Study On Chinese Medicine Tablet Coating Based On Neural Network Modeling

Tablet coating is a kind of method for Chinese traditional medicine's coating, which principle is set the coating materials and Chinese medical tablets such as, into a pot that is the environment with high temperature and negative pressure. Because the coating process works in a special way, that each control parameter has very strong nonlinear and coupling each other, that is very difficult to form an effective model to simulate it only in traditional ways. All the processes are basically depended on manual operation, so the product's quality is not guaranteed and the rates of substandard products are quite high.The artificial neural net work theory provides a new idea for solving these kinds of nonlinear system and modeling optimization problems.The dissertation focused on the control features of Chinese traditional medicine coating , formed an artificial neural network at first which is composed by the input processing parameters, by the yield with coating rates to an active evolution neural network model. And then, trained the network with those processing parameters swatches collected from the working processes, to find the relation between the input parameters and output yield rates. Finally, use the global search capability method GA to optimize the input parameters for high product rates.The main context as follows:1. In the artificial intelligence aspect, based on the bionics principle, merge voluntary mutation and BP neural network features, developed a adaptive and efficient intelligent algorithm, active evolution neural network method. Compared with simple genetic neural network, some solutions concerning of the convergence speed and efficiency were put forward.2. In the artificial intelligence aspect, primarily based on the studied automatic control parameters, and the reformed GBB-45 coating machine, researched the coating modeling. The primary effective aspects, including input wind temperature, out put wind temperature, the coating pot air pressure and pump rotational speed, were provided for establishing a operable prediction model.3. In the Chinese traditional drug coating modeling aspect, this paper discusses the difficult of the coating system control, the process is primary depended on the human experience to control the machine, but the results were always not forecasted. So trying to model the machine's process and form an active evolution neural network model to approach the real output. In the real produce of Jiangxi Herbfine Hi-tech Co.Ltd, the results show that the model is an efficient method.