Study On Modeling By Support Vector Machine And Fuzzy Neural Network Control Of Wood Drying

Wood resource is a kind of green materials, which could regeneration and recycling. Wood production belongs to the basic industrial in national economy, and maintains important position. Our nation’s status quo is lacking of forest resources, and now a day, how to improve wood utilization is becoming one of the most significant projects. Wood drying is an important link in wood processing duo to the fact that the quality of drying wood could directly affect the wood utilization and the quality of wood production. Adopting reasonable drying technology and improving the control strategy could be the most effective way to ensure the drying result, increase the drying quality and cut down cost of drying wood could.Wood drying process maintains the characteristics of large lagging, strong coupling, and non-linear, and normally the circumstances in the wood drying kiln is extremely complicated. Therefore, it is hard to build the precision math model to describe this system. In our nation, manual adjustment is widely used in wood drying kiln to control drying process, which would be hold in low control precision of the temperature and humidity, unsatisfactory drying effect and hardly to expand the production scale.A modeling method, which is based on support vector machine, is proposed in this paper. The support vector machine has the ability of self-learning and approaching any non-linear function. Wood drying process includes two part of model structures that are standard model and control model. The temperature, humidity and moisture content data during the drying process were used to build the wood drying process standard model, which is based on support vector machine. Predicting the wood moisture content by standard model could provide precise and real time data of the wood drying process. Wood drying process control model was the model between control signal, temperature and humidity. Based on wood drying craft, the temperature and humidity of drying process are decided by opening degree of heating valve, steam injection valve and moisture valve. Therefore, the opening degree of heating valve, steam injection valve and moisture valve were picked as input; temperature and humidity were picked as output to build support vector machine wood drying process control model.In this paper, the structure, parameters and sample data of support vector machine was optimized by several intelligent algorithms. It could effetely improve the accuracy of the model to accomplish prediction and modeling on line of wood drying process. Firstly, the least square was used in this paper to optimize the structure of support vector machine; secondly, grid search method, genetic algorithms and particle swarm optimization were used to optimize the penalty factor and the kernel function parameters of the model; finally, wood drying process on-line model that was based on weighted fuzzy membership was built through pretreatment of sample data by weighted fuzzy membership method and improved off-line support vector machine modeling method. This model could reflect operating status in current time, and could forecast the operating status next time. Therefore, this model reflected dynamic characteristic of wood drying system, and avoided the problem of model mismatch, which could cause by-changing the inner circumstance, changing the condition of actuators, external disturbance and manual adjustment.A T-S fuzzy neural network temperature and humidity controller, which is optimized by genetic algorithms, was proposed in this paper for improving the drying quality and accurately controlling the temperature and humidity of wood drying process. Based on fuzzy control, fuzzy algorithm was used to remove the coupling relationship between inner temperature and humidity of wood drying kiln; the self-learning and adaptive ability of neural network were used to accomplish the fuzzy logic of the-whole non-linear process; and the parameters of neural network was optimized and trained by GA to improve the self-learning and adaptive ability of the system. This controller combined fuzzy logic, neural network and genetic algorithms, which could express each advantage. It may have stronger approximation ability to the nonlinear systems; faster training convergence speed and more stability of the algorithm, and it could fulfill the temperature and humidity control demand of the wood drying system.It could maintain the significant meaning of studying wood drying craft, fulfilling intelligent control of wood drying process, improving drying quality, reducing energy consumption and decreasing the drying time by building the standard model and control model of wood drying process and by designing and optimizing the fuzzy neural network controller.