Surrogate-Assisted Reliability-Based Design Optimization in a Composite Structure under Constraints

Kumar, Dinesh; Verma, Richa; Kobayashi, Kazuma; Alam, Syed Bahauddin

Permalink

https://hdl.handle.net/2142/121818 Copy

Description

Title

Surrogate-Assisted Reliability-Based Design Optimization in a Composite Structure under Constraints

Author(s)

• Kumar, Dinesh
• Verma, Richa
• Kobayashi, Kazuma
• Alam, Syed Bahauddin

Issue Date

2023

Keyword(s)

• Surrogate modeling
• RBDO
• Composite structure
• Optimization
• Kriging

Abstract

These days, computational models are used in designing engineering components. Taking different sources of uncertainties in the optimization process is necessary for an optimal and realistic design using computer simulations. Without considering these uncertainties in the design process, the manufactured structure of an engineering component might fail when used in different conditions. Reliability-based design optimization (RBDO) are very advanced algorithm for computationally designing robust engineering applications. However, only a little work has been seen in composite applications using RBDO. In the RBDO process, different sources of uncertainties can be included in the optimization process in the early stage of a design process. The RBDO process is computationally very time-consuming and needs a lot of data storage. Usually, the RBDO process requires a vast number of computer simulations. In literature, numerous methods for RBDO have been proposed. These methods combine robustness analysis with optimization algorithms. In this work, to avoid running a large number of simulations, we first construct a surrogate model and use it for the robustness analysis. For surrogate modeling, the Kriging method is used. Further, for RBDO, sequential quadratic programming (SQP), a gradient-based approach, is combined with the Kriging method. In terms of application, we have used a composite beam. FEM solver Abaqus is used for composite structure analysis. An in-house code for surrogate modeling is combined with the sequential quadratic programming algorithm and Abaqus using Python script. The uncertainties are introduced in the material properties, and the cost function is considered the composite material's strength. Finally, the convergence of the cost function and constraints are shown with iteration. The optimal results are compared with the case without consideration of uncertainties.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121818

Owning Collections