Machine Learning in the Production Process Control of Metal Melting

Abstract

While cast iron is occasionally melted in a unique kind of furnace, in nowadays, electric induct-ion heaters or arc furnaces with electricity are more frequently used. The regulation of the metal smelting process using machine learning is presented in this research. Nonlinear relationships between process parameters define metal melting as a dynamic manufacturing process. The objective is to establish "smart foundries," which are establishments with centralized management of supply chains, production processes, and customer orders. The idea of a smart foundry is still in its infancy, therefore it requires a workforce that understands its relevance and is willing to follow technological growth. One method to approach the concept of the smart foundry is to apply artificial intelligence (AI) techniques in the R&D of manufacturing processes in foundries, in addition to deploying them in conventional production structures. The manufacturing of cast iron is a topic of our investigation in this specific instance. The neural network is machine learning methods that have been used. Their use is to forecast the quantity of alloying additives needed to produce white cast iron with the required chemical makeup. In the training phase & testing phases, the neural network model outperformed the previously investigated model, making it suitable for use in the management of the production of cast iron. We utilize Neural Networks for our experimentation and compared them with previous results. The results of NN are more accurate. The supplied neural network (NN 4-14-4) is made up of an input layer with four inputs, an output layer with four outputs, and a single hidden layer of 14 neurons. Neurons in the input and hidden layers of a NN employ sigmoid functions, whilst those in the output layer use linear transfer. The MSE-mean square error of the neural network (NN4-14-4) during the training stage is 0.14, which is a very impressive result.