Neural Network-Based Reliability Analysis Of Structures With The Local Stress Concentration

In the practical engineering, the damage and destroy of the structure often occur at the stress concentration position, therefore, it becomes very important for stress concentration factors (SCF) to be determined. The expression of stress concentration factors can be solved difficultly using theoretical methods, and so experimental methods are applied widely. Applying experimental methods, the numerical solution can be obtained accurately, however, the explicit expression of stress concentration factors can't be given directly and which will bring some difficulties for the farther reliability analysis and design. Thereby, it is necessary to seek the approximate method that can give the formulation of stress concentration factors with simple form and nice precision on the base of the experiment data. An important purpose of determining the stress concentration factors is that the stress concentration factors will be combined with the modern design method and applied to the practical engineering structure. The reliability and safety of engineering structures is one of the major objectives of structure design. All kinds of uncertain factors consist in the practical engineering structures universally, such as load and structural random parameters and they have randomicity that is inherent in the practical engineering problem. The loads that are applied to the engineering structures are random, thus which leads to the appearance of random structure system with random parameters. Hence, it is very important for design purposes to study the reliability of random structure system, because the reliability research of random structure system could help the designer to establish acceptable tolerances on structures and control the fluctuations of the system parameters for safe operations. In the end, the reasonable structure could be designed with enough reliability and little cost. At the present time, as an important index of product quality and one of the most important technical indexes, the reliability has been given more attention in engineering. In the practical mechanical components, the most damaged location is usually appeared at the stress concentration position. The stress concentration is an important factor that affects the reliability of mechanical components, and so the research of the corresponding reliability problem has an important theoretic and engineering practical significance in the case of the stress concentration. Neural network (NN) is a complex network system consists of many nerve centers with different frame. The form and function of each nerve center is simple, but the network system could realize different complex functions. Because NN is a complex non-linear dynamics system with many eximious characteristics, such as self-organizing, self-adapting and self-learning abilities, and so NN has been used successfully in different studies. In the structure reliability problem considered in the case of the stress concentration, on the basis of the structural reliability research actuality, neural network technology is applied to the reliability analysis field in the paper. And the NN with Back Propagation (BP) algorithm is adopted, because it is the most adult and is used widely in practical engineering. BP neural network could approach any continuous function and gain higher simulation precision. At present, more attention has been given in the stress concentration problem, and modern structural reliability theory and neural network technique have been developed rapidly. But it is difficult for the structural reliability to be solved in the case of the stress concentration. Neural network (NN) technique has been used in this paper to simulate the relationship between the basic random variables and the stress concentration factors, and the explicit expression of the stress concentration factors can be obtained directly. On the basis of this, the expression of the stress concentration factors given by NN technique and the reliability theory have been combined to solve the structural reliability problems in the case of the stress concentration, and then the corresponding methods of reliability analysis have been presented. The paper has five chapters in all and the main productions obtained are following as: Chapter One Introduction This chapter detailedly introduces the stress concentration of the practical engineering structure, uncertain factors of the engineering design, the structure reliability and the development of neural network technology and gives the research content of this paper. Chapter Two The basic theory of reliability and the base of neural network This chapter expounds the basic theory of reliability and neural networktechnology, neural network model, neural network structure, neural network merit, the basic character of neural network etc. in all. Chapter Three Neural network–based structural reliability analysis of the local stress concentration In the study of the local stress concentration, more attention has been focused on the confirmation of stress concentration factors at present, and has been seldom done on the corresponding reliability problem in the world. However, in the practical engineering structure, the local stress concentration is an important matter that leads to the structural failure. For the stress concentration is an important factor that affects the structure reliability, the research of the corresponding reliability problem has an important theoretic and engineering practical significance in the case of the stress concentration. In the reliability problem considered in the case of the stress concentration, the research work has been done in this paper. Firstly, neural network technique has been used to simulate the expression of the stress concentration factors, the expression obtained by NN is explicit and high precision could be gained with NN's eminent characteristics. Secondly, the explicit expression of the stress concentration factors given by NN and reliability theory have been combined to deal with the reliability problems in the case of the stress concentration, and then the corresponding methods of reliability analysis have been investigated in the work. Chapter Four FEM-NN-MCS-Based Estimation of Stress Concentration Factors in Reliability Analysis For the reliability analysis of complex structures, the stochastic finite element method (SFEM) has been applied widely in practical engineering at the present. SFEM could solve normal distribution parameters and low non-linear problems effectively, but it is difficult to deal with non-normal distribution parameters and high non-linear problems. The finite element method (FEM), neural network technique (NN) and the Monte Carlo simulation method (MCS) have been combined to solve the structural reliability problems in the case of the stress concentration, so a FEM-NN-MCS method has been presented for the reliability analysis of complex structures in this paper. The method could effectively deal with random distribution parameters and high non-linear problems, at the same time, which could solve the precision problem exited from theoretical viewpoint. In this paper, the finite element method (FEM) is used to gain the SCF database as an alternative approach. Although the numerical solution of the SCF...