intelligent identification methods can be used to approximate the nonlinear relationship of grain drying systems,and then combine with related control algorithms,such as PID(Proportional Integral Derivative)control,adaptive control,model predictive control,backstepping control,sliding mode control,evolutionary algorithms,or their combinations to construct intelligent controllers.This paper is supported by the China National Common Weal Industrial Special Scientific Research Funds for Grain Industry(no.201413006).Aiming at the difficulties of automatic control in grain drying process before storage,such as large lag,non-linearity,large overshoot and strong coupling,the intelligent modeling and intelligent control methods of grain drying process have been developed by taking the combined multi-functional grain drying experimental system as the research object.The main research work and innovations of this paper are as follows:1.Based on the study of the deep bed drying mechanism model of the wheat mixed flow continuous drying process,several intelligent modeling methods for grain drying process have been studied for infrared radiation and convection grain drying process.Combined with the drying mechanism model,the BP neural network method,the nonlinear autoregressive neural network model and the improved particle swarm optimization support vector machine regression algorithm have been introduced to establish the new dynamic process model of infrared radiation and convection(IRC)grain drying process,and the effectiveness of the models’ predictive performance have been verified.Based on the principle of heat and mass transfer,the mathematical mechanism models of the wheat mixed flow drying process for the combined multi-functional grain drying experiment system which worked in the cycling drying mode and the continuous drying mode have been established,respectively.The two models’ numerical simulation have been made and analyzed by programming in MATLAB,and the prediction results with data collected from actual grain drying experiments have verified the model’s reliability.The established continuous drying mechanism model can be used to simulate the actual continuous drying process to verify the feasibility of the control algorithms presented in this paper in the control simulation study.The Back Propagation(BP)neural network prediction model for the combined infrared radiation and convection(IRC)grain drying process has been studied in this study,and the prediction results are analyzed and verified by the practical grain drying data.A nonlinear autoregressive with exogenous input nonlinear(NARX)model for the IRC grain drying process is presented,and the prediction and application results are analyzed and verified.The modeling accuracy is further verified by comparing with the predicted performance of the established linear autoregressive exogenous(ARX)input model.A support vector machine regression(SVR)modeling method based on an improved particle swarm optimization(PSO)is proposed for the IRC grain drying process in this paper.This method proposes an improved PSO algorithm by introducing a linear decreasing weighting equation based on fitness deviation in the standard PSO algorithm,and it has improved the optimization ability of the standard PSO algorithm.The proposed improved PSO algorithm is used to optimize the parameters of the SVR model of grain drying,which makes the support vector machine regression model of grain drying process has a higher prediction accuracy.By analyzing the prediction results and comparing with the prediction performances of other IRC grain drying process models,the modeling accuracy of the improved particle swarm optimization support vector machine model is further verified.2.A performance objective function is proposed by considering grain drying quality and energy loss control.Combined with genetic algorithm,biological immune feedback algorithm,fuzzy control and support vector machine algorithm,a series of intelligent control methods are proposed.The simulations have verified that these intelligent control algorithms have good performances in the control of grain drying process,such as rapidity,stability,accuracy,and the anti-interference ability.Based on the genetic optimization algorithm and the biological immune feedback algorithm,combined with the fuzzy control and the PID(Proportional Integral Derivative)control algorithm,two improved fuzzy immune PID control algorithms(a genetically-optimized fuzzy immune PID controller and a genetically-optimized double fuzzy immune PID controller)are proposed to control the outlet grain moisture of the combined multi-functional grain drying experiment system.The simulation results have showed that the two improved fuzzy immune PID algorithms can adapt to the complex grain drying process control,and have good performances in terms of rapidity,stability,accuracy,and the anti-interference ability,of which the genetically-optimized double fuzzy immune PID control algorithm has better control performance.In view of the advantages of support vector machine modeling method,two kinds of support vector machine controller for grain drying process are studied and proposed:a genetically-optimized support vector regress(SVR)internal model PID controller(GO-SVR-IMPC)and a genetically-optimized SVR indirect inverse model PID controller(GO-SVR-IIMCPID).The results of the tracking control simulation,the anti-interference test and the robust test have demonstrated the effectiveness of the two SVR control algorithms for grain drying process,of which the GO-SVR-IIMCPID control algorithm has better control performances.The problems of large hysteresis control caused by the lag delay of the drying section and the grain draining section in the control process are analyzed and demonstrated through the timing sequence relationship of feedback-adjustment-output effect.3.Based on the above intelligent control method,a practical grain drying control system is established,which is superior to the existing expert control system and manual control system in terms of grain drying quality and energy consumption.Based on the GO-SVR-IIMCPID controller,the field control for the continuous drying process of wheat mixed flow has been carried out and compared with the control performances of manual control and the existing expert control system,respectively.The comparison results show that the control effect based on the GO-SVR-IIMCPID algorithm is better than the other two kinds of controller. |
