Abstract
Global sensitivity analysis (GSA) is a commonly used approach to explore the contribution of input variables to the model output and identify the most important variables. However, performing GSA typically requires a large number of model evaluations, which can result in a heavy computational burden, particularly when the model is computationally expensive. To address this issue, an efficient Sobol index estimator is proposed in this paper using polynomial chaos expansion (PCE) enhanced Gaussian process regression (GPR) method, namely PCEGPR. The orthogonal polynomial functions of PCE method are incorporated into GPR surrogate model to construct the kernel function. An estimation scheme based on fixed-point iteration and leave-one-out cross-validation error is presented to determine the optimal parameters of PCEGPR method. The analytical expressions of main and total sensitivity indices are also derived by considering the posterior predictor and covariance of PCEGPR surrogate model. The effectiveness of the proposed estimator is demonstrated by four numerical examples.
Similar content being viewed by others
Data availability
Data generated during this study are available from the corresponding author upon reasonable request.
References
Li S, Tang ZC (2018) An efficient numerical simulation method for evaluations of uncertainty analysis and sensitivity analysis of system with mixed uncertainties. Adv Mech Eng 10(10):1687814018800533
Cheng K, Lu ZZ, Ling CN, Zhou ST (2020) Surrogate-assisted global sensitivity analysis: an overview. Struct Multidiscip Optim 61(3):1187–1213
Sun X, Choi Y, Choi J (2020) Global sensitivity analysis for multivariate outputs using polynomial chaos-based surrogate models. Appl Math Model 82:867–887
Wang P, Li CY, Liu FC, Zhou HY (2021) Global sensitivity analysis of failure probability of structures with uncertainties of random variable and their distribution parameters. Eng Comput. https://doi.org/10.1007/s00366-021-01484-7
Wei PF, Lu ZZ, Song JW (2015) Variable importance analysis: a comprehensive review. Reliab Eng Syst Saf 142:399–432
Wang Lu, Zhang XB, Li GJ, Lu ZZ (2022) Credibility distribution function based global and regional sensitivity analysis under fuzzy uncertainty. Eng Comput 38:1349–1362
Saltelli A, Annoni P (2010) How to avoid a perfunctory sensitivity analysis. Environ Model Softw 25(12):1508–1517
Kucherenko S, Song SF, Wang L (2019) Different numerical estimators for main effect global sensitivity indices. Reliab Eng Syst Saf 165:222–238
Borgonovo E, Hazen GB, Plischke E (2016) A common rationale for global sensitivity measures and their estimation. Risk Anal 36(10):1871–1895
Janon A, Klein T, Lagnoux A, Nodet M, Prieur C (2014) Asymptotic normality and efficiency of two Sobol index estimators. ESAIM-Probab Stat 18:342–364
Sobol IM (2001) Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates. Math Comput Simul 55:271–280
Burnaev E, Panin I, Sudret B (2017) Efficient design of experiments for sensitivity analysis based on polynomial chaos expansions. Ann Math Artif Intell 81(1–2):187–207
Qian E, Peherstorfer B, O’Malley D, Vesselinov W, Willcox K (2018) Multifidelity Monte Carlo estimation of variance and sensitivity indices. SIAM/ASA J Uncertain Quantif 6(2):683–706
Damblin G, Ghione A (2021) Adaptive use of replicated Latin hypercube designs for computing Sobol’ sensitivity indices. Reliab Eng Syst Saf 212:107507
Janouchova E, Kucerova A (2013) Competitive comparison of optimal designs of experiments for sampling-based sensitivity analysis. Comput Struct 124:47–60
Borgonovo E, Plischke E (2016) Sensitivity analysis: a review of recent advances. Eur J Oper Res 246(3):869–887
Viana FAC, Gogu C, Goel T (2021) Surrogate modeling: tricks that endured the test of time and some recent developments. Struct Multidiscip Optim 64(5):2881–2908
Bhosekar A, Ierapetritou M (2018) Advances in surrogate based modeling, feasibility analysis, and optimization: a review. Comput Chem Eng 108:250–267
Zhu ZG, Ji HB, Li L (2023) Deep multi-modal subspace interactive mutual network for specific emitter identification. IEEE Trans Aerosp Electron Syst. https://doi.org/10.1109/TAES.2023.3240115
Yang HQ, Wang ZH, Song KL (2022) A new hybrid grey wolf optimizer-feature weighted-multiple kernel-support vector regression technique to predict TBM performance. Eng Comput 38:2469–2485
Schöbi R, Sudret B, Wiart J (2015) Polynomial-chaos-based Kriging. Int J Uncertain Quantif 5:171–193
Shang XB, Ma P, Chao T, Yang M (2020) A sequential experimental design for multivariate sensitivity analysis using polynomial chaos expansion. Eng Optim 52(8):1382–1400
Kaintura A, Spina D, Couckuyt I, Knockaert L, Bogaerts W, Dhaene T (2017) A kriging and stochastic collocation ensemble for uncertainty quantification in engineering applications. Eng Comput 33(4):935–949
Liu FC, He PF, Dai Y (2023) A new Bayesian probabilistic integration framework for hybrid uncertainty propagation. Appl Math Model 117:296–315
Jiang C, Hu Z, Liu YX, Mourelatos ZP, Gorsich D, Jayakumar P (2020) A sequential calibration and validation framework for model uncertainty quantification and reduction. Comput Methods Appl Mech Eng 368:113172
Yang MD, Zhang DQ, Han X (2022) Efficient local adaptive kriging approximation method with single-loop strategy for reliability-based design optimization. Eng Comput 38(3):2431–2449
Sudret B (2008) Global sensitivity analysis using polynomial chaos expansions. Reliab Eng Syst Saf 93(7):964–979
Crestaux T, Le Maitre O, Martinez JM (2009) Polynomial chaos expansion for sensitivity analysis. Reliab Eng Syst Saf 94(7):1161–1172
Palar PS, Tsuchiya T, Parks GT (2016) Multi-fidelity non-intrusive polynomial chaos based on regression. Comput Methods Appl Mech Eng 307:489–490
Palar PS, Zuhal LR, Shimoyama K, Tsuchiya T (2018) Global sensitivity analysis via multi-fidelity polynomial chaos expansion. Reliab Eng Syst Saf 170:175–190
Bhattacharyya B (2020) Global sensitivity analysis: A Bayesian learning based polynomial chaos approach. J Comput Phys 415:109539
Guo L, Narayan A, Zhou T (2018) A gradient enhanced ℓ1-minimization for sparse approximation of polynomial chaos expansions. J Comput Phys 367:49–64
Peng J, Hampton J, Doostan A (2016) On polynomial chaos expansion via gradient-enhanced ℓ1-minimization. J Comput Phys 310:440–458
Chen LM, Qiu HB, Jiang C, Xiao M, Gao L (2018) Support Vector enhanced Kriging for metamodeling with noisy data. Struct Multidiscip Optim 57(4):1611–1623
Marrel A, Iooss B, Laurent B, Roustant O (2009) Calculations of Sobol indices for the Gaussian process metamodel. Reliab Eng Syst Saf 94(3):742–751
De Lozzo M, Marrel A (2016) Estimation of the derivative-based global sensitivity measures using a Gaussian process metamodel. SIAM-ASA J Uncertain Quantif 4(1):708–738
Zhou YC, Lu ZZ, Cheng K, Yun WY (2019) A Bayesian Monte Carlo-based method for efficient computation of global sensitivity indices. Mech Syst Signal Process 117:498–516
Cheng K, Lu ZZ (2018) Adaptive sparse polynomial chaos expansions for global sensitivity analysis based on support vector regression. Comput Struct 194:86–96
Tang KK, Congedo PM, Abgrall R (2016) Adaptive surrogate modeling by ANOVA and sparse polynomial dimensional decomposition for global sensitivity analysis in fluid simulation. J Comput Phys 214:557–589
Lin Q, Hu DW, Hu JX, Cheng YS, Zhou Q (2021) A screening-based gradient-enhanced Gaussian process regression model for multi-fidelity data fusion. Adv Eng Inform 50:101437
Arthur CK, Temeng VA, Ziggah YY (2020) Novel approach to predicting blast-induced ground vibration using Gaussian process regression. Eng Comput 36(1):2942
Hadigol M, Doostan A (2018) Least squares polynomial chaos expansion: a review of sampling strategies. Comput Methods Appl Mech Eng 332:382–407
Cheng K, Lu ZZ, Zhou YC, Shi Y, Wei YH (2017) Global sensitivity analysis using support vector regression. Appl Math Model 49:587–598
Moghaddam VH, Hamidzadeh J (2016) New Hermite orthogonal polynomial kernel and combined kernels in support vector machine classifier. Pattern Recogn 60:921–935
Cheng K, Lu ZZ, Zhen Y (2019) Multi-level multi-fidelity sparse polynomial chaos expansion based on Gaussian process regression. Comput Methods Appl Mech Eng 349:360–377
Yan L, Duan XJ, Liu BW, Xu J (2018) Gaussian processes and polynomial chaos expansion for regression problem: linkage via the RKHS and comparison via the KL divergence. Entropy 20(3):191
Cheng K, Lu ZZ, Xiao SN, Oladyshkin S, Nowak W (2022) Mixed covariance function Kriging model for uncertainty quantification. Int J Uncertain Quantif 12(3):17–30
Liu HT, Cai JF, Ong YS (2017) An adaptive sampling approach for kriging metamodeling by maximizing expected prediction error. Comput Chem Eng 106:171–182
Tipping ME (2001) Sparse Bayesian learning and the relevance vector machine. J Mach Learn Res 1(3):211–244
Shang XB, Chao T, Ma P, Yang M (2020) An efficient local search-based genetic algorithm for constructing optimal Latin hypercube design. Eng Optim 52(2):271–287
Shang XB, Su L, Fang H, Zeng BW, Zhang Z (2023) An efficient multi-fidelity Kriging surrogate model-based method for global sensitivity analysis. Reliab Eng Syst Saf 229:108858
Kucherenko S, Feil B, Shah N, Mauntz W (2011) The identification of model effective dimensions using global sensitivity analysis. Reliab Eng Syst Saf 96(4):440–449
Wu ZP, Wang WJ, Wang DH, Zhao K, Zhang WH (2019) Global sensitivity analysis using orthogonal augmented radial basis function. Reliab Eng Syst Saf 185:291–302
Wipf DP, Rao BD (2004) Sparse Bayesian learning for basis selection. IEEE Trans Signal Process 52(8):2153–2164
Acknowledgements
This work is supported by the Natural Science Foundation of Heilongjiang Province of China [grant number: LH2023F022], the National Natural Science Foundation of China [grant numbers: 62173103, 52171299], and the Fundamental Research Funds for the Central Universities of China [grant numbers: 3072022JC0402, 3072022JC0403].
Funding
Natural Science Foundation of Heilongjiang Province of China, LH2023F022, National Natural Science Foundation of China, 62173103, 52171299, Fundamental Research Funds for the Central Universities of China, 3072022JC0402, 3072022JC0403.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that there are no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In the general case, the random noise \(\upsilon\) with zero mean and \(\sigma^{2}\) variance, i.e., \(\upsilon \sim \boldsymbol{\mathcal{N}}\left( {0,\sigma^{2} } \right)\) is considered in the training data. Then, the covariance matrix in Eq. (20) can be written as \(\Lambda = \Phi^{T} W\Phi + \sigma^{2}I\), where \(I\) denotes the identical matrix. The maximum likelihood method is used to compute \(\mu\) and \(W\). The likelihood function is written as
Taking the logarithm operator of likelihood function, the log-likelihood function is given as
Taking the derivative of \(\boldsymbol{\mathcal{L}}\left( {\mu ,W} \right)\) with respect to \(\mu\) and setting it to zero, we have
Then, the maximum likelihood estimation of \(\mu\) is given as
The second term of \(\boldsymbol{\mathcal{L}}\left( {\mu ,W} \right)\) in Eq. (67) can be written as
where \(C = \sigma^{ - 2} \Phi \Phi^{T} + W^{ - 1}\). According to Woodbury inversion [54], \(\Lambda^{ - 1}\) is reformulated as
and the vector of PCE coefficients PCE in Eq. (25) is also deduced as
Based on Eqs. (71) and (72), the third term of \(\boldsymbol{\mathcal{L}}\left( {\mu ,W} \right)\) is derived as [49, 54]
Taking the derivative of \(\boldsymbol{\mathcal{L}}\left( {\mu ,W} \right)\) with respect to \(w_{{{\varvec{\alpha}}_{i} }}\) and forcing it to zero, the following equation is obtained
where \(C_{{{\varvec{\alpha}}_{i} {\varvec{\alpha}}_{i} }}^{ - 1}\) is the ith diagonal element of \(C^{ - 1}\). Then the estimation of \(w_{{{\varvec{\alpha}}_{i} }}\) is given as
In Sect. 4.1, the random noise is not taken into consideration in the GPR modeling, i.e., \(\sigma^{2} = 0\). Therefore, Eq. (69) is transformed as
and \(C^{ - 1} = \sigma^{2} \left( {\Phi \Phi^{T} + \sigma^{2} W^{ - 1} } \right)^{ - 1}\) is equal to 0. Then, \(\hat{w}_{{{\varvec{\alpha}}_{i} }} = \lambda_{{{\varvec{\alpha}}_{i} }}\) and
To compute the values of \(\hat{\mu }\) and \(\hat{\user2{w}}\), the fixed-point iteration is used to solve Eqs. (76) and (77) in Sect. 4.1.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Shang, X., Zhang, Z., Fang, H. et al. Global sensitivity analysis using polynomial chaos expansion enhanced Gaussian process regression method. Engineering with Computers 40, 1231–1246 (2024). https://doi.org/10.1007/s00366-023-01851-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00366-023-01851-6