No abstract

In the development of complex engineering systems, many engineers from different disciplines collaborate to identify feasible designs that satisfy all system requirements. Analytical target cascading (ATC) is a method for the design optimization of hierarchical, multilevel systems and has been successfully employed in the design of complex engineering systems. In this paper, we propose a novel data-driven set-based ATC (SBATC) method for hierarchical design optimization problems using machine learning techniques. The proposed SBATC offers two advantages in engineering processes. First, it decomposes hierarchical design optimization problems into sets of suboptimization problems. The feasible regions of the suboptimization problems are explored and cascaded to lower-level optimization problems. Using the set-based approach, couplings between two adjacent levels in the optimization process are not required. Second, an efficient strategy is employed to determine feasible regions based on Bayesian active learning (BAL). In BAL, the Gaussian process (GP) of all cost functions is trained. An acquisition function that combines the probability of feasibility and entropy search is evaluated using posterior distributions of the trained GP. The acquisition function is maximized to generate new sampling points around the feasible regions by balancing the exploitation and exploration of the design space. To verify the effectiveness of the proposed method, numerical examples of hierarchical optimization problems are evaluated.

The purpose of this paper is to propose a new Set-based concurrent engineering method using Bayesian active learning and to show an application to a multi-disciplinary design optimization problem. In the early stages of the system design process, it is required to set a target value considering the uncertainty of design conditions. If any change of design condition occurs by an external factor in the later development process, the predefined target value cannot be held, and critical rework can be inevitable. To avoid this issue, it is important in the early design stage to solve not only a single target solution but also feasible design solutions that satisfy all multi-disciplinary requirements. In order to discover the feasible region with limited resources, an efficient sampling strategy using CAE simulation is necessary. In this study, a sampling strategy based on Bayesian active learning is proposed to discover a feasible region of multi-disciplinary constraints concurrently. In the proposed method, Gaussian Process models of the multi-disciplinary constraints are trained. Based on posterior distributions of trained Gaussian Processes, new acquisition function by combining two different types of acquisition functions, Probability of Feasibility and Entropy Search is proposed and maximized to generate new sampling points to improve the prediction accuracy of feasible region effectively. To show the effectiveness of the proposed Set-based concurrent engineering method to a multi-disciplinary design problem, a suspension design problem is demonstrated.

No abstract

No abstract