Algorithm Based On Ga Combined With Bp And Its Combine Harvesters Threshing Performance Modeling Applications

Seed-husking plant is the key unit of combine,whose improvement and perfection is mainly based on the nonlinear mathematical model of the threshing performance with its influencing factors. In the process of threshing,the significance of factors,which influence threshing performance,not only include some certain factors such as feed quantity,cylinder rotary speed,distance between cylinder and concave,thresher concave radian but also include some uncertain factors such as the crop's variety,water content and the proportion of cereal straw. For this reason,the seed-husking plant is system with a character of uncertain,multi-input-output and complex nonlinear. Under the uncertain conditions,the purpose of setting up nonlinear mathematical model of the threshing performance with its influencing factors is to optimize all parameters of construction and motion so that we can get the optimum threshing performances.The key of BP neural network is the capacity of parallel computing,distributed saving,self-studying,fault-tolerant and nonlinear function approximating. In this paper,we first analyze each factor of influencing threshing performance,and deficiency of all traditional methods such as single factor,orthogonal experiment,variance analysis and regression analysis,which have been used to study the threshing performance. In the basis of above analysis,we propose a new method of threshing performance modeling-a BP neural network. By use the new ways of threshing performance modeling-a BP neural network,we can obtain the optimum model of threshing performance,which can better describe the seed-husking plant's feature of complex nonlinear,multi-input-output and indefinite.At present,during the course of BP neural network's learning and training,we often adopt the algorithm of back-error propagation,which is based on global error function's gradient descent,whose essence is a point-to-point search algorithm. Because the weight space formed by the BP neural network's global error function,which include extreme point,is a hyper-surface,considering the initial parameters of BP neural network's structure,so the BP neural network has a inherent deficiency of easily falling into local minimum. Genetic algorithm has the feature of parallel and globaloptimization,which mimic Darwin's evolutional model of survival-of-the-fittest. After analysis of each characters of BP algorithm and GA algorithm,against the BP neural network's deficiency of easily falling into local minimum,we propose to use a algorithm combined GA with BP to overcome BP neural network's inherent deficiency.After analysis of the feasibility of algorithm combined BP with genetic,we explore how to combine GA algorithm with BP algorithm,give a detailed algorithm and concrete implementation process of GA-BP neural network's hybrid studying-training. At the same time,we give a neural network's topological structure,which will use to solve the concrete problem oriented to threshing performance modeling.We used Delphi language to develop a GA-BP neural network's simulation software in this paper,which implemented the thoughts of threshing performance modeling. Using this simulation software,we gave a experiment on the speed-controlled threshing unit for wheat offered by the college engineering of LUOYANG,the result of test verified the feasibility of threshing performance modeling.In the end of this thesis,the application prospect and further research domains of GA-BP neural network are presented.