Abstract
In numerical simulations, a high-fidelity (HF) simulation is generally more accurate than a low-fidelity (LF) simulation, while the latter is generally more computationally efficient than the former. To take advantages of both HF and LF simulations, a multi-fidelity surrogate (MFS) model based on moving least squares (MLS), termed as adaptive MFS-MLS, is proposed. The MFS-MLS calculates the LF scaling factors and the unknown coefficients of the discrepancy function simultaneously using an extended MLS model. In the proposed method, HF samples are not regarded as equally important in the process of constructing MFS-MLS models, and adaptive weightings are given to different HF samples. Moreover, both the size of the influence domain and the scaling factors can be determined adaptively according to the training samples. The MFS-MLS model is compared with three state-of-the-art MFS models and three single-fidelity surrogate models in terms of the prediction accuracy through multiple benchmark numerical cases and an engineering problem. In addition, the effects of key factors on the performance of the MFS-MLS model, such as the correlation between HF and LF models, the cost ratio of HF to LF samples, and the combination of HF and LF samples, are also investigated. The results show that MFS-MLS is able to provide competitive performance with high computational efficiency.
Similar content being viewed by others
References
Absi GN, Mahadevan S (2019) Simulation and sensor optimization for multifidelity dynamics model calibration. AIAA J 58(2):879–888
Alexandrov NM, Lewis RM, Gumbert CR, Green LL, Newman PA (2001) Approximation and model management in aerodynamic optimization with variable-fidelity models. J Aircr 38(6):1093–1101
Bai J, Meng G, Zuo W (2019) Rollover crashworthiness analysis and optimization of bus frame for conceptual design. J Mech Sci Technol 33(7):3363–3373
Basheer IA, Hajmeer M (2000) Artificial neural networks: fundamentals, computing, design, and application. J Microbiol Methods 43(1):3–31
Batra R, Pilania G, Uberuaga BP, Ramprasad R (2019) Multifidelity information fusion with machine learning: a case study of dopant formation energies in Hafnia. ACS Appl Mater Interfaces 11(28):24906–24918
Berci M, Toropov VV, Hewson RW, Gaskell PH (2014) Multidisciplinary multifidelity optimisation of a flexible wing aerofoil with reference to a small UAV. Struct Multidisc Optim 50(4):683–699
Breitkopf P, Naceur H, Rassineux A, Villon P (2005) Moving least squares response surface approximation: formulation and metal forming applications. Comput Struct 83(17–18):1411–1428
Cai X, Qiu H, Gao L, Shao X (2017) Metamodeling for high dimensional design problems by multi-fidelity simulations. Struct Multidisc Optim 56(1):151–166
Clarke SM, Griebsch JH, Simpson TW (2005) Analysis of support vector regression for approximation of complex engineering analyses. J Mech Des 127(6):1077–1087
Durantin C, Rouxel J, Désidéri JA, Glière A (2017) Multifidelity surrogate modeling based on radial basis functions. Struct Multidisc Optim 56(5):1061–1075
Eldred M, Giunta A, Collis S (2004) Second-order corrections for surrogate-based optimization with model hierarchies. In 10th AIAA/ISSMO multidisciplinary analysis and optimization conference.
Eldred MS, Giunta AA, Collis SS (2004) Second-order corrections for surrogate-based optimization with model hierarchies. Collect Tech Pap - 10th AIAA/ISSMO Multidiscip Anal Optim Conf 3:1754–1768.
Fang H, Horstemeyer MF (2006) Global response approximation with radial basis functions. Eng Optim 38(04):407–424
Fernández-Godino MG, Park C, Kim NH, Haftka RT (2016) Review of multi-fidelity models. http://arxiv.org/abs/1609.07196
Fernández-Godino MG, Park C, Kim NH, Haftka RT (2019) Issues in deciding whether to use multifidelity surrogates. AIAA J 57(5):2039–2054
Forrester AI, Keane AJ (2009) Recent advances in surrogate-based optimization. Prog Aerosp Sci 45(1–3):50–79
Forrester AI, Sóbester A, Keane AJ (2007) Multi-fidelity optimization via surrogate modelling. Proc R Soc A 463(2088):3251–3269
Forrester A, Sobester A, Keane A (2008) Engineering design via surrogate modelling: a practical guide. Wiley, Hoboken
Goel T, Haftka RT, Shyy W, Queipo NV (2007) Ensemble of surrogates. Struct Multidisc Optim 33(3):199–216
Haftka RT (1991) Combining global and local approximations. AIAA J 29(9):1523–1525
Han ZH, Görtz S (2012) Hierarchical kriging model for variable-fidelity surrogate modeling. AIAA J 50(9):1885–1896
Han ZH, Görtz S, Zimmermann R (2013) Improving variable-fidelity surrogate modeling via gradient-enhanced kriging and a generalized hybrid bridge function. Aerosp Sci Technol 25(1):177–189
Hao P, Feng S, Zhang K, Li Z, Wang B, Li G (2018) Adaptive gradient-enhanced kriging model for variable-stiffness composite panels using Isogeometric analysis. Struct Multidisc Optim 58(1):1–16
Hao P, Feng S, Li Y, Wang B, Chen H (2020) Adaptive infill sampling criterion for multi-fidelity gradient-enhanced kriging model. Struct Multidisc Optim 62:1–21
Hutchison MG, Unger ER, Mason WH, Grossman B, Haftka RT (1994) Variable-complexity aerodynamic optimization of a high-speed civil transport wing. J Aircr 31(1):110–116
Jin R, Chen W, Simpson TW (2001) Comparative studies of metamodeling techniques under multiple modelling criteria. Struct Multidisc Optim 23(1):1–13
Jones DR, Schonlau M, Welch WJ (1998) Efficient global optimization of expensive black-box functions. J Global Optim 13(4):455–492
Kaplan AG, Kaplan YA (2020) Developing of the new models in solar radiation estimation with curve fitting based on moving least-squares approximation. Renewable Energy 146:2462–2471
Kleijnen JP (2008) Response surface methodology for constrained simulation optimization: an overview. Simul Model Pract Theory 16(1):50–64
Krishnamurthy T (2005) Comparison of response surface construction methods for derivative estimation using moving least squares, kriging and radial basis functions. In 46th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials conference
Lancaster P, Salkauskas K (1981) Surfaces generated by moving least squares methods. Math Comput 37(155):141–158
Lee J, Yong SC (2011) Role of conservative moving least squares methods in reliability based design optimization: a mathematical foundation. J Mech Design 133(12):121005
Li J, Wang H, Kim NH (2012) Doubly weighted moving least squares and its application to structural reliability analysis. Struct Multidisc Optim 46(1):69–82
Liu Y, Collette M (2014) Improving surrogate-assisted variable fidelity multi-objective optimization using a clustering algorithm. Appl Soft Comput 24:482–493
Lü Q, Xiao ZP, Ji J, Zheng J, Shang YQ (2017) Moving least squares method for reliability assessment of rock tunnel excavation considering ground-support interaction. Comput Geotech 84:88–100
Mainini L, Maggiore P (2012) A Multifidelity Approach to Aerodynamic Analysis in an Integrated Design Environment. In53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference 20th AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA
Majdisova Z, Skala V (2017) Radial basis function approximations: comparison and applications. Appl Math Model 51:728–743
Myers RH, Montgomery DC, Anderson-Cook CM (2016) Response surface methodology: process and product optimization using designed experiments. Wiley, Hoboken
Napier N, Sriraman SA, Tran HT, James KA (2020) An artificial neural network approach for generating high-resolution designs from low-resolution input in topology optimization. J Mech Des 142:1
Park C, Haftka RT, Kim NH (2018) Low-fidelity scale factor improves Bayesian multi-fidelity prediction by reducing bumpiness of discrepancy function. Struct Multidisc Optim 58(2):399–414
Sacks J, Welch WJ, Mitchell TJ, Wynn HP (1989) Design and analysis of computer experiments. Stat Sci 4:409–423
Smola AJ, Schölkopf B (2004) A tutorial on support vector regression. Stat Comput 14(3):199–222
Song X, Lv L, Sun W, Zhang J (2019) A radial basis function-based multi-fidelity surrogate model: exploring correlation between high-fidelity and low-fidelity models. Struct Multidisc Optim 60(3):965–981
Sun G, Li G, Zhou S, Xu W, Yang X, Li Q (2011) Multi-fidelity optimization for sheet metal forming process. Struct Multidisc Optim 44(1):111–124
Tao J, Sun G (2019) Application of deep learning based multi-fidelity surrogate model to robust aerodynamic design optimization. Aerosp Sci Technol 92:722–737
Toal DJ (2015) Some considerations regarding the use of multi-fidelity Kriging in the construction of surrogate models. Struct Multidisc Optim 51(6):1223–1245
Viana FA, Simpson TW, Balabanov V, Toropov V (2014) Metamodeling in multidisciplinary design optimization: how far have we really come? AIAA J 52(4):670–690
Viana FA (2010) SURROGATES Toolbox User’s Guide, Version 2.1, http://sites.google.com/site/felipeacviana/surrogatestoolbox
Wang B (2015) A local meshless method based on moving least squares and local radial basis functions. Eng Anal Boundary Elem 50:395–401
Wang GG, Shan S (2007) Review of metamodeling techniques in support of engineering design optimization. J Mech Des 129(4):370–380
Zhang Y, Kim NH, Park C, Haftka RT (2018) Multifidelity surrogate based on single linear regression. AIAA J 56(12):4944–4952
Zhou Q, Shao X, Jiang P, Zhou H, Shu L (2015) An adaptive global variable fidelity metamodeling strategy using a support vector regression based scaling function. Simul Model Pract Theory 59:18–35
Zhou Q, Wang Y, Choi SK, Jiang P, Shao X, Hu J (2017) A sequential multi-fidelity metamodeling approach for data regression. Knowl-Based Syst 134:199–212
Funding
This research is financially supported by the National Key Research and Development Program of China (Grant No. 2018YFB1700704).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Replication of results
The main codes and raw data are submitted as supplementary materials.
Additional information
Responsible Editor: Vassili Toropov
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the supplementary material.
Appendices
Appendix 1: 16 Test Functions
No. | HF/LF | Test functions | D | S | r 2 |
---|---|---|---|---|---|
1 | HF | \({y}_{h}={(6x-2)}^{2}\text{sin}(12x-4)\) | 1 | (0,1]D | 0.58 |
LF | \({y}_{l}=0.56{y}_{h}+10\left(x-0.5\right)-5\) | ||||
2 | HF | \({y}_{h}=\text{sin}\left(2\pi \left(x-0.1\right)\right)+{x}^{2}\) | 1 | [0,1]D | 0.86 |
LF | \({y}_{l}=\text{sin}\left(2\pi \left(x-0.1\right)\right)\) | ||||
3 | HF | \({y}_{h}=x\text{sin}\left(x\right)/10\) | 1 | [0,10]D | 0.73 |
LF | \({y}_{l}=x\text{sin}\left(x\right)/10+x/10\) | ||||
4 | HF | \({y}_{h}=\text{cos}(3.5\pi x)\text{exp}(-1.4x)\) | 1 | [0,1]D | 0.75 |
LF | \({y}_{l}=\text{cos}\left(3.5\pi x\right)\text{exp}\left(-1.4x\right)+0.75{x}^{2}\) | ||||
5 | HF | \({y}_{h}={4{x}_{1}}^{2}-{2.1x}_{1}^{4}+{{\frac{1}{3}x}_{1}^{6}+x}_{1}{x}_{2}-4{x}_{2}^{2}+4{x}_{2}^{4}\) | 2 | [− 2,2]D | 0.77 |
LF | \({y}_{l}={2{x}_{1}}^{2}-{2.1x}_{1}^{4}+{{\frac{1}{3}x}_{1}^{6}+0.5x}_{1}{x}_{2}-4{x}_{2}^{2}+2{x}_{2}^{4}\) | ||||
6 | HF | \(y_{h} = \left[ {x_{2} - 1.275\left( {\frac{{x_{1} }}{\pi }} \right)^{2} + 5\frac{{x_{1} }}{\pi } - 6} \right]^{2} + 10\left( {1 - \frac{1}{8\pi }} \right){\text{cos}}\left( {x_{1} } \right)\) | 2 | \({x}_{1}\in [-\text{5,10}]\) \({x}_{2}\in [\text{0,15}]\) | 0.98 |
LF | \({y}_{l}=\frac{1}{2}{[{x}_{2}-1.275{\left(\frac{{x}_{1}}{\pi }\right)}^{2}+5\frac{{x}_{1}}{\pi }-6]}^{2}+10(1-\frac{1}{8\pi })\text{cos}({x}_{1})\) | ||||
7 | HF | \({y}_{h}={[1-2{x}_{1}+0.05\text{sin}(4\pi {x}_{2}-{x}_{1})]}^{2}+{[{x}_{2}-0.5\text{sin}(2\pi {x}_{1})]}^{2}\) | 2 | [0,1]D | 0.85 |
LF | \({y}_{l}={[1-2{x}_{1}+0.05\text{sin}(4\pi {x}_{2}-{x}_{1})]}^{2}+4{[{x}_{2}-0.5\text{sin}(2\pi {x}_{1})]}^{2}\) | ||||
8 | HF | \(y_{h} = \sum\nolimits_{i = 1}^{2} {x_{i}^{4} - 16x_{i}^{2} + 5x_{i} }\) | 2 | [− 3,4]D | 0.83 |
LF | \(y_{l} = \sum\nolimits_{i = 1}^{2} {x_{i}^{4} - 16x_{i}^{2} }\) | ||||
9 | HF | \({y}_{h}=\frac{1}{6}[\left(30+5{x}_{1}\text{sin}\left(5{x}_{1}\right)\right)\left(4+\text{exp}\left(-5{x}_{2}\right)\right)-100]\) | 2 | [0,1]D | 0.88 |
LF | \({y}_{l}=\frac{1}{6}[\left(30+5{x}_{1}\text{sin}\left(5{x}_{1}\right)\right)\left(4+\frac{2}{5}\text{exp}\left(-5{x}_{2}\right)\right)-100]\) | ||||
10 | HF | \({y}_{h}=\text{cos}({x}_{1}+{x}_{2})\text{exp}({x}_{1}{x}_{2})\) | 2 | [0,1]D | 0.86 |
LF | \({y}_{l}=\text{cos}[0.6\left({x}_{1}+{x}_{2}\right)]\text{exp}(0.6{x}_{1}{x}_{2})\) | ||||
11 | HF | \(y_{h} = \sum\nolimits_{i = 1}^{3} {0.3\sin \left( {\frac{16}{{15}}x_{i} - 1} \right) + \left[ {{\text{sin}}\left( {\frac{16}{{15}}x_{i} - 1} \right)} \right]^{2} }\) | 3 | [− 1,1]D | 0.40 |
LF | \(y_{l} = \sum\nolimits_{i = 1}^{3} {0.3\sin \left( {\frac{16}{{15}}x_{i} - 1} \right) + 0.2\left[ {{\text{sin}}\left( {\frac{16}{{15}}x_{i} - 1} \right)} \right]^{2} }\) | ||||
12 | HF | \({y}_{h}={({x}_{1}-1)}^{2}+{({x}_{1}-{x}_{2})}^{2}+{x}_{2}{x}_{3}+0.5\) | 3 | [0,1]D | 0.69 |
LF | \({y}_{l}=0.2{y}_{h}-0.5{x}_{1}-0.2{x}_{1}{x}_{2}-0.1\) | ||||
13 | HF | \(y_{h} = \sum\nolimits_{i = 1}^{5} {\left[ {100\left( {x_{i}^{2} - x_{i + 1} } \right)^{2} + \left( {x_{i} - 1} \right)^{2} } \right]}\) | 6 | [0,1]D | 0.54 |
LF | \(y_{l} = \sum\nolimits_{i = 1}^{5} {\left[ {100\left( {x_{i}^{2} - 4x_{i + 1} } \right)^{2} + \left( {x_{i} - 1} \right)^{2} } \right]}\) | ||||
14 | HF | \(y_{h} = \sum\nolimits_{i = 1}^{8} {x_{i}^{4} - 16x_{i}^{2} + 5x_{i} }\) | 8 | [− 3,3]D | 0.74 |
LF | \(y_{l} = \sum\nolimits_{i = 1}^{8} {0.3x_{i}^{4} - 16x_{i}^{2} + 5x_{i} }\) | ||||
15 | HF | \(y_{h} = \sum\nolimits_{{i = 1}}^{2} {\left[ {\left( {x_{{4i - 3}} + 10x_{{4i - 2}} } \right)^{2} + 5\left( {x_{{4i - 1}} - x_{{4i}} } \right)^{2} + \left( {x_{{4i - 2}} - 2x_{{4i - 1}} } \right)^{4} + 10\left( {x_{{4i - 3}} - x_{{4i}} } \right)^{4} } \right]}\) | 8 | [0,1]D | 0.66 |
LF | \(y_{l} = \sum\nolimits_{{i = 1}}^{2} {\left[ {\left( {x_{{4i - 3}} + 10x_{{4i - 2}} } \right)^{2} + 125\left( {x_{{4i - 1}} - x_{{4i}} } \right)^{2} + \left( {x_{{4i - 2}} - 2x_{{4i - 1}} } \right)^{4} + 10\left( {x_{{4i - 3}} - x_{{4i}} } \right)^{4} } \right]}\) | ||||
16 | HF | \(y_{h} = \sum\nolimits_{i = 1}^{10} {{\text{exp}}\left( {x_{i} } \right)\left[ {A\left( i \right) + x_{i} - {\text{ln}}\left( {\sum\nolimits_{k = 1}^{10} {{\text{exp}}\left( {x_{k} } \right)} } \right)} \right]}\) A = [− 6.089, − 17.164, − 34.054, − 5.914, − 24.721, − 14.986, − 24.100, − 10.708, − 26.662, − 22.662, − 22.179] | 10 | [− 2,3]D | 0.94 |
LF | \(y_{l} = \sum\nolimits_{i = 1}^{10} {{\text{exp}}\left( {x_{i} } \right)\left[ {B\left( i \right) + x_{i} - {\text{ln}}\left( {\sum\nolimits_{k = 1}^{10} {{\text{exp}}\left( {x_{k} } \right)} } \right)} \right]}\) B = [− 10, − 10, − 20, − 10, − 20, − 20, − 20, − 10, − 20, − 20] |
Appendix 2: Results of 16 test functions
Function | MFS-MLS | MFS-RBF | CoRBF | LR-MFS | RBF | MLS | PRS |
---|---|---|---|---|---|---|---|
1 | 0.996 ± 0.000 | 0.969 ± 0.052 | 0.987 ± 0.070 | 0.996 ± 0.000 | 0.756 ± 0.192 | 0.812 ± 0.157 | 0.250 ± 0.084 |
2 | 0.999 ± 0.010 | 0.999 ± 0.000 | 0.981 ± 0.012 | 0.975 ± 0.010 | 0.933 ± 0.048 | 0.956 ± 0.028 | 0.334 ± 0.093 |
3 | 0.994 ± 0.000 | 0.845 ± 0.372 | 0.999 ± 0.000 | 0.994 ± 0.000 | 0.572 ± 0.251 | 0.734 ± 0.230 | 0.064 ± 0.047 |
4 | 0.999 ± 0.000 | 0.982 ± 0.044 | 0.973 ± 0.024 | 0.966 ± 0.022 | 0.863 ± 0.168 | 0.784 ± 0.135 | 0.023 ± 0.015 |
5 | 0.834 ± 0.057 | 0.810 ± 0.045 | 0.677 ± 0.149 | 0.513 ± 0.079 | 0.515 ± 0.096 | 0.383 ± 0.206 | 0.666 ± 0.135 |
6 | 0.874 ± 0.047 | 0.883 ± 0.046 | 0.441 ± 0.406 | 0.876 ± 0.048 | 0.625 ± 0.152 | 0.523 ± 0.245 | 0.515 ± 0.165 |
7 | 0.960 ± 0.015 | 0.931 ± 0.038 | 0.942 ± 0.102 | 0.752 ± 0.065 | 0.815 ± 0.100 | 0.502 ± 0.253 | 0.574 ± 0.161 |
8 | 0.817 ± 0.051 | 0.811 ± 0.053 | 0.738 ± 0.332 | 0.817 ± 0.051 | 0.529 ± 0.109 | 0.259 ± 0.210 | 0.254 ± 0.120 |
9 | 0.963 ± 0.026 | 0.968 ± 0.012 | 0.974 ± 0.031 | 0.930 ± 0.017 | 0.899 ± 0.061 | 0.693 ± 0.270 | 0.916 ± 0.046 |
10 | 0.990 ± 0.010 | 0.970 ± 0.028 | 0.965 ± 0.055 | 0.892 ± 0.034 | 0.941 ± 0.058 | 0.935 ± 0.098 | 0.954 ± 0.148 |
11 | 0.928 ± 0.018 | 0.848 ± 0.063 | 0.771 ± 0.082 | 0.774 ± 0.063 | 0.911 ± 0.030 | 0.457 ± 0.070 | 0.759 ± 0.111 |
12 | 0.985 ± 0.005 | 0.962 ± 0.014 | 0.949 ± 0.070 | 0.858 ± 0.267 | 0.898 ± 0.045 | 0.984 ± 0.014 | 0.975 ± 0.046 |
13 | 0.720 ± 0.098 | 0.708 ± 0.050 | 0.427 ± 0.196 | 0.582 ± 0.064 | 0.505 ± 0.089 | 0.254 ± 0.197 | 0.483 ± 0.242 |
14 | 0.835 ± 0.019 | 0.820 ± 0.031 | 0.025 ± 0.018 | 0.824 ± 0.029 | 0.688 ± 0.069 | 0.452 ± 0.117 | 0.042 ± 0.000 |
15 | 0.975 ± 0.005 | 0.945 ± 0.011 | 0.920 ± 0.015 | 0.919 ± 0.013 | 0.932 ± 0.015 | 0.975 ± 0.006 | 0.958 ± 0.096 |
16 | 0.896 ± 0.014 | 0.893 ± 0.012 | 0.711 ± 0.013 | 0.892 ± 0.012 | 0.831 ± 0.021 | 0.832 ± 0.026 | 0.147 ± 0.071 |
Rights and permissions
About this article
Cite this article
Wang, S., Liu, Y., Zhou, Q. et al. A multi-fidelity surrogate model based on moving least squares: fusing different fidelity data for engineering design. Struct Multidisc Optim 64, 3637–3652 (2021). https://doi.org/10.1007/s00158-021-03044-5
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00158-021-03044-5