Research On Modeling And Fast Optimization For Skeletal Structure Of Simplified Car Body

Automotive industry, "Eleventh Five-Year Plan", targetedly put forward that:"Large Automotive Group must have a platform for independent product, assembly of R & D capabilities; Backbone enterprises must have the ability to match among car body, powertrain and chassis; Parts enterprises must master the major powertrain components and key core technologies and R & D capabilities with the platform synchronization and the development of own-brand cars was encouraged". Structural design of car body has become indispensable technology of China's self-developed car to build the core competitiveness of enterprises. Especially in recent years, due to energy shortages and fierce market competition, lightweight body has become a hot topic of the current global as well as the development of China's automobile industry. During the conceptual design stage (CDS), the modeling and fast optimization of for skeletal structure of simplified car body are investigated in this thesis, supported by National Natural Science Foundation of China (No.: 50975121), Science and Technology Development Plan of Jilin province (No.: 20096004), Graduate Innovation Fund of Jilin University (No.: 2009-2007) and FAW Group Science and Technology Innovation Project (No.: 0837 and 093715).In the past topological relationships of carrying parts and some basic and important parameters of car body were considered during the CDS, however, As the body design technology has matured, as well as a gradual increase in computer-aided design capabilities, a number of important properties during the detailed design stage of the car body, such as thin-walled beam, cross-sectional shape design, cross-sectional mechanical properties optimization, and the research of joint flexibility and plastic hinge, have been analyzed and optimized in advance during the CDS. How to build the SSSBC more accurately reflect the actual performance of the car body is the foundation and guide for the conceptual design and directly related to the correctness of the follow-up finite element analysis and structural optimization. So it is necessary to develop specific finite element, targeted car body features.In chapter 2, the thin-walled closed single cell, double cell and open cross-section of the mechanical properties of the formula are deduced. The cross-sectional warping deformation of short and plump thin-walled beam, affected by constraint rotation, is studied. Beam element stiffness matrix with warping degree of freedom is derived. Then, considering the joint flexible, the semi-rigid beam element, respectively, connected at both ends of three extensional and three rotaional spring elements is innovatively developed. Based on the relationship of internal nodes and external nodes of semi-rigid connection beam elements, internal degrees of freedom are reduced at the element level, so as to facilitate the finite element modeling. By adjusting the spring stiffness, the connection of the thin-walled beams is in the condition of intermediate state of semi-rigid connections between ideal pinned and rigid connection. Using this beam element, the modeling of SSSBC is more precise. In addition, a triangular plate element is developed to simulate the body floor, roof, rear fender, after hoardings as well as the glass sheet structure. From the introduction of thin plate bending theory, the displacement shape function and stiffness matrix of the element are derived, as well as the construction of Hammer numerical integration scheme are given in detail. Finally, the cross-sectional properties and joint stiffness are extracted from the detailed finite element model of a car body and assigned to the SSSBC. Based on adjoint variable method the sensitivity of static stiffness for cross-sectional parameter of beam element is analyzed. The sensitivity of cross-sectional area is much smaller than that of moment of inertia, and therefore the optimization of mechanical parameters of car body can not take into account cross-sectional area parameters, and only consider changes of the moment of inertia I y , I z,J on the stiffness of the car body. This conclusion provides a basis for mathematical modeling of the optimization problems for SSSCB of the next chapter.In chapter 3, two-level and collaborative optimization model for SSSCB is researched under the displacement and frequency constraints. At present a lot of literatures, on the optimization of truss structures, focus on all kinds of optimization algorithms applied to the simple truss optimization; while few literatures are pertaining to the optimization of beam frame structure. Qian Lingxi et al stated:" This is a practical optimization problem with multiple variables." In order to handling the thorny multiple design variables of beam element, A common approach is to assign a specific cross section (rectangular pipe, circular, shaped, etc.) to the beam element or selecting from a cross-sectional library. When the designer of the car body designs the cross-sectional geometry, they not only considers the design of mechanical property but also consider the shape, layout and other factors, such as the A column, which is necessary to meet the requirements of mechanical properties, while cross-sectional geometry can not obstruct the driver's sight. A group of cross-sectional mechanical properties can correspond to multiple sets of cross-sectional geometry, and then the designer consolidates other factors to select the corresponding cross-sectional geometry. Therefore, providing the optimal cross-sectional mechanical properties meeting design goals (mass, stiffness, frequency) has very important significance. Structural optimization of SSSCB based on mechanical property of element is presented in this thesis. The weight objective of car body is translated into constraint and design variables are translated into the objective. Then, a multi-objective optimization model is obtained to minimize the variable of cross-sectional mechanical properties. Semi-rigid connection stiffness ri of beam element is associated with the cross-sectional moment of inertia I y , I z,J . It is very difficult to forecast the relationship between connection stiffness and the moment of inertia of beam element. This article assumes that joint flexibility is proportional to the moment of inertia of beam element, so the joint flexibility can be expressed by the moment of inertia. Next, the level of the cross-sectional geometry optimization is carried out, which should transfer the optimal mechanical property as constraints, and the smallest cross-sectional area as the objective function. Optimal cross-sectional area returns to the optimization level of mechanical properties. Two level optimization are iterated until convergence. SSSCB is of small-scale structures (less than 500 degrees of freedom), so the finite element analysis for each calculation is smaller. As a result, the search capabilities of optimization algorithm should be the first place. Improved nondominated sorting genetic algorithm is introduced to solve this two-level and collaborative optimization model. In this paper, an implicit constraints handling approach is presented based on sort and superior relationship without penalty function. By the handling of implicit constraints, feasible and the non-feasible solution ultimately maintain a balance. Population has a good diversity so as to search out the global optimal solution. Numerical examples verify the effectiveness of the algorithm.In chapter 4, the problem of the static and dynamic reanalysis is described first. Fox's multi-point approximation reanalysis and combined approximation reanalysis are introduced, and the convergence of the latter has been studied. Taking the advantages of multi-point approximate reanalysis and combined approximation reanalysis, this chapter proposes the innovative multi-point combination approximate reanalysis algorithm, which has a true sense of the global approximation and local approximation characteristics. Then the error assessment of the static and dynamic reanalysis problem are introduced to facilitate the combination of reanalysis algorithms and the structural optimization. Numerical examples show that the proposed algorithm is more accurate than the CA method, but the stability of the algorithm remains to be discussed further.In chapter 5, an adaptive reanalysis approach for GA structural optimization, extended from Kirsch's CA method, is presented. The rules for selecting initial number of basis vectors are given. The required number of basis vectors for each generation can be adaptively determined, so a bridge between reanalysis method and GA is built. For problems of large scale, the trade-off crossover and mutation probability is selected to keep the population diversity and enhance the efficiency of reanalysis-based GA. Form the results of three examples it is possible to say that reanalysis-based GA is a very efficient and high-quality approximate optimization method. Adaptive reanalysis is more suitable for GA than traditionally gradient optimization algorithms, because of the GA's population searching characteristic.In chapter 6, Structural optimization with frequency constraints is highly nonlinear dynamic optimization problems, so optimal criteria (OC) method, which is easily trapped into the local optima, is difficult to search the global optimal solution. On the other hand, genetic algorithm (GA) for structural optimization is very robust, but it is also computationally intensive and hence slower than OC methods. To solve this problem a hybrid OC-GA method for truss optimization is presented in this paper. First the optimality criterion is developed for multiple frequency constraints based on differentiation of the Lagrangian function. Then, upon the sensitivity analysis, the most efficient variables are identified and modified as iteration scheme. Finally, OC method serves as a local search operator and is integrated with GA. The numerical results indicate that the hybrid OC-GA has powerful capacity in searching for more optimal results and requiring less computational effort.In chapter 7, the requirements of the CAE analysis and optimization platform of SSSCB is analyzed and then system-level structure is also divided. The whole platform is divided into a relatively independent five-leve framework: data management layer, the body design layer, finite element analysis layer, general optimization layer and data output layer. The message and parameters are transmitted among these five levels by a well-defined interface. Based on Managed DirectX, a graphics engine has been developed for 3D rendering of SSSCB. Data structure of a double-linked list is coded, and object-oriented finite element program and optimization algorithms are programmed. The proposed object-oriented optimization framework is divided into three components: a module component, an optimizer component, and a simulation component. Finally, a graphical user interface is developed to interact among graphics, finite element analysis, structural optimization and users.The final chapter is the summary and prospects for the whole thesis.