Details

Autor(en) / Beteiligte

• Igarashi, Hajime,

Titel

Topology Optimization and AI-Based Design of Power Electronic and Electrical Devices : Principles and Methods

Auflage

First edition

Ort / Verlag

Cambridge, MA : Academic Press,

Erscheinungsjahr

[2024]

Beschreibungen/Notizen

• Includes bibliographical references and index.
• Front Cover -- Topology Optimization and AI-based Design of Power Electronic and Electrical Devices -- Copyright -- Contents -- Preface -- Nomenclature -- 1 Equations of electromagnetic field -- 1.1 Maxwell equations -- 1.2 Conservation laws -- 1.2.1 Conservation of electric charge -- 1.2.2 Conservation of energy -- 1.2.3 Conservation of momentum -- 1.3 Static fields -- 1.3.1 Electrostatic field -- 1.3.2 Magnetostatic field -- 1.4 Quasistatic fields -- 1.4.1 Magneto-quasistatic field -- 1.4.2 Electro-quasistatic field -- 1.4.3 Magneto- and electro-quasistatic approximations -- 1.5 Electromagnetic waves -- 1.5.1 Wave equation -- 1.5.2 Displacement current in a power device -- 1.6 Boundary conditions -- 1.6.1 Boundary conditions on a material surface -- 1.6.2 Electric and magnetic walls -- 1.6.3 Periodic boundary condition -- 1.6.4 Boundary conditions for wave propagation -- 1.6.5 Impedance boundary conditions -- 1.7 Summary -- 2 Modeling of electromagnetic systems -- 2.1 Permanent magnet (PM) -- 2.2 Energy and force -- 2.2.1 Magnetic field -- 2.2.2 Electric field -- 2.3 Inductance -- 2.3.1 Definition of inductanace -- 2.3.2 Differential inductance -- 2.3.3 Effect of an air gap -- 2.4 Skin and proximity effects -- 2.5 Loss analysis -- 2.5.1 Classical eddy current loss -- 2.5.2 Hysteresis loss -- 2.5.3 Steinmetz equation -- 2.5.4 Steinmetz equation in time domain -- 2.5.5 Eddy current loss considering skin effect -- 2.5.6 Mathematical models of magnetic hysteresis -- 2.6 Modeling of electric motors -- 2.6.1 Circuit equation of moving object and electromagnetic forces -- 2.6.2 Equations for PM motors -- 2.6.3 d-q transformation -- 2.6.4 Motor control -- 2.6.5 Behavior model of an electric motor -- 2.6.6 Torque component separation -- 2.7 Summary -- 3 Finite element method for electromagnetic field -- 3.1 Two-dimensional analysis.
• 3.1.1 Two-dimensional magnetostatic field -- 3.1.2 Consideration of magnetic saturation -- 3.1.3 Coupling with an electric circuit -- 3.1.4 Treatment of a permanent magnet -- 3.2 Three-dimensional analysis -- 3.2.1 Thee-dimensional electrostatic field -- 3.2.2 Thee-dimensional magnetostatic field -- 3.2.3 Edge elements -- 3.2.4 Finite element analysis with an edge element -- 3.2.5 Compatibility -- 3.2.6 Thee-dimensional magneto-quasistatic field -- 3.2.6.1 Time-domain analysis -- 3.2.6.2 Numerical stability in time-domain analysis -- 3.2.7 Analysis of a three dimensional electro-quasistatic field -- 3.2.8 Analysis of the three-dimensional wave equation -- 3.3 Finite elements -- 3.3.1 Simplex elements -- 3.3.2 Hexahedral element -- 3.3.3 Other finite elements -- 3.4 Computation of electromagnetic force -- 3.5 Summary -- 4 Numerical methods for electromagnetic field analysis -- 4.1 Homogenization method -- 4.1.1 Homogenization of a laminated steel plate -- 4.1.2 Ollendorff formula -- 4.1.2.1 Macroscopic permeability -- 4.1.2.2 Other homogenization methods -- 4.1.3 Homogenization of a winding coil -- 4.1.4 Unit cell approach -- 4.1.4.1 Linear system -- 4.1.4.2 Consideration of magnetic saturation -- 4.1.5 Soft magnetic composite (SMC): Homogenization of a heterogenous material -- 4.1.5.1 Modeling of SMC -- 4.1.5.2 Modeling with discrete element method -- 4.1.6 Expression using an equivalent circuit -- 4.1.6.1 Laminated steel sheets -- 4.1.6.2 Winding coil -- 4.1.6.3 Equivalent circuit for electromagnetic devices -- 4.1.6.4 Physical interpretation of a Cauer circuit -- 4.2 Model-order reduction -- 4.2.1 Principal component analysis -- 4.2.2 Proper orthogonal decomposition (POD) -- 4.2.3 Equivalent circuit obtained via PVL method -- 4.2.3.1 Formulation of PVL method -- 4.2.3.2 Synthesis of equivalent circuits -- 4.2.4 Direct synthesis of a Cauer circuit.
• 4.2.4.1 CVL method -- 4.2.4.2 Simple example -- 4.3 Summary -- 5 Optimization methods -- 5.1 Introduction -- 5.2 Basics of deterministic methods -- 5.2.1 Mathematical properties -- 5.2.2 Steepest descent method -- 5.2.3 Adjoint variable method -- 5.3 Method of Lagrange multiplier -- 5.3.1 Equality-constrained minimization problem -- 5.3.2 Inequality-constrained minimization problem -- 5.3.3 Augmented Lagrangian method -- 5.3.4 Numerical example -- 5.4 Method of moving asymptotes -- 5.4.1 Principle and method -- 5.4.2 Simple example -- 5.5 Genetic algorithm -- 5.5.1 Principle and method -- 5.5.1.1 Design of chromosome -- 5.5.1.2 Algorithm -- 5.5.1.3 Building block hypothesis -- 5.5.2 Real-coded genetic algorithm -- 5.5.3 Real-coded ensemble crossover -- 5.5.4 Micro-genetic algorithm -- 5.5.5 Robust genetic algorithm -- 5.5.6 Consideration of constraints -- 5.5.7 Numerical examples -- 5.6 Covariance Matrix Adaptation Evolution Strategy: CMA-ES -- 5.6.1 Normal distribution -- 5.6.2 Geometry of Gaussian function -- 5.6.3 Principle and method of CMA-ES -- 5.6.4 Treatment of constraints in CMA-ES -- 5.6.5 Numerical example 1: Optimization of magnetization distribution -- 5.6.6 Numerical example 2: Comparison of GA and CMA-ES -- 5.7 Genetic algorithm for multi-objective optimization -- 5.7.1 Principle and method -- 5.7.2 Non-dominated sorting genetic algorithm: NS-GAII -- 5.7.3 Treatment of constraints -- 5.7.4 Numerical example -- 5.8 Simulated annealing -- 5.8.1 Principle and method -- 5.8.2 Quantum and emulated quantum annealing -- 5.9 Summary -- 6 Topology optimization -- 6.1 Introduction -- 6.1.1 Features of parameter and topology optimization -- 6.1.2 Comparison of PO with TO -- 6.2 Topology optimization (TO) methods -- 6.2.1 Overview -- 6.2.2 Density method -- 6.2.3 Level-set method -- 6.2.4 Naive ON-OFF method.
• 6.2.5 Numerical example of naive ON-OFF method -- 6.2.6 Hybridization of ON-OFF and level-set methods -- 6.3 TO based on Gaussian basis functions -- 6.3.1 Principle and methods -- 6.3.2 Numerical example: PM motor -- 6.3.2.1 Single-objective optimization -- 6.3.2.2 Multi-objective optimization -- 6.3.3 Numerical example: experimental validation for PM motor model -- 6.3.4 Numerical example: wireless power transfer -- 6.3.5 Numerical example: wireless power transfer considering eddy currents -- 6.3.6 Numerical example: microstrip lines -- 6.4 Advanced TO using Gaussian basis functions -- 6.4.1 Consideration of a motor-control system -- 6.4.2 Consideration of mechanical strength -- 6.4.3 Hybridization of TO and PO -- 6.4.4 Multi-material optimization -- 6.4.5 2.5D topology optimization -- 6.5 Discussions -- 6.5.1 Comparison of topology optimization methods -- 6.5.2 Challenging problems in topology optimization -- 6.6 Summary -- 7 Basics of machine learning -- 7.1 Introduction -- 7.2 What is a surrogate model -- 7.3 When surrogate models are effective -- 7.4 Offline and online surrogate models -- 7.4.1 Curse of dimensionality -- 7.4.2 Determination of hyper-parameters in a surrogate model -- 7.4.3 Sampling -- 7.5 Least squares method -- 7.6 Minimum norm solution and generalized inverse matrix -- 7.7 Method of maximum likelihood -- 7.7.1 Application to least squares method -- 7.7.2 Application to classification -- 7.8 Response surface methods -- 7.9 Neural networks -- 7.9.1 Learning based on steepest descent method -- 7.9.2 Back propagation -- 7.9.3 Error functions for NN -- 7.9.4 Classification using NN -- 7.10 Regression tree -- 7.10.1 Principle and method -- 7.10.2 Tree-based methods -- 7.11 Numerical examples 1: optimal design using neural network -- 7.12 Numerical examples 2: comparison of surrogate models -- 7.13 Bayesian optimization.
• 7.13.1 Gaussian process regression -- 7.13.2 Derivation of expected value and variance -- 7.13.3 Numerical examples -- 7.13.3.1 Function of a single variable -- 7.13.3.2 Optimization of magnetic shield -- 7.14 Summary -- 8 Optimal design based on machine learning -- 8.1 Direct inverse modeling -- 8.1.1 Principle and method -- 8.1.2 Numerical example -- 8.2 Adaptive surrogate model -- 8.2.1 Principle and method -- 8.2.2 Numerical example -- 8.3 Integrated design using Monte Carlo tree search -- 8.3.1 Principle and method -- 8.3.2 Numerical example 1: PM motor -- 8.3.3 Numerical example 2: DC-DC converter -- 8.4 Summary -- 9 Optimal design based on deep learning -- 9.1 Introduction -- 9.2 Convolutional neural network (CNN) -- 9.3 Fast optimization using deep learning -- 9.3.1 Classification of torque performances -- 9.3.2 Optimization based on classification -- 9.4 Regression based on material and magnetic field distributions -- 9.5 Regression of torque and flux functions -- 9.6 Explainable AI -- 9.7 Variational autoencoder -- 9.7.1 Kullback-Leibler divergence -- 9.7.2 Variational autoencoder -- 9.7.3 Numerical examples -- 9.7.4 Application of variational autoencoder -- 9.8 Summary -- A Maxwell stress tensor in orthogonal curvilinear coordinates -- B Newton-Raphson method -- C Differential forms -- D Mathematical properties of FE matrices -- D.1 Gradient and rotation matrices -- D.2 Condition required for interpolation functions -- D.3 A method and A-φ method -- Bibliography -- Index -- Back Cover.
• Description based on print version record.