Intelligent processing , injection

This work combines the design of experiment (DOE) methodology and neural network to develop a computational-intelligence-based injection molding quality predictor. The DOE obtains the optimal process windows... 

Shelesh-Nezhad, K. Siores, E. Intelligent system for plastic injection molding process design. J. Mater. Process. Technol. 1997, 63, 458-462. 

Someone hands you a bottle and asks whether it was produced by extrusion or injection blow molding. How can you tell What if they do not have the bottle with them—what can you ask about the bottle or the product that would let you make an intelligent guess about which process was used Think of at least three different questions you would ask. 

Gas-assist injection moulding is becoming a frequently-used process for forming hollow, stiff parts. Foamed material, sandwich structures with compatible core and face sheets, and sandwich constructions with recycled material are other examples of how intelligent material utilisation can lead to source reduction. 

Quality of injection-molded parts is determined by a lot of factors, such as machine, plastic material, operation conditions and others. Normally, using numerical or experimental method to examine and control all parameters is still very difficult, hi this study, we proposed a computational intelligence-based method to obtain the optimal process windows systematically, hi addition to this, this method can be used as an on-line monitor to predict part quality from process dynamics data. This method combines design of experiment (DOE) and neural network techniques for intelligent quality prediction. This approach is a potential method to improve the molding stabihty and molded part quality. 

This study combines design of experiment (DOE) method and neural network to develop an intelligent injection molded part quality predictor. The DOE obtains the optimal process windows and neural network is used as the part quality predictor. Various quality predictors are proposed and compared by the real molding trial data in order to compare the feasibility and prediction accuracy of part quality predictors. 

---