Press release of Impress Holdings Co., Ltd. (2024.10.25) Impress Holdings Co., Ltd. Analyze discrepancies and variations between simulation and experimental results and develop superior products! “Quantification of uncertainty for CAE utilization – Surrogate modeling using Gaussian process regression and design of experiments” On October 25, 2024, Kindai Kagakusha Co., Ltd., which is involved in the publishing business of specialized books in the field of science and engineering in the Impress Group, published the book “Quantification of Uncertainty for CAE Utilization -” from the Kindai Kagakusha Digital Label (*). We have published “Surrogate Modeling Using Gaussian Process Regression and Design of Experiments” (Author: Arisu Toyonori). (*What is Kindai Kagakusha Digital? Kindai Kagakusha is an on-demand publishing label that makes full use of digital technology in collaboration with authors on a project basis.For more information, please see here
https://www.kindaikagaku.co.jp/kdd/scheme/
https://prcdn.freetls.fastly.net/release_image/5875/6247/5875-6247-f1134987723e0a55c5fc8dcb21484faa-1130×1604.jpg ●Bibliographic information [Book title] Quantification of uncertainty for CAE utilization -Surrogate modeling using Gaussian process regression and design of experiments- [Author] Arihiro Toyonori [Specifications] A5 size, paperback, monochrome print version/partial color electronic version, 244 pages of text [Print version standard price] 3,000 yen (excluding tax) [Electronic version standard price] 3,000 yen (excluding tax) [ISBN] (Hardcover book)
978-4-7649-0714-0 C3041 [ISBN] (POD)978-4-7649-6088-6 C3041 [Product URL] https://www.kindaikagaku.co.jp/book_list/detail/9784764960886/ ●Content introduction At the manufacturing site, Mr. A: “We are consulting with you regarding a new product designed by Mr. B. Actually, 5% non-conformity occurred during shipping inspection.” Mr. B: “I can’t believe it because I’ve confirmed it with simulations many times. Is there a problem with the testing method?” Mr. A: “The measuring instruments used for inspection are calibrated regularly, so there is no problem.” Mr. B: “…” Such conflicts of opinion may occur. The approach to solving these problems is the theme of this book, “quantification of uncertainty.” This book introduces the
mathematical background of “machine learning,” “Gaussian process regression,” and “design of experiments” necessary for uncertainty modeling, and then introduces SmartUQ, a commercial solution that supports uncertainty quantification. I will. We hope that this book will contribute in some small way to solving the problems faced by our readers and making the manufacturing process more efficient and sound. ●Author introduction Yuuko Toyonori Advisor, Measurement Engineering System Co., Ltd. 1972: Graduated from the Department of Mathematical Engineering, Faculty of Engineering, Kyoto University Joined Yokogawa Electric Corporation (currently Yokogawa Electric Corporation) 2009: Retired from Yokogawa Electric Corporation 2013: Current position ●Table of contents Chapter 1 The role of experiment and simulation 1.1 Positioning of experiments and simulations 1.2 Deductive and inductive approaches 1.3 Surrogate model 1.4 Utilization of knowledge and data from real space 1.5 Symbols and notation Chapter 2 Machine Learning 2.1 What is machine learning? 2.2 Specific examples of machine learning 2.3 Machine learning classification 2.4 Curve fitting using polynomials 2.5 Generalization performance and cross-validation Column: Feature extraction Chapter 3: Uncertainty and probability distribution 3.1 What is uncertainty? 3.2 Error 3.3 Variation and bias 3.4 Distribution of measured values 3.5 Probability distribution 3.6 Normal distribution 3.7 Multivariate normal distribution 3.8 Revisited: Least Squares 3.9 Summary Column: Central limit theorem Chapter 4 Linear Regression Model 4.1 Simple regression/multiple regression 4.2 Linear regression 4.3 A simple example of linear regression 4.4 Issues with linear regression models Column: words and feature vectors Chapter 5 From Gaussian process to Gaussian process regression 5.1 Dual
representation of Bayesian estimation 5.2 Gaussian process 5.3 Derivation of Gaussian process regression 5.4 Challenges of Gaussian process regression Column: Gaussian process regression and neural networks Chapter 6 Learning Hyperparameters 6.1 Properties of hyperparameters 6.2 Naive approach to maximum likelihood estimation 6.3 Hyperparameter optimization 6.4 MCMC method 6.5 Various optimization problems using gradients 6.6 Unconstrained problem 6.7 Constrained problems 6.8 Examples of constrained problems Column: Optimization using gradient method Chapter 7 Gaussian process calculation package 7.1 Available Gaussian process calculation packages 7.2 Hyperparameter optimization problem for Gaussian process regression model 7.3 GPML 7.4 GPy 7.5 Various uses of GPy 7.6 GPy implementation example: Men’s 100m world record 7.7 Summary Column: Normalization and standardization of training data Chapter 8 Design of Experiments and V&V Process 8.1 What is experimental design? 8.2 Orthogonal array 8.3 Latin Hypercube Sampling 8.4 Gaussian process regression and surrogate models 8.5 Discussion: Error propagation 8.6 Real-life problems: experiments and simulations 8.7 Quality Assurance Activities and V&V Process 8.8 Validation of experimental and simulation results Column: Standards for various V&V processes Chapter 9 Integrated Solutions for Uncertainty Quantification 9.1 What is SmartUQ? 9.2 Weight reduction and fatigue strength of iron brackets 9.3 NACA Airfoil: Optimization of Aircraft Wing Shape 9.4 Conclusion Appendix A: Formulas for vectors and matrices A.1 Matrix multiplication, transposition, and tracing A.2 Inverse matrix A.3 Differentiation A.4 Determinant A.5 Eigenvalues, eigenvectors A.6 Definiteness of real symmetric matrices Appendix B Formulas for normal and multivariate normal distributions B.1 Normal distribution B.2 Multivariate normal distribution Appendix C’Nonlinear Programming Formulas C.1 Vector differential operator C.2 Convex functions and convex sets [Kinda Kagakusha Digital]
https://www.kindaikagaku.co.jp/kdd/index.htm Kindai Kagakusha Digital is a 21st century science and technology publishing label promoted by Kindai Kagakusha Co., Ltd. By actively leveraging digital power, we propose a speedy, on-demand, and sustainable publishing model. [Kindaikagakusha Co., Ltd.] https://www.kindaikagaku.co.jp/ Kindai Kagakusha Co., Ltd. (Headquarters: Chiyoda-ku, Tokyo, Representative Director and President: Hiroaki Otsuka) was founded in 1959. We are developing a publishing business that covers a wide range of specialized fields of science and engineering, including academic books centered on mathematics, mathematical sciences, information science, and information engineering, as well as textbooks for science and engineering universities. In order to meet modern needs that require not only basic knowledge of natural science but also its advanced use, we cover a wide range of subjects, from classics to the latest interdisciplinary fields. In addition, we collaborate with major academic societies, associations, and prominent research institutions to pursue an academic level that will become a global standard. [Impress Group]
https://www.impressholdings.com/ A media group whose holding company is Impress Holdings Co., Ltd. (Headquarters: Chiyoda-ku, Tokyo, Representative Director: Daisuke Matsumoto, Stock Code: TSE Standard Market 9479). We are developing highly specialized media & services and solution businesses with the main themes of “IT,” “music,” “design,” “mountains/nature,” “aviation/railway,” “mobile services,” and “academic/science and engineering.” Furthermore, we also develop and operate content business platforms. [Contact information] Kindai Kagakusha Co., Ltd. TEL: 03-6837-4828 Email:
kdd-qa@kindaikagaku.co.jp