Black-box topology optimization (BBTO) uses evolutionary algorithms and other soft computing techniques to generate near-optimal topologies of mechanical structures. Although evolutionary algorithms are widely used to compensate the limited applicability of conventional gradient optimization techniques, methods based on BBTO have been criticized due to numerous drawbacks. In this paper, we discuss topology optimization as a black-box optimization problem. We review the main BBTO methods, discuss their challenges and present approaches to relax them. Dealing with those challenges effectively can lead to wider applicability of topology optimization, as well as the ability to tackle industrial, highly-constrained, nonlinear, many-objective and multimodal problems. Consequently, future research in this area may open the door for innovating new applications in science and engineering that may go beyond solving classical optimization problems of structural engineering. Furthermore, algorithms designed for BBTO can be added to existing software toolboxes and packages of topology optimization.
The work at hand addresses engineers, designers and scientists who face the challenging Task of devising concept structures in a virtual product design process that involves more and more sophisticated physical simulations. Using methods of evolutionary optimization and machine learning, this dissertation explores a novel generic topology optimization algorithm, which is able to provide concept designs even for problems involving complex, black-box simulations. A self-contained learning component utilizes physical simulation data to generate a search direction. The generic topology optimization is studied in conjunction with statistical models such as neural networks or support vector regression. In empirical experiments, the novel method reproduces reference structures with Minimum compliance and provides innovative solutions in the domain of vehicle crashworthiness optimization.
Contents
Symbols and Abbreviations XIV
Abstract XVII
Zusammenfassung XVIII
1 Introduction 1
2 Fundamentals...
In this paper we propose a novel method for the topology optimization of mechanical structures, based on a hybrid combination of a neuro-evolution with a gradient-based optimizer. Conventional gradient-based topology optimization requires problem-specific sensitivity information, however this is not available in the general case. The proposed method substitutes the analytical gradient by an artificial neural network approximation model, whose parameters are learned by an evolutionary algorithm. Advantageous is that the number of parameters in the evolutionary search is not directly coupled to the mesh of the discretized design, potentially enabling the optimization of fine discretizations. Concretely, the network maps features, obtained for each element of the discretized design, to an update signal, that is used to determine a new design. A new network is learned for every iteration of the topology optimization. The proposed method is evaluated on the minimum compliance design problem, with two different sets of features. Feasible designs are obtained, showing that the neural network is able to successfully replace analytical sensitivity information. In concluding remarks, we discuss the significant improvement that is achieved when including the strain energy as feature.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.