Mg Cored wire method is a type of spheroidization treatment method. It has been introduced by many foundries from the viewpoints of improving the environment safety of factory workers, automation and productivity. However, the prediction of residual Mg content depends on experience because various factors complexly affect each other and the effects of each factor on Mg yield have not been clarified theoretically.

  In this research, we regard the relationship between these factors of spheroidization and residual Mg content as a nonlinear optimization problem. Therefore, we attempted to predict residual Mg content in the ladle and the product after pouring, which was spheroidized by Mg Cored wire method using the layered neural network (ANN). In ANN learning, the method of using the actual measured data as it is, the method of utilizing engineering knowledge, and the method of correcting data noise were examined. In addition, we constructed ANN for the relationship between the operation conditions of the spheroidizing treatment and residual Mg content in the ladle, and the relationship between the residual Mg content of the ladle and pouring conditions and residual Mg content of the product.

  As a result, the predicted residual Mg content in the ladle using ANN could be estimated with the correlation coefficient R ＝ 0.91. According to the constructed ANN, the higher the residual Mg content in the ladle, the lower is the yield rate of the Mg in the product. The yield rate of Mg could be improved to 54% by reducing the residual Mg content in the ladle to 0.042%. It was also clarified that the residual Mg content is the same in the ladle and the product. When this result was verified in an actual plant using 25% Mg wire, the yield rate of Mg improved by 53% at a monthly average and to about 65% at the highest value.

  The graphite shapes of flake graphite cast iron can be classified into five types of A, B, C, D, E or six types when containing chill structures. Although the method of numerically determining the quality of spheroidal graphite cast iron with nodularity, has been suggested, there is no method to numerically qualify the quality of flake graphite in flake cast iron.

  In this study, we cast test pieces with a diameter of 30mm to determine the graphite structures of flake graphite casting by AI. That is, we let the AI determine the structure of flake graphite of gray iron from types A to chill. As a result, we will be able to predict the tensile strength from the graphite structure which is determined by AI, hardness, and composition. We used Convolutional Neural Network (CNN) for structure determination and Deep Neural Network (DNN) for tensile strength with AI.

  Specifically, we obtained the following results.

  As there is a mixture of several types of graphite are mixed from type A to chill in the flake graphite test piece or in flake graphite photo, the graphite shapes should be classified with the graphite ratio. In addition, the high accuracy of the graphite classification by AI was confirmed by the following : (a) evaluation of AI prediction with Confusion Matrix, (b) comparison with ISO, (c) comparison with graphite type index point (GTP) by Kanno, and (d) the high relationship coefficient (R ＝ 0.97) with regard to the prediction of tensile strength.

  Slag in a melt must be removed because it causes product defects. In the slag removal process, flux such as limestone is put into the molten metal to remove the slag generated on the surface of the molten metal. No attempts have been made to automatically recognize the state of the cast iron molten surface and to remove slag. In this study, the position of the slag in the cast iron molten surface is automatically detected by performing image recognition by deep learning. In addition, the state of the molten metal surface is determined to recognize the state in which the slag removal process has been completed.

  As a result, it has been found that the slag position can be determined with high accuracy by using pix2pix to distinguish the furnace wall from the molten surface and to classify the image of the molten metal surface using a convolutional neural network. In addition, by classifying the slag image using the convolutional neural network, it is possible to distinguish between the state in which the slag is floating and the state in which the flux is floating.