Surrogate‐assisted design optimization of photonic directional couplers

Abstract

SUMMARYIn this paper, computationally efficient simulation-driven design optimization of integrated photonic couplers is discussed. The design process is conducted using variable-fidelity electromagnetic (EM) simulations and surrogate-based optimization (SBO) techniques. Two design case studies are considered: conventional directional coupler and a curved coupler. A particular choice of the SBO approach relies on a visual inspection of the low-and high-fidelity coupler model responses as well as their correlations. The first structure is optimized using adaptively adjusted design specifications technique. Because of limited correlations between the high-and low-fidelity EM models of the second structure, its optimal design is obtained using the additive response correction method. Subsequently, the design is fine-tuned using local response surface approximations. As demonstrated through numerical results, utilization of the SBO techniques allows for considerable reduction of the design optimization time compared with direct handling of the high-fidelity model. The total optimization cost of conventional and curved coupler corresponds to only 13.5 and 23.7 high-fidelity model simulations, respectively.

Rapid simulation-driven design of compact photonic Y-junction by variable-dimensional sequential approximate optimization

“…Set n = 3; 2. Find the optimized design of the Y-junction x k * = argmin{x n : U(R(x n ))} using pattern search [4]; 3. Set n = n + 3; 4.…”

“…The first approximation is obtained through direct search but carried out for low-dimensional parameterization of only four variables. The RSA model is constructed as a second-order polynomial without mixed terms and the sampling scheme is the star distribution [4] that involves only 2k + 3 samples. The total cost of the above optimization scheme is a few dozen of evaluations of the EM model R in Step 2 of the process, and 180 evaluations for the sequential approximate optimization process (Steps 3-7).…”

“…These difficulties can be alleviated using unconventional structure topologies which, however, can be obtained only by means of simulation-driven design methods. Unfortunately, reliable evaluation of photonic structure responses requires high-fidelity EM simulations and thus their design using conventional techniques is often computationally infeasible [4].…”

Inverse surrogate modeling for low-cost geometry scaling of microwave and antenna structures

“…EC 33,4 as well asin microwave and antenna engineeringadaptively adjusted design specifications (Bekasiewicz and Koziel, 2015b).…”

Editorial for the special issue on advances in simulation‐driven modeling and optimization of microwave/RF circuits

“…In Rayas‐Sanchez et al, EM‐based optimization of microstrip and SIW interconnects is presented using low‐order polynomial surrogates. Application of adaptively adjusted design specification technique and sequential approximate optimization is considered in Bekasiewicz and Koziel in the context of design optimization of photonic directional couplers. Cost‐efficient design of microstrip patch antennas is demonstrated in Aoad et al using inverse artificial neural networks models.…”