Casting Defect Diagnosis Based On Genetic Algorithm, Neural Network Model

In this essay a neural network model for the diagnoses of grey iron casting & nodular iron casting defects is presented.First,Some commen and typical defects and their relative influential factors in foundry are classified,then sumed up at large,also according to the practical situation in foundry some effective methods are put forward to prevent defects from occuring; Secondly,a neural network model for the diagnoses of grey iron casting & nodular iron casting based on genetic algorithms is constructed on the basis of creating a database about casting defects-influential factors;Finally, a Applications Software for the diagnoses of casting defects is developed by using Borland C++ Builder 4.0, it can diagnose casting defects fast and effectively and give some relative preventive methods,thus the problem quality is improved.In order to facilitate the user this software has a full chinese interface.In a word, it lower waste products rate efficiently. To improve the speed for the learning of network weights in this model, genetic algorithms that is popular recently is applied ,also to try a new weights learning method.this method is more efficient than BP algorithms in practice.