Abstract: Based on the neural network model for diagnosing defects of ductile iron castings, an investigation in influence factors of abnormal nodularizing and decaying defects in nodular iron casting is conducted by MATLAB NEURAL NETWORK TOOLBOX. The simulation results show that the abnormal nodularizing and decaying defects can be eliminated effectively by controlling chemical composition and pouring temperature. The method proves itself an aid to improving the quality of the castings and controlling the process of ductile cast iron production, making it easy to analyze the cause producing defects of ductile iron casting in practical production.