Study Of High-temperature Creep Life Prediction Of 9-12% Cr Heat-resistant Steel Based On Microstructure Characteristics

9-12%Cr heat-resistant steel has good high temperature resistance,corrosion resistance,oxidation resistance and excellent creep resistance.Moreover,it provides excellent long-term high temperature strength and is low cost compared to austenitic steel and nickel-based alloys.In order to ensure the safe and efficient operation of these high-temperature and high-pressure components,it is necessary to accurately predict and evaluate the creep life of 9-12%Cr heat-resistant steel.The existing extrapolation of long-term creep life based on laboratory short-term creep data leads to excessive prediction of creep life.One of the reasons is that the change of microstructure under long-term service is not considered.Therefore,based on the phenomenological model,physical mechanism model and machine learning model,this paper establishes a high temperature creep life prediction model considering the influence of microstructure,and comprehensively evaluates the accuracy of predicted life.(1)In this paper,the new parameter master curve of Exponential function based on TTP method is used to predict the high temperature creep life and rupture strength of 9-12%Cr heat-resistant steel by using the latest creep database published by NIMS in 2020.The results show that the Exponential function has higher prediction accuracy than the traditional polynomial function in the low stress range beyond the experimental creep data,and the violin plot distribution range of the predicted life error(t _P-t_r)fluctuates the least.However,with the increase of temperature and the decrease of stress,the disadvantage of the phenomenological parameter method is the overestimation of creep life and rupture strength.(2)In order to solve the drawbacks of traditional phenomenological prediction methods,based on the creep cavity rupture theory,this paper analyzes the microstructure and geometric characteristics of the creep cavity,and establishes a creep life prediction model called the cavity rupture life coefficient U*,and predicts the creep life and rupture strength of 9-12%Cr heat-resistant steel.The results show that the creep rupture strength curves of U*method at different stress and temperature levels are basically consistent with the NIMS test data,especially at low stress,with an average fitting degree R~2=0.99.The R~2,RSS,RMSE and AARE values of the model evaluation are compared with the TTP method of the polynomial function parameter master curve and the NIMS Assessment-2020 and ECCC Assessment-2019.It is concluded that most of the prediction data points of the U*method is basically on the t_r=t _P line,and the prediction accuracy is the highest.(3)To address the difficulties encountered in predicting the high-temperature creep life based on the microstructure physical mechanism model with experimental means and quantified microstructure at low stresses,and the problems of many model parameters and the inability to guarantee the accuracy and precision of the prediction results,the RPROP-NNs method was first used to predict the high-temperature creep life and rupture strength of 9%Cr steel without coupled microstructure parameters.The results show that the RSS and RMSE of the RPROP-NNs method are 1 to 2 orders of magnitude lower than those of the TTP method and the U*method of the main curve of the exponential function parameters.And AARE is also around 2%,which is the lowest.At the same time,the creep fracture strength value extrapolated to 250,000hours by RPROP-NNs method is basically consistent with NIMS Assessment-2020 and ECCC Assessment-2019.However,in the newly collected microstructure data of 9%Cr steel,based on the RPROP-NNs method,the microstructure characteristics related to the creep process(including martensite laths,subgrains,dislocation density and precipitated phase)are calculated and introduced in this paper.The obvious changes are related to the macroscopic properties,the correlation between the microstructure characteristics is revealed,and the creep life of the steel studied is predicted and the error analysis is carried out.