الاء عبد السلام

Hardness – Strength Correlation Properties of Carbon Steel and Aluminium Alloys Using Feed Forward Artifical Neural Network In this thesis, that aims to build two mathematical model to predict mechanical properties of low/medium carbon steel and AA 2219 Al alloy. The first model that have different heat treated low and medium carbon steels with as received state and second model have different aging time of AA 2219 Al alloy. The samples were tested using hardness and tensile testing, two models were predicted based on feed forward neural networks were developed using back propagation based on gradient descent learning. For first model, the influence of (0.1, 0.25, and 0.5) carbon percentages and heat treatment on the mechanical properties of low and medium carbon steel was considered. The heat treatments were quenching for one hour at different temperatures (850, 910 and 950 ᵒC) then cooling in water and tempering for one hour also at (300, 500, and 750 ᵒ C) then cooling in air. The behavior of mechanical properties of the specimen was estimated utilizing Vickers hardness testing and tensile test. The values of both tests for the quenched specimens were relatively higher than those of the as received samples. The data obtained then were feed to neural network for predict the mechanical properties of low/medium carbon steel. In neural networks were used three layers, first layers take the carbon percentage, quenching temperature and tempering temperature as inputs, and second layers take yield strength, tensile strength and Vickers hardness as outputs, and last one hidden layers. The correlation coefficients for the three responses versus the experimental data were 0.9815, 0.9602 and 0.9764 for yield strength, tensile strength and Vickers hardness, respectively. The second model is proposed to predict the ageing time from the required mechanical properties (yield strength, tensile strength, ductility, and average hardness) of AA2219 Al alloy. The network takes the mentioned mechanical properties as inputs and return the ageing time as output. The results showed that the proposed model can be used to find the proper ageing time for the required mechanical properties. The correlation coefficient r for this model response against the experimental data was 0.9905. The achievement of two artificial neural networks models were very good and showed that models could be utilized for predicting the mechanical properties of low/medium carbon steel at various temperatures as functions of carbon percentage and heat treatment, also for aging time for AA2219 Al alloy.