2018Winter Simulation ConferenceRequires access

Initial sampling using multi-fidelity information in simulation optimization of manufacturing systems

Ziwei Lin, Shichang Du, Andréa Matta

Open publisher page 1 citations

Abstract

High-fidelity models are capable of providing accurate estimates but slow in execution. On the other hand, estimates provided by low-fidelity models are biased but fast. The knowledge embedded in low-fidelity models might be helpful for simulation optimization algorithms. Several multi-fidelity modeling algorithms have been proposed in literature, whereas currently only high-fidelity information is used in the initial sampling phase. This poster provides an algorithm to allocate high-fidelity budgets using multi-fidelity information in order to contain a fixed number of good solutions in the initial design. Results show that the proposed sampling policy can allocate more budgets in promising areas.

About this research paper

What this paper is about

High-fidelity models are capable of providing accurate estimates but slow in execution. On the other hand, estimates provided by low-fidelity models are biased but fast. The knowledge embedded in low-fidelity models might be helpful for simulation optimization algorithms. Several multi-fidelity modeling algorithms have been proposed in literature, whereas currently only high-fidelity information is used in the initial sampling phase. This poster provides an algorithm to allocate high-fidelity budgets using multi-fidelity information in order to contain a fixed number of good solutions in the initial design. Results show that the proposed sampling policy can allocate more budgets in promising areas.

Why it matters

OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

High-fidelity models are capable of providing accurate estimates but slow in execution. On the other hand, estimates provided by low-fidelity models are biased but fast. The knowledge embedded in low-fidelity models might be helpful for simulation optimization algorithms. Several multi-fidelity modeling algorithms have been proposed in literature, whereas currently only high-fidelity information is used in the initial sampling phase. This poster provides an algorithm to allocate high-fidelity budgets using multi-fidelity information in order to contain a fixed number of good solutions in the initial design. Results show that the proposed sampling policy can allocate more budgets in promising areas.

Key concepts: Fidelity, High fidelity, Computer science, Sampling (signal processing), Mathematical optimization, Engineering, Mathematics, Telecommunications

Related papers

Back to paper searchBrowse research topicsOriginal source
Initial sampling using multi-fidelity information in simulation optimization of manufacturing systems — Research Paper | ScholarLens