Research On High Temperature Water Stress Corrosion Cracking Prediction Of Austenitic Stainless Steel Based On Data Mining

As the basic equipment to promote China’s economic development,special pressure equipment has been extensively used in various fields such as petrochemical and nuclear power,but its potential failure risk threatens the safety of national life and property.Therefore,it is important to study the damage prediction of pressure equipment,especially the frequent and complex high temperature water stress corrosion cracking(HTWSCC).In this paper,based on the HTWSCC behavior data of austenitic stainless steel,a key material for pressure equipment,the whole process prediction models from crack emergence to expansion were constructed and validated by using data mining techniques.The main work is as follows:(1)Based on the susceptibility data,the classical classification algorithms of data mining are mixed into Hybrid model and the optimization effect of three different algorithms are compared,among which the particle swarm optimization(PSO)algorithm performs the best.Finally,the improved PSO(IPSO)algorithm is combined with the Hybrid model to propose the IPSO-Hybrid austenitic stainless steel HTWSCC susceptibility prediction model.It is verified that the average accuracy of IPSO-Hybrid model is 0.857,up to 0.926,and the effect is good.(2)Based on the crack growth rate data,the prediction effects of different classical regression algorithms are first compared,and the results show that the BPNN has the best performance.Then,the optimization effects of three different algorithms on BPNN are compared,among which the PSO has the best performance.Finally,the adaptive particle swarm optimization(APSO)algorithm is combined with BPNN,and the APSO-BPNN austenitic stainless steel HTWSCC crack growth rate prediction model is proposed.It has been verified that the average fitting degree between the predicted rate and the actual rate of the APSO-BPNN model is 0.80,up to 0.93,and the precision is high.