Study On Behavior Of Low-cycle Short Fatigue Cracks Under Complex Stress States And High Temperatures

With the advance and development of the technology, all kinds of mechanical products or equipments advance gradually toward the direction of high-temperature, high pressure, high-speed and large-scale. Fatigue fracture becomes the main failure modes of structures and components. According to statistics, more than about80%of the failures in modern industrial fields relate to fatigue mechanism and fatigue fracture occurs mostly under the condition of the complex stress states. Fewer and fewer long cracks appear in the materials and engineering structures, whereas short cracks are mostly observed. The existing approaches mainly concentrated on a single crack behavior, whereas the material fatigue damage process is actually caused by all cracks. Therefore, futher study on the collective evolution of short fatigue cracks at high temperatures and complex stress states can better reveal the material fatigue performance and effectively prolong the service life of materials. Researches indicated that short fatigue cracks’behavior has has a strong collective behavior and and random statistical complexity, and the traditional methods have some limitations in analysing them. Therefore, it is necessary to put forward a new approach which can allow for and implement collective evolution of short fatigue cracks. The new approach can futher explore the damage law of short fatigue cracks, reveal the physical mechanism of fatigue damage, and then provide a theoretical basis for the safety and reliability evaluation of structures.The numerical simulation, the fatigue damage law and the life prediction were studied in this thesis. Let the low carbon steel (0.20wt%C) as the experimental object and the collective short fatigue cracks at complex stress states and high temperatures as the research object in this thesis. The study contents are mainly followed:carrying out the numerical simulation of collective short fatigue cracks’evolution based on the Monte Carlo method; the study on damage law of collective short fatigue cracks by using fractal method; the test data training and the prediction of short fatigue cracks’characterization parameters by using GA-BP neural network.The high temperature low cycle fatigue tests on low carbon steel were carried out. Complex stess state was achieved by changing the notch radius of the specimens. During the fatigue test, the damage evolution process of material was tracked and recorded by using interrupt test method and the application of metallurgical microscope and measuring equipment. The characteristics of material damage evolution and initiation, propagation law of short fatigue cracks were obtained accordingly. The several cyclic process of fatigue test was numerical simulated by the application of ANSYS. The influence rule of the notched size load amplitude and other parameters on the distribution of nodal stress and strain were obtained.The evolution of short fatigue cracks under the conditions of high temperature and low cycle were numerical simulated. The material mesoscopic structural model and cracks’ initiation and propagation model was proposed based on Monte Carlo random ideas. The models consider the influence of barriers between the grain boundaries and interference between the cracks. The evolution process of collective short fatigue cracks reappeared by Monte Carlo simulation. The Monte Carlo simulation data was in good agreement with the test data.The parameters to characterize the evolution of short fatigue cracks were studied. Image processing technology was applied to analyse the test data. Otsu algorithm was used to achieve binary image segmentation. Binarization and fractal dimension calculation of test photos with short fatigue cracks were achieved by self-programming. The collective evolution of short fatigue cracks were fractal reflected by the law of fractal dimension evolution with the fatigue life. The results showed that fractal dimension can be seen as a new parameter for describing damage evolution process. A model was performed based on a Paris-type equation for the evolution of fracal dimensions as a function of the cyclic plastic strain energy density. The experimental results of fractal dimension growth rate and FEM analysis of the range of cyclic plastic strain energy density were in good agreement with the proposed model. This good match also demonstrated that the model is valid for computing fatigue collective damage life caused by all fatigue cracks with acceptable accuracy.Some parmeters of fatigue test results were training by neural network. Genetic algorithm was proposed to improve the weights and thresholds of BP neural network. The network design combined with genetic algorithm and BP neural network was also introduced. GA-BP neural network training took load amplitude, strain amplitude, temperature, fractal dimension, notched radius, life percentage and other test data as the input parameters, respectively took the crack number density, the crack growth rate and fatigue life as the output parameter. The network training results were in good agreement with the experimental results. The model based on GA-BP neural network was reasonable and effective to simulate the collective evolution behavior of short fatigue cracks.