The Causes And Prediction Of Quality Defects For The Micro Injection Molding Parts

Now micro injection molding has been widely used. But its technology is immature, it is needed to learn micro injection molding process in the micro level combining with the physical characters of the micro plastics. It is become more important that how to monitor and control the molding quality of the micro injection plastics parts.Because the micro injection plastic parts are so small, the quality evaluation by measuring the part dimensions is difficult. It is well-known that the molding quality of the part depends mostly on the quality of the mould cavity. If the mould cavity is made strictly, the denser the molding part is in the injection molding process, the closer to the cavity the molding part is, and the replicability is good. So the quality of the micro injection plastic parts can be monitored by weightTo study the causes that affect micro part molding process and the effective evaluation for the micro part quality, the micro gear is studied. The micro gear mold for planetary reducer which has four cavities is designed. Two experimental groups are finished using the orthogonal experiment. The experimental factors are injection speed, injection pressure, packing time etc and the weight of the product is the experimental evaluation results. One of the experiments is used as the train data for the quality prediction modle, the other is used as the reference data, comparing with the quality prediction modle.Using the evaluation criteria of the product weight, the factors which affect the quality of the molding are get by the Range analysis. The injection pressure, melt temperature, injection speed, packing pressure, and packing time affect on the weight in the descending order.The BP neural network, improved neural network, mixed polynomial neural network, mixed neural network, wavelet neural network have been studied. Using the injection process parameters as the input variable, the weight of parts in one mold as the output, the prediction modle is built. Through the actual test, the BP neural network and the wavelet network show a good performance. The relative errors are separately 0.6% and 0.75%. Others don't perform well.After getting the end face picture of the gear, through the image methed, the points on the gear tooth are got. The molding quality is evaluated by comparing the radius of curvature of the corresponding points with the corresponding points on the mold cavity.