The Research Of The Soft Computing Method For The Process Optimization Of Gas-assisted Injection Molding

In the manufacturing process of plastic gas-assisted injection molding, the optimizationof the process parameters can use numerical simulation technology to achieve maximumefficiency and shorted the production cycle. With the rapid development of the plastic industry,people take extensive researches on the process optimization of the injection molding withnumerical simulation technology. And based on the numerical simulation technology, somescholars import artificial intelligence method to the process optimization filed of the plasticinjection molding, which can search the optimal process parameters in the molding process.Numerical simulation technology and artificial intelligence method are both useful formanufacturing process, but these methods also have some problems like time-consuming,experimental results only restricted to particular conditions, and required superior with theoperator. This paper puts forward a soft computing method on these problems. This methodestablishes agent model for the pipe parts first, then use the results from the orthogonalexperimental design to train the artificial neural network, and have the trained artificial neuralnetwork to substitute the process of the numerical simulation. This method can obtain goodprocess parameters much faster, and help workers promoting manual testing's efficiency.The order of presentation is as follows:Chapter 1 briefly review the study of the gas-assisted injection molding technology athome and abroad, and points out the deficiency of the present research work, then put forwardwhat this paper will study.In chapter 2, have researched the tubular products which is a large family of the gasassistedmolding products. Firstly establish a initial agent model of soft computing methods with the feature of the models. Based on the initial agent model, import the shape parametersof the model, and the length correction coefficient of the initial agent model, and adoptmathematical analysis method to establish the function relation between the exteriorparameters of the model and the length correction coefficient. Finally obtain the agent modelwhich can well reflect the relative airway length of the original model.In chapter 3, with the agent model as the research object, adopt the appearance parametersof the agent model and molding process parameters as the factors of the test to design theorthogonal experiment, and analysis the experimental results by single evaluation andcomposite evaluation, then obtain every factor'impacts to the different evaluation indicator,and optimize the process conditions of the gas-assisted injection molding. And adjust theseparameters to design several orthogonal experiments, to get enough experimental data to trainthe artificial neural network.In chapter 4, have studied the optimization method of the gas-assisted injection moldingprocess which is based on the BP neural network, With the BP neural network, we can obtainthe approximate relationship between the process parameters and the quality evaluationindicator, and get the quality evaluation indicator of the agent model matching the processparameters rapidly. And have confirmed the soft computing which is based on the agent modeland BP Neural Network.In chapter 5, with the auto rearview mirror product as example, application of the softcomputing method in irregular tubular products was studied. The research shows that the softcomputing method also can be applied for irregular tubular gas-assisted injection moldingproducts, to help workers to promote manual testing's efficiency.In chapter 6, all achievements of the dissertation are summarized and the further researchwork is put forward.