Polynomial Dimensional Decomposition Research Articles

Global sensitivity analysis with limited data via sparsity-promoting D-MORPH regression: Application to char combustion

In uncertainty quantification, variance-based global sensitivity analysis quantitatively determines the effect of each input random variable on the output by partitioning the total output variance into contributions from each input. However, computing conditional expectations can be prohibitively costly when working with expensive-to-evaluate models. Surrogate models can accelerate this, yet their accuracy depends on the quality and quantity of training data, which is expensive to generate (experimentally or computationally) for complex engineering systems. Thus, methods that work with limited data are desirable. We propose a diffeomorphic modulation under observable response preserving homotopy (D-MORPH) regression to train a polynomial dimensional decomposition surrogate of the output that minimizes the number of training data. The new method first computes a sparse Lasso solution and uses it to define the cost function. A subsequent D-MORPH regression minimizes the difference between the D-MORPH and Lasso solution. The resulting D-MORPH based surrogate is more robust to input variations and more accurate with limited training data. We illustrate the accuracy and computational efficiency of the new surrogate for global sensitivity analysis using mathematical functions and an expensive-to-simulate model of char combustion. The new method is highly efficient, requiring only 15% of the training data compared to conventional regression.

An adaptive data-driven subspace polynomial dimensional decomposition for high-dimensional uncertainty quantification based on maximum entropy method and sparse Bayesian learning

Polynomial dimensional decomposition (PDD) is a surrogate method originated from the ANOVA (analysis of variance) decomposition, and has shown powerful performance in uncertainty quantification (UQ) accuracy and convergence recently. However, complex high-dimensional problems result in a large number of polynomial basis functions, leading to heavy computational burden, and the probability distributions of the input random variables are indispensable for PDD modeling and UQ, which may be unavailable in practical engineering. This study establishes an adaptive data-driven subspace PDD (ADDSPDD) for high-dimensional UQ, which employs two types of data for modeling the PDD basis function and the low-dimensional subspace directly, namely, the data of input random variables and the input-response samples. Firstly, we propose a data-driven zero-entropy criterion-based maximum entropy method for reconstructing the probability density functions (PDF) of input variables. Then, with the aid of the established PDFs, a data-driven subspace PDD (DDSPDD) is proposed based on the whitening transformation. To recover the subspace of the function of interest accurately and efficiently, we put forward an approximate active subspace method (AAS) based on the Taylor expansion under some mild premises. Finally, we integrate an adaptive learning algorithm into the DDSPDD framework based on the sparse Bayesian learning theory, obtaining our ADDSPDD; thus, the real subspace and the significant PDD basis functions can be identified with limited computational budget. We validate the proposed method by using four examples, and systematically compare four existing dimension-reduction methods with the AAS. Results show that the proposed framework is effective and the AAS is a good choice when the corresponding assumptions are satisfied.

Random vibration analysis of an uncertain vehicle-track coupled system based on a polynomial dimensional decomposition

ABSTRACT In this paper, the uncertainty propagation of an uncertain vehicle-track coupled system (VTCS) subjected to track irregularity is quantified using a polynomial-dimensional decomposition (PDD). Firstly, the conditional power spectral density (PSD) of response related to uncertain parameters is derived using the pseudo-excitation method. Secondly, the PDD surrogate model that can describe the probabilistic characters of the uncertainty propagation is established by conducting the dimensional decomposition to the conditional PSD with component functions and performing the Fourier expansion to the component functions. Finally, the dimensional reduction integration and Gauss integration are introduced to overcome the difficulty of high-dimensional integration when calculating the expansion coefficients. In numerical example, the proposed method is applied to the uncertainty quantification of vertical vibration response of an uncertain VTCS, the accuracy and efficiency of the proposed method are verified by comparing the results of the PDD method with that of Monte-Carlo simulation.

Dynamic Responses Analysis of Bridges With Uncertain Parameters Under Moving Loads

This paper focuses on the dynamic responses analysis of bridges with uncertain parameters under moving loads and proposes the polynomial dimensional decomposition method for the analysis of structural responses induced by moving loads. The uncertain parameters were regarded as independent random variables and the random function of responses above these variables was constructed. The dimensional decomposition of the function was further performed using a group of component functions with increasing number of variables and the approximate expressions of the component functions were derived using Fourier polynomial expansion. The expansion coefficients were efficiently calculated by introducing dimension-reduction integration to reduce the integration dimensions. In the numerical examples, the analysis of responses of bridges with uncertain parameters under moving loads was performed and the accuracy and efficiency of the proposed method was verified by comparing the results of the proposed method with that of Monte Carlo simulation.

Nonstationary random vibration analysis of structures with uncertain parameters based on a polynomial dimensional decomposition

AbstractBased on polynomial dimensional decomposition (PDD), a novel method for analysing nonstationary random vibration of structures with uncertain parameters is proposed in this paper. The uncertain structural parameters are assumed to be independent random variables, with certain types of probability distributions; thus, the nonstationary random vibration responses of the structures become stochastic functions of these uncertain parameters. A dimensional decomposition and Fourier expansion were applied to calculate the statistical characteristics of the nonstationary random vibration responses of the structures. To calculate the conditional power spectral density (CPSD) and conditional standard deviation (CSD) of the nonstationary random vibration responses, CPSD and CSD were explicitly expressed using expansion coefficients and polynomial basis. Subsequently, the probability distributions of the structural responses can be effectively calculated in terms of these explicit expressions. Dimension‐reduction integration and Gauss numerical integration were applied to calculate expansion coefficients. The validity of the proposed method was verified by numerical examples, including a single‐degree‐of‐freedom system and a multi‐degree‐of‐freedom system (reinforced concrete shear wall). The results indicate that the proposed method is as accurate as Monte Carlo simulation (MCS), whereas the proposed method requires considerably smaller number of nonstationary random vibration analyses than MCS. Moreover, the nonstationary random vibration analysis of a long‐span cable‐stayed bridge demonstrates that the proposed method is feasible for quantifying the uncertainties of the nonstationary random vibration responses of complex engineering structures with uncertain parameters.

An adaptive polynomial dimensional decomposition method and its application in reliability analysis

PurposeThe polynomial dimensional decomposition (PDD) method is a popular tool to establish a surrogate model in several scientific areas and engineering disciplines. The selection of appropriate truncated polynomials is the main topic in the PDD. In this paper, an easy-to-implement adaptive PDD method with a better balance between precision and efficiency is proposed.Design/methodology/approachFirst, the original random variables are transformed into corresponding independent reference variables according to the statistical information of variables. Second, the performance function is decomposed as a summation of component functions that can be approximated through a series of orthogonal polynomials. Third, the truncated maximum order of the orthogonal polynomial functions is determined through the nonlinear judgment method. The corresponding expansion coefficients are calculated through the point estimation method. Subsequently, the performance function is reconstructed through appropriate orthogonal polynomials and known expansion coefficients.FindingsSeveral examples are investigated to illustrate the accuracy and efficiency of the proposed method compared with the other methods in reliability analysis.Originality/valueThe number of unknown coefficients is significantly reduced, and the computational burden for reliability analysis is eased accordingly. The coefficient evaluation for the multivariate component function is decoupled with the order judgment of the variable. The proposed method achieves a good trade-off of efficiency and accuracy for reliability analysis.

An adaptive sparse polynomial dimensional decomposition based on Bayesian compressive sensing and cross-entropy

An adaptive sparse polynomial dimensional decomposition based on Bayesian compressive sensing and cross-entropy

A novel polynomial dimension decomposition method based on sparse Bayesian learning and Bayesian model averaging

The polynomial dimensional decomposition (PDD), an orthogonal polynomial-based metamodel, has received increasing attention in uncertainty quantification (UQ). Nevertheless, for complex high-dimensional problems, its computational burden may become unaffordable, which is usually called curse of dimensionality. To solve this problem, sparse regression methods can be considered to establish a sparse PDD model. However, when the design samples are limited, their computational accuracy may be low due to the enormous size of the polynomial bases. Aimed at this issue, we proposed a novel sparse PDD metamodel based on the Bayesian LASSO (least absolute shrinkage and selection operator) method and an adaptive candidate basis selection and model updating method (CBSMU). Firstly, to improve the CPU efficiency, an analytical Bayesian LASSO based on the sparse Bayesian learning is used for the regression analysis, which replaces the time-consuming Markov chain Monte Carlo sampling of the traditional method with the efficient iteration algorithm for calculating the posterior estimations. Then, to reduce the size of the polynomial bases, this study proposes the adaptive CBSMU for screening the significant candidate polynomial bases and updating the metamodel sequentially. The CBSMU can find out the candidate bases that contributes to improve the prediction accuracy in the view of Bayesian model averaging. Thus, during the process of the sparse PDD modeling, the size of candidate bases is relatively small, which facilitates to improve the final computational accuracy when the design samples are limited. We verify the proposed method using three high-dimensional numerical examples, and apply it to solve one complex high-dimensional engineering problem. The results show that the proposed method is more accurate for UQ than the two common methods with the same computational costs, and is well-suited for solving complex high-dimensional structural dynamic problem.

Iterative Polynomial Dimensional Decomposition approach towards solution of structural mechanics problems with material randomness

One of the major difficulties in solving stochastic mechanics problems is the curse of dimensionality, where an exponential increase in the dimension of the problem is encountered with the increase in the number of random variables and/or order of expansion considered in any approximation. A prominent method in addressing the curse of dimensionality is ANOVA dimension Decomposition (ADD), which represents a mathematical function with multiple lower variate functions. These lower variate functions are represented using orthogonal polynomials, which yields Polynomial Dimensional Decomposition (PDD). In recent articles, the authors proposed an Iterative Polynomial Chaos (ImPC) based method for the solution of structural mechanics problems, where computational efficacy of ImPC was demonstrated against Polynomial Chaos (PC). In ImPC, the problems are solved iteratively using smaller sizes of PC expansions. Thus, it reduces the curse of dimensionality of PC expansion. The PDD reduces the size of the system matrix by considering a fewer number of random variables at a time, while ImPC can be considered to solve each components of PDD iteratively so that a converged solution can be achieved without increasing the order of expansion, which is termed as iterative PDD in the present study. Thus, the overall convergence can be achieved with a lesser size of the system matrix, which enables to perform analyses with a lesser computational facility. Further, the stiffness matrix size can be reduced by considering the random field at Gauss points instead of the mid point. Numerical studies with both Gaussian and non-Gaussian random field of Young’s modulus are conducted, and computational efficiency of the iterative PDD is compared with that of PDD, ImPC, and first order perturbation method. The iterative PDD is observed to be computationally less demanding and exhibits reduced dimensional curse.

A polynomial dimensional decomposition framework based on topology derivatives for stochastic topology sensitivity analysis of high-dimensional complex systems and a type of benchmark problems

In this paper, a new computational framework based on the topology derivative concept is presented for evaluating stochastic topological sensitivities of complex systems. The proposed framework, designed for dealing with high dimensional random inputs, dovetails a polynomial dimensional decomposition (PDD) of multivariate stochastic response functions and deterministic topology derivatives. On one hand, it provides analytical expressions to calculate topology sensitivities of the first three stochastic moments which are often required in robust topology optimization (RTO). On another hand, it offers embedded Monte Carlo Simulation (MCS) and finite difference formulations to estimate topology sensitivities of failure probability for reliability-based topology optimization (RBTO). For both cases, the quantification of uncertainties and their topology sensitivities are determined concurrently from a single stochastic analysis. Moreover, an original example of two random variables is developed for the first time to obtain analytical solutions for topology sensitivity of moments and failure probability. Another 53-dimension example is constructed for analytical solutions of topology sensitivity of moments and semi-analytical solutions of topology sensitivity of failure probabilities in order to verify the accuracy and efficiency of the proposed method for high-dimensional scenarios. Those examples are new and make it possible for researchers to benchmark stochastic topology sensitivities of existing or new algorithms. In addition, it is unveiled that under certain conditions the proposed method achieves better accuracies for stochastic topology sensitivities than for the stochastic quantities themselves.

Stochastic isogeometric analysis in linear elasticity

A new stochastic method, integrating spline dimensional decomposition (SDD) and isogeometric analysis (IGA), is proposed for solving stochastic boundary-value problems from linear elasticity. The method, referred to as SDD–IGA, involves Galerkin isogeometric analysis as a deterministic solver for governing partial differential equations and a novel Fourier-like orthogonal spline expansion generated from the analysis-of-variance decomposition of a high-dimensional function. For the stochastic part of the SDD–IGA method, an innovative dimension-reduction integration technique is presented for efficiently calculating the expansion coefficients. Analytical formulae have been derived to calculate the second-moment properties of an SDD–IGA approximation for a general output random variable of interest. Numerical examples demonstrate the capability of a low-order SDD–IGA in efficiently delivering probabilistic solutions with an approximation quality as good as, if not better than, that obtained from a high-order polynomial dimensional decomposition. The proposed SDD–IGA method is most suitable in the presence of locally nonlinear or nonsmooth behavior commonly found in applications.

The Applications of Order Reduction Methods in Nonlinear Dynamic Systems

Two different order reduction methods of the deterministic and stochastic systems are discussed in this paper. First, the transient proper orthogonal decomposition (T-POD) method is introduced based on the high-dimensional nonlinear dynamic system. The optimal order reduction conditions of the T-POD method are provided by analyzing the rotor-bearing system with pedestal looseness fault at both ends. The efficiency of the T-POD method is verified via comparing with the results of the original system. Second, the polynomial dimensional decomposition (PDD) method is applied to the 2 DOFs spring system considering the uncertain stiffness to study the amplitude-frequency response. The numerical results obtained by the PDD method agree well with the Monte Carlo simulation (MCS) method. The results of the PDD method can approximate to MCS better with the increasing of the polynomial order. Meanwhile, the Uniform-Legendre polynomials can eliminate perturbation of the PDD method to a certain extent via comparing it with the Gaussian-Hermite polynomials.

Forward and backward uncertainty quantification with active subspaces: Application to hypersonic flows around a cylinder

We perform a Bayesian calibration of the freestream velocity and density starting from measurements of the pressure and heat flux at the stagnation point of a hypersonic high-enthalpy flow around a cylinder. The objective is to explore the possibility of using stagnation heat flux measurements, together with pressure measurements, to rebuild freestream conditions since such measurements are available for recent space missions but not exploited for freestream characterization. First, we formulate an algorithm of mesh adaptation, enabling accurate numerical solutions in an automatic way for a given set of inputs. Secondly, active subspaces are used to find a low-dimensional dependence structures in the input-to-output map of the forward numerical solver. Then, surrogate models on the active variables are used to accelerate the forward uncertainty propagation by Monte Carlo sampling and the Markov Chain Monte Carlo sampling of the posterior distribution for Bayesian inversion. A preliminary sensitivity analysis with sparse Polynomial Dimensional Decomposition is performed on the chemical model of the air mixture, to determine the most influential uncertain chemical parameters in the forward problem. Then, the forward and backward methodologies are applied to the simulation of a hypersonic flow around a cylinder, in conditions for which experimental data are available, revealing new insights towards the potential exploitation of heat flux data for freestream rebuilding.

Dimensionwise multivariate orthogonal polynomials in general probability spaces

This paper puts forward a new generalized polynomial dimensional decomposition (PDD), referred to as GPDD, encompassing hierarchically organized, measure-consistent multivariate orthogonal polynomials in dependent random variables. In contrast to the existing PDD, which is valid strictly for independent random variables, no tensor-product structure or product-type probability measure is imposed or necessary. Fundamental mathematical properties of GPDD are examined by creating dimensionwise decomposition of polynomial spaces, deriving statistical properties of random orthogonal polynomials, demonstrating completeness of orthogonal polynomials for requisite assumptions, and upholding mean-square convergence to the correct limit. The GPDD approximation proposed should be effective in solving high-dimensional stochastic problems subject to dependent variables with a large class of non-product-type probability measures.

Adaptive sparse polynomial dimensional decomposition for derivative-based sensitivity

For applications, such as the plasma-coupled combustion system we consider, derivative-based sensitivity indices (DSI) are known to have several advantages over Sobol's total sensitivity indices, especially for small sample sizes. Several properties of derivative-based sensitivity measures are leveraged to develop a new and efficient numerical approach to estimate them. It is based on computing the DSI measures by effectively cost-free Monte Carlo sampling of an adaptively constructed orthogonal polynomial surrogate with uncertain input parameters that can have arbitrary probability distributions. The adaptivity reduces the number of necessary model evaluations, which is demonstrated both in a constructed example (the Moon function) and in two plasma-combustion systems with up to 55 uncertain parameters. Unimportant parameters are successfully identified and neglected with a low number of model evaluations, which makes it an attractive non-intrusive approach when adjoint solutions are unavailable to provide sensitivity information.

Stochastic design optimization accounting for structural and distributional design variables

PurposeThis paper aims to present a new method, named as augmented polynomial dimensional decomposition (PDD) method, for robust design optimization (RDO) and reliability-based design optimization (RBDO) subject to mixed design variables comprising both distributional and structural design variables.Design/methodology/approachThe method involves a new augmented PDD of a high-dimensional stochastic response for statistical moments and reliability analyses; an integration of the augmented PDD, score functions, and finite-difference approximation for calculating the sensitivities of the first two moments and the failure probability with respect to distributional and structural design variables; and standard gradient-based optimization algorithms.FindingsNew closed-form formulae are presented for the design sensitivities of moments that are simultaneously determined along with the moments. A finite-difference approximation integrated with the embedded Monte Carlo simulation of the augmented PDD is put forward for design sensitivities of the failure probability.Originality/valueIn conjunction with the multi-point, single-step design process, the new method provides an efficient means to solve a general stochastic design problem entailing mixed design variables with a large design space. Numerical results, including a three-hole bracket design, indicate that the proposed methods provide accurate and computationally efficient sensitivity estimates and optimal solutions for RDO and RBDO problems.

Kriging-sparse Polynomial Dimensional Decomposition surrogate model with adaptive refinement

Uncertainty Quantification and Sensitivity Analysis problems are made more difficult in the case of applications involving expensive computer simulations. This is because a limited amount of simulations is available to build a sufficiently accurate metamodel of the quantities of interest.In this work, an algorithm for the construction of a low-cost and accurate metamodel is proposed, having in mind computationally expensive applications. It has two main features. First, Universal Kriging is coupled with sparse Polynomial Dimensional Decomposition (PDD) to build a metamodel with improved accuracy. The polynomials selected by the adaptive PDD representation are used as a sparse basis to build a Universal Kriging surrogate model. Secondly, a numerical method, derived from anisotropic mesh adaptation, is formulated in order to adaptively insert a fixed number of new training points to an existing Design of Experiments.The convergence of the proposed algorithm is analyzed and assessed on different test functions with an increasing size of the input space. Finally, the algorithm is used to propagate uncertainties in two high-dimensional real problems related to the atmospheric reentry.

Uncertainty quantification under dependent random variables by a generalized polynomial dimensional decomposition

This paper is concerned with uncertainty quantification analysis of complex systems subject to dependent input random variables. The analysis focuses on a new, generalized version of polynomial dimensional decomposition (PDD), referred to as GPDD, entailing hierarchically ordered measure-consistent multivariate orthogonal polynomials in dependent variables. Under a few prescribed assumptions, GPDD exists for any square-integrable output random variable and converges in mean-square to the correct limit. New analytical formulae are proposed to calculate the mean and variance of a GPDD approximation of a general output variable in terms of the expansion coefficients and second-moment properties of multivariate orthogonal polynomials. However, unlike in PDD, calculating the coefficients of GPDD requires solving a coupled system of linear equations. Besides, the variance formula of GPDD contains extra terms due to statistical dependence among input variables. The extra terms disappear when the input variables are statistically independent, reverting GPDD to PDD. Two numerical examples, the one derived from a stochastic boundary-value problem and the other entailing a random eigenvalue problem, illustrate second-moment error analysis and estimation of the probabilistic characteristics of eigensolutions.

Statistical moment analysis of nonlinear rotor system with multi uncertain variables

Abstract The statistical moments of the dynamical system models with uncertainties are analyzed in this paper. The analytical and numerical cases of the polynomial dimensional decomposition to solve the dynamical responses are provided. First, the polynomial dimensional decomposition method is applied to study a two-degree-of-freedom spring system model with stiffness uncertainty. Second, a linear rotor system model with eight random variables is discussed in the cases of different polynomial orders. Third, a rotor system model supported by cubically nonlinear stiffness is established by the Newton’s second law. The amplitude-frequency responses of nine and twelve uncertain variables are calculated by combining the harmonic balance method and the polynomial dimensional decomposition method. The accuracy of the polynomial dimensional decomposition method is verified via comparing with the Monte Carlo Simulation method. The applications of the polynomial dimensional decomposition method in the nonlinear rotor systems can provide theoretical guidance to study complex rotor-bearing systems in the future.

Statistical moment analysis of multi-degree of freedom dynamic system based on polynomial dimensional decomposition method

The first two moments of the steady-state response of the spring and rotor models are determined by the polynomial dimensional decomposition (PDD) method and the Monte Carlo simulation (MCS) method in this paper. Both the analytical and numerical cases of the PDD method in the dynamical models are provided, and the response of the numerical case is calculated via the harmonic balance and PDD methods. The PDD can describe the amplitude–frequency characteristics of the dynamical models with random variables except the frequencies around the resonant frequencies. These results are shown as a three-DOF spring model with stiffness uncertainty, rotor models with four and nine random variables and nonlinear rotor model with thirteen random variables, respectively. The effects of larger uncertainties and polynomial order are also highlighted. The efficiency of the PDD method is verified via comparing with the MCS method. The applications of the PDD method to the rotor model can provide guidance to further study the rotor systems supported by bearings.