Research And Design Of A Wood Drying Control System Based On Artificial Neural Networks

Due to the development and increasing requirements of modern society, the consumptions of natural resources, especially of forest resources, increase rapidly. In consequence of this process, not only has the world forest resource decreased in great magnitude extent, but also is the global eco-system being ruined massively, posing devastating threaten to the living and development of human being. How to effectively exploit limited forest resources, keep energy consumptions in low level, maintain highly efficient utilization of natural resources, has become one of the most urgent focuses for international society.In terms of the living-tree timber quantity stock per person, China is a country with very low forest resource storages. The governments of different levels in China must pay more attentions to forest resource management and bring forward suitable development strategies.Wood drying is a key technical approach, which can improve lumber qualities, avoid large losses of wood products as well as utilize wood materials effectively. Conventional wood drying control methods are based on traditional PID(Proportion Integration Differentiation,PID)control algorithms and the whole control process proceeds according to previous technical experiences. Consequently, the environmental parameters being supervised and controlled can not maintain auto-tuning, thus lacking technical flexibilities. Given this situation, the utilizations of the modern sophisticated control technology have become a urgent and must-do strategy in the domain of modern drying controlling technology.The controlling process of wood drying is a manmade-supervised one, during which the temperature and relative humidity inside drying chambers are recorded, compared with the targeted values and then adjusted accordingly, so that the moisture content of drying timber samples can approach to the objective ones with reasonable drying rates. Taking the characteristics of the controlling process of wood drying in mind , through the study of adding variable momentum BP algorithm, is introduced into the golden section method on momentum adjustment. Preliminary simulation tests showed that the fluctuation phenomenon during the initial study phase of this network can be avoided, effectively decreasing the times of network training and increasing the convergence rates. Through the combination of BP neural network technology and PID controlling algorism, wood drying process can be controlled by a hybrid BP neural network and PID controlling system. Simultaneously, this new controlling system was modeled and simulated by using the tool box Simulink of Matlab software.The traditional PID control system as well as the hybrid BP(Back Propagation,BP) neural network and PID controlling strategy were simulated and compared with each other. Results showed that the threshold response curves of the hybrid neural network PID controlling system have many advantages, i.e., rapid rising speeds, high manipulation precision, excellent steady-state qualities, low transfer time, hyper tuning quantities, making this hybrid controlling strategy highly favorable for multiple-variables, nonlinear, pure time delay processes, in this case, the wood drying. In conclusion, applying the hybrid neural network PID algorism to wood drying controlling process is feasible. This controlling strategy can realize the online optimization of controlled parameters. Further more, it can improve the controlling precision level of wood drying, guarantee drying quality effectively, keep low levels of energy consumption and control production cost.