Effect of Tempering Temperature on Hydrogen Desorption of AISI4140: Neural Networks Analysis

Abstract

Over the past ten years, Cr-Mo is used for the manufacture of hydro treating reactors, in order to meet the demand of the refining industry, steelmakers have had to develop CrMo steel, which until then was widely used for the manufacture of devices operating at high temperatures under hydrogen pressure The purpose of this work is to study the susceptibility of steel 42CrMo4 (AISI4041) to hydrogen this steel is a martensitic, that is widely use especially in the hydrogen storage industry, this one is tempered at different temperatures (500, 550, 600, 650 and 700℃), the application of artificial neural networks is proposed to predict the optimum tempering temperature to have a minimal quantity of hydrogen by analyzing the hydrogen concentration parameter that escapes through the desorption phenomenon through time at room temperature the Mathematical model proposed by the neural network method has proved this and this has been validated once more by comparing it with a global statistical study following a coupling between the variables “time / concentration”, “Tempering Temperature / concentration”. In addition, the data used in the model were taken from real experimental data and are arranged in a format of two input parameters: time and tempering temperature, and output parameter is hydrogen concentration. The model was arbitrated using Mean Absolute Error (MAE) and average quadratic error (MSE) and correlation coefficient and model performance values found as 0.0686 and 0.0071, 98.87%, 0.0071 for the training part and 0.0916 and 0.0112, 99.10%, 0.0112 for the test. Finally, the major conclusions of this research show that ANN as powerful computational techniques in modeling of nonlinear systems, can be reliably used in the prediction and correlation to obtain tempering temperature to get lower hydrogen concentration lesser.