Prediction of elements on corrosion behavior of zinc-alümium alloys using artificial neural networks

Keywords: artificial neural network, zinc-aluminum alloys, corrosion Depending on the progress in information and communication technologies, systems have been developed to simulate like human behavior. In neural network, human brain work structure has been considered and modeled as connected units like neurons. Each unit of this system carried out the characteristics of a simplified nerve cell. The complexity of such a system is specified by the number of connections, layers, and nodes within the neural-network. In this study, it is aimed to obtain the predicted values of weight-loss changes in time periods occurred as a result of corrosion of zinc-aluminum alloys. Therefore, a neural network is designed according to the result values obtained from experiments. With using a backward calculation algorithm in neural network, experimental result in corrosion behavior of zinc-aluminum alloys have been predicted. Consequently, it has become possible to predict the weight-loss occurs on the unit area of the material (Zinc-Aluminum alloy) in determined time-periods. In spite of pretty much differences between used data groups, average deviation of six different ANN concerned algorithm estimation which is used in five different Zn-Al alloy group training has found under %5. Average deviation in algorithm which gives best result in Zn-Al alloy group is under %3. Neural networks brings entirely different approach to calculations. It uses current samples instead of programming to solve a specific problem. Metallurgical and Materials Engineering is an practical science as the other engineering sciences. Hence several subjects of Metallurgical and Materials Engineering utilize neural network.