| The duplex stainless steel with austenite and ferrite two phase structure, therefore it combines both the performance of austenite stainless steel and ferrite stainless steel. It has been extensively used in several industrial applications. However, the difference stress and strain distribution on phase boundary of austenite and ferrite under hot working conditions often lead to the formation drawback of surface cracks and edge cracks, which will seriously affect the quality of products and subsequent processing. Therefore, the hot ductility and deformation behavior of duplex stainless steel were studied in this paper, attempts to establish a theoretical basis for optimizing hot rolling and hot forging processes.In this paper, The elevated temperature compressive tests (deformation temperature from 900°C to 1200°C, strain rate from 0.1s-1 to 50s-1) of 2205 duplex stainless steel was performed on Gleeble-3800 thermal mechanical simulator. The elevated temperature compressive deformation behavior and microstructure evolution of 2205 duplex stainless steel were analyzed. Based on the Dynamic Materials Model and destabilization criterion, the processing map was constructed in the experimental conditions. The results show that flow stress level increase actually increases with strain rate and decreases with deformation temperature. It is well established that high temperature and low strain rate lead to the onset of dynamic recrystallization. The constitutive equation of peak stress and the relationship between Z parameter and peak stress are formulated with Sellars'hyperbolic sine model. According to the results of processing map and microstructure characterizations under different deformation conditions, an optimum hot working regime was established: temperature range of 1150°C～1200°C and strain rate of 0.1 s?1. As for its complicated dvnamic response characters during deformation at elevated temperatures, BP neural network was used for establishing elevated temperature constitutive model and the processing map of 2205 duplex stainless steel based on the experimental data. At the same time, the normalization method was employed for avoiding over-fitting in the model. The results of the study show that prediction values bascily coincided with values of BP neural network.It is highly complexity non-liners problem that chemistry composion and thickness of stainless steel clad sheet affected on shear strength in stainless steel clad sheet. However, it can not found law to use traditional method of analysis eauation and numerical simulation. So the prediction model of shear strength in stainless steel clad sheet by artificial neural network and genetic algorithm have been developed. Creq value, Creq/Nieq value, thicknesses of base and cladding plates were employed as inputs while the shear strength of stainless steel clad sheet was taken as output. The optimal unit number in hidden layer was determined by try and error measure. The BP neural network model architecture was chosen to be 4-7-1. Three training algorithms were compared with train error, test error, number of epochs at stopping points, which were Levenberg-Marquardt, Quick-Propagation, Standard back-propagation algorithms. Levenberg-Marquardt algorithms which has minimum test error and fastest computation speed was chosen as training algorithm in this paper. In addition, early stopping algorithm method was employed for avoiding over-fitting in the model. GA was used to optimize the power and threshold values of the developed BP neural network model, which made the ARE of shear strength was only 1.69%. The developed optimal model shows that it is effective and credible in practical application. |
PDF Full Text Request