Evidence Hub

Surrogate-Assisted Multiobjective Co-Design of Fractional-Order Aerospace DC–DC Converters under Thermal, Transient, and Fault-Survivability Constraints

Demi̇r T, koçyiğit A

2026

Abstract

Abstract More-electric aircraft power converters are designed under competing requirements rather than a single performance target: rapid bus-voltage recovery, low semiconductor loss, moderate passive-component size, thermal headroom, and fault ride-through must be achieved simultaneously. This paper formulates these requirements as a six-variable co-design problem for a three-phase interleaved 270 V-to-28 V, 800W DC–DC converter controlled by a fractional-order P Iλ law. The decision vector contains switching frequency, phase inductance, output capacitance, fractional integral order, and two controller gains. A physics-based averaged model supplies the training data, while a machine-learning layer is used only to accelerate designspace screening. Five thousand Latin-hypercube designs are evaluated with the nonlinear averaged model; 3595 satisfy the imposed dynamic and hardware constraints. An ensemble feasibility classifier attains 96.4% test accuracy, 99.0% precision, and 96.0% recall. A multilayer surrogate for the transient objective has a mean absolute percentage error of 7.18% and a median error of 3.56%; because its accuracy deteriorates near saturation boundaries, every reported Pareto design is re-evaluated with the original physics model. Screening 60,000 candidates yields 26,209 high-confidence feasible points, from which 68 trade-off designs are retained by a bank of Dirichlet scalarizations and subsequently verified. A balanced design at fs = 246.23 kHz, L =…

Read the paper

Related studies

Go deeper in Evidence Hub

Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.

Open this study Sign up free