| The impeller material FV520B-I of centrifugal compressor was applied to the high speed rotary environment.Under the action of cyclic load,the fatigue life of FV520B-I was often more than 10~7 cycles,entering the very high cycle fatigue range.The very high cycle fatigue life of FV520B-I was one of the important factors affecting the life of centrifugal compressors.The research on the very high cycle fatigue life of FV520B-I can reduce the probability of accidents caused by fatigue failure and improve production safety.Therefore,this thesis mainly studied the very high cycle fatigue damage behavior and mechanism of FV520B-I,extracted metal magnetic memory detection characteristic parameters and established the relationship between the fatigue damage and metal magnetic memory.The above work not only provided research support for improving the service life of centrifugal compressor impeller,but also provided guidance for preventing accidents of centrifugal compressor due to material fatigue life problems.The fatigue properties of dog-bone-shaped specimens of FV520B-I base metal and the welded joint up to 10~9 cycles were obtained.The very high cycle fatigue behavior and mechanism of FV520B-I base metal and welded joint dog-bone-specimens were analyzed by fatigue fracture morphology observation,characteristic region size measurement,strength model prediction,etc.The S-N curves of the dog-bone-shaped specimens were lower than that of the funnel-shaped specimens,and the S-N curves of welded joint specimens moved downward compared with that of base metal specimens.For all types of specimens,cracks initiating from the surface matrix and cracks initiating from subsurface inclusions were the main cracking modes in the high cycle fatigue and the very high cycle fatigue,respectively.Large welding defects were introduced in the welding process,the fatigue life of dog-bone-specimens with welded joints was more dispersed,and S-N curve drop down.The Murakami model and the modified model considering the influence of inclusion shape were used to predict the fatigue strength of10~9cycles.The predicted results were close to the test results of base metal specimens,and the predicted values of the modified model were closer to the test results than Murakami model.Electromagnetic resonance fatigue testing machine and ultrasonic vibration fatigue testing machine were used to take high cycle and very high cycle fatigue tests on FV520B-I.After a certain number of cycles,metal magnetic memory testing equipment was used to test the specimen.The metal magnetic memory equipment can obtain the normal magnetic field intensity Hp_y and the slope K of Hp_y under different cycles.We studied the influence law of normal magnetic field intensity Hp_y and slope K to fatigue cycle,fatigue damage location and crack size.The analysis results showed that in the very high cycle fatigue test,the normal magnetic field intensity Hp_y and its slope K changed after cracks appeared in the specimen,and the position of the normal magnetic field intensity Hp_y and K changed were corresponding to the crack location of the specimen.The normal magnetic field intensity Hp_y of a very high cycle fatigue specimen changed before the obvious cracks appeared.The variation range of Hpy of the normal magnetic field intensity increases after obvious cracks appeared.The normal magnetic field intensity Hp_y increased after obvious cracks appeared.Considering the prediction effect of metal magnetic memory detection in high cycle fatigue,it showed that metal magnetic memory detection is feasible for predicting the very high cycle fatigue life.The slope K changed more obviously than the normal magnetic field intensity Hp_y.The slope K had zero point at the crack location and was not easily affected by the environment.There was an obvious crack in one specimen,but the normal magnetic field intensity Hp_y and its slope K did not change.The reason for this situation was that the crack width of the specimen is too small,which lead to the closure of the crack surface.Leakage magnetic field didn’t form on the specimen surface.Therefore,the normal magnetic field intensity Hp_y and its slope K did not change.Finally,several methods were used to predict the very high cycle fatigue life of FV520B-I.The prediction methods include the traditional model Paris formula,machine learning and deep learning.Collecting the data of FV520B-I in very high cycle fatigue test,and analyzing the predicted results of the three methods.The results show that the accuracy of prediction to the traditional model Paris formula and most machine learning models were low,while the accuracy of prediction of stochastic forest algorithm and deep neural network learning were high,and the determination coefficient R~2 was more than0.9.By analyzing three methods,it provided the basis for the physical mechanism-data-driven prediction about very high cycle fatigue life of FV520B-I. |
PDF Full Text Request