Numerically efficient algorithm for compact microwave device optimization with flexible sensitivity updating scheme

Abstract

An efficient trust-region algorithm with flexible sensitivity updating management scheme for electromagnetic (EM)-driven design optimization of compact microwave components is proposed. During the optimization process, updating of selected columns of the circuit response Jacobian is performed using a rank-one Broyden formula (BF) replacing finite differentiation (FD). The FD update is omitted for directions sufficiently well aligned with the recent design relocation. As the algorithm converges, the alignment threshold is gradually reduced so that the condition for using BF becomes less stringent. This allows for further reduction of the number of EM simulations involved in the optimization process. The presented flexible Jacobian update scheme allows for considerable reduction of the computational cost with only slight degradation of the design quality. Robustness of the presented algorithm is validated through multiple optimization runs from random initial designs. The verification experiments are conducted for a range of microwave components, including a compact microstrip coupler as well as a threesection compact microwave resonant cell-based impedance transformer. The effects of the alignment threshold value on the computational efficiency of the algorithm and the design quality are investigated. Significant savings reaching 50% as compared to the reference algorithm are demonstrated. K E Y W O R D SBroyden update, design closure, EM-driven design, gradient search, microwave optimization, trust-region framework

Improved Modeling of Microwave Structures Using Performance-Driven Fully-Connected Regression Surrogate

“…The methods for accelerating EM-driven design procedures have been researched for over two decades [6], [7]. Available approaches include incorporation of adjoint sensitivities [8], [9], or sparse Jacobian updates [10], [11] into gradient-based algorithms, utilization of custom EM solvers [12], as well as mesh deformation techniques [13]. Other methods rely on variable-fidelity or variable-resolution techniques [14], [15], often combined with physics-based surrogate modeling procedures (e.g., space mapping [16], response correction [17], Bayesian model fusion [18], cokriging [19]).…”

Accelerated multiobjective design of miniaturized microwave components by means of nested kriging surrogates

“…On a practical side, multiobjective design is often handled by reformulating the problem into a single‐objective task . The reason is strictly opportunistic: the abundance of single‐objective optimization methods and their reliable implementations .…”

“…On a practical side, multiobjective design is often handled by reformulating the problem into a singleobjective task. 35 The reason is strictly opportunistic: the abundance of single-objective optimization methods and their reliable implementations. [36][37][38] The mentioned reformulation can be realized either through objective aggregation (weighted sum methods 39 and goal attainment method 40 ) or by casting all but a primary goal into constraints.…”

Rapid Optimization of Compact Microwave Passives Using Kriging Surrogates and Iterative Correction

“…A conceptually straightforward approach (yet complex implementation-wise) is to incorporate adjoint sensitivities into gradient-based algorithms [15], [16] although availability of this technology is limited throughout commercial simulation packages. Another strictly algorithmic method is to suppress finitedifferentiation sensitivity updates, e.g., by monitoring Jacobian variability [17] or design relocation [18]. In recent years, utilization of fast surrogate models has been gaining considerable popularity.…”