Study On The Technology And Method Of Intellectualized Defect Diagnosis Of Injection Parts

The quality of injection molding part is relative deeply to material properties, mold design, plastic processing parameters setup, and environment condition.Of all the factors, the process setup has more complicated effect on part quality. The conventional method diagnoses plastic part's flaws manually, which deeply depend on the operator's knowledge and experience. With artifical intelligence developing, it provides a new method to diagnose the defect of injection part. Base on the study of plastic part faults, reasons and solutions,combining the characteristic of analytical diagnosis processing for defects, an method of the part flaws diagnosis based on fuzzy neural network is present.As far as the structure of neural network, two layers and three level fuzzy-neural network with hiberarchy reasoning relationship is constructed. Meanwhile, with the method of several small-scale neural network structures substituting the single large one, which has solved the problem that the single large neural network need more time to calculate more data in sample study-train and analytical diagnosis, study and reasoning velocity of the neural network has been effectively improved. With two additional layers, fuzzy layer and remove fuzzy layer, added before and after neural network, the uncertainty caused by defects in conventional diagnosis system and the effect of fuzzy description on the analytical diagnosis for part defect are all resolved preferably, and the accuracy of analytical diagnosis system is increased. In order to accelerate calculation and avoid the vibration of BP network, mixed method is built to calculate the weights and thresholds of multi-layer neural network by utilize BP algorithm and terraced optimizations, which results in rapid computing of weights and thresholds.The knowledge base is the basis and premise of intelligent diagnosis system to make reasoning. The experts' knowledge and experience of solving the problem is converted into the training samples for fuzzy neraul network.Through training,the knowledge implicated in the samples is saved as the weights and threshold of the neraul network.By this means,the knowledge base of fuzzy neraul network is formed. Meanwhile,the knowledge of solution to reasons is expressed into rules and saved to form the reason to solution knowledge base.An intelligent system of the injection part flaws diagnosis is developed using Access2000,Matlab and VB6.0 . Finally, considered two defects on plastic part in practical production as samples, the diagnosis result is the same to the practical result.