Estimation of the thermal conductivity of slags by using artificial neural network

Neural network is a strong data prediction technique has a similarity with human brain consists of ten billion nerves called neurons. Biological neuron is a vital unit to understand and to learn the nature of universe. In order to create convenient response throughout neural network system consists of some operational unit such as dendrite who receive the signals from the outside, soma evaluates and conveys the signals through the next layer and axon deliver the responses through the synapses. Neurons talks each other by input when neuron received inputs regarding with various weights from multiple cells. Aggregation of the inputs called "intensity" multiplied by regarding weights. If intensity of the signals is enough to cross over the certain threshold, then ignition process starts and signals are allowed to go ahead to be transmitted all the way through the axon and synapse to be delivered next neuron. Artificial neural network (ANN) is extremely reliable method to be used for estimation because it simulates the human brain behavior mostly based on learned experience from the past. Same phenomenon happens in the ANN system where errors computed for each output and propagated backwards through the layer or units inside the network in order to correct the weights to get better results till minimize the errors which called "learning". Slags have an important role to get desired clean and quality production. "Look after the slag and the metal will look after itself" is strong expression to understand how they are essentials in the industries. It covers the metal from the oxygen and prevent the oxidation, takes off the unwanted elements and prevent the heat losses from the crucibles. Therefore, thermal conductivity of the molten melts was reviewed as one of the important element of the thermal properties of the slags in present study. Thermal conductivity is the diffusivity constant of the conductive term of the heat transfer equation that represent ability of a specific material to transfer heat, given a temperature gradient and length (W/m. K). Flow of energy during the process affects the slag temperature, phase composition, and therefore other thermochemical properties. Present study, thermal conductivity of Na2O – B2O3 – SiO2 and CaO – Al2O3 – SiO2 systems were reviewed at various temperatures. Literature research was studied and data samples collected either in mass or mole fraction of the regarding liquid slag system associated temperatures in order to create database which constitutes learning dataset of the ANN enables to present study to estimate the thermal conductivity of the liquid area in ternary phase diagrams. It was taken into consideration that simulated or estimated thermal conductivity values are coherent with experimental values. According to estimated dataset, iso-thermal contours are depicted in ternary phase diagrams.