An improved weighted response surface method based on vector projection sampling

The response surface method was proposed as a collection of statistical and mathematical techniques that are useful for modeling and analyzing a system which is influenced by several input variables. This method gives an explicit approximation of the implicit limit state function of the structure through a number of deterministic structural analyses. However, the position of the experimental points is very important to improve the accuracy of the evaluation of failure probability. In the paper, the experimental points are obtained by using Givens transformation in such way these experimental points nearly close to limit state function. A Numerical example is presented to demonstrate the improved accuracy and computational efficiency of the proposed method compared to the classical response surface method. As seen from the result of the example, the proposed method leads to a better approximation of the limit state function over a large region of the design space, and the number of experimental points using the proposed method is less than that of classical response surface method.

An improvement of the response surface method based on reference points for structural reliability analysis

A new response surface method (RSM) for slope reliability analysis was proposed based on Gaussian process (GP) machine learning technology. The method involves the approximation of limit state function by the trained GP model and estimation of failure probability using the first‐order reliability method (FORM). A small amount of training samples were firstly built by the limited equilibrium method for training the GP model. Then, the implicit limit state function of slope was approximated by the trained GP model. Thus, the implicit limit state function and its derivatives for slope stability analysis were approximated by the GP model with the explicit formulation. Furthermore, an iterative algorithm was presented to improve the precision of approximation of the limit state function at the region near the design point which contributes significantly to the failure probability. Results of four case studies including one nonslope and three slope problems indicate that the proposed method is more efficient to achieve reasonable accuracy for slope reliability analysis than the traditional RSM.

Abstract: The surrogate model method is widely used in structural reliability analysis to approximate complex limit state functions. Accurate results can only be obtained when the surrogate model for the limit state function is approximated sufficiently close to the failure region. This study develops a novel local approximation method for efficient structural reliability assessment. The adaptive Markov chain simulation is utilized to generate samples in the failure region (the “region of most interest”). The support vector regression technique is then used to obtain an explicit approximation of the original complex limit state function around the region of most interest. Four examples are given to demonstrate the application and efficiencies of the proposed method.

The response surface method (RSM) is a powerful approach for carrying out reliability analysis for complicate engineering with implicit limit state functions. The quality of a response surface mainly depends on the choice of response surface function and the selection of sample points. To investigate the influence of the types of response surface functions and to reduce the computation efforts, two new sampling methods and a hybrid RSM are proposed. In the hybrid RSM, four types of response surface functions and three sampling methods are involved, and each Response surface function can be connected to each sampling method. The four response surface functions are quadratic polynomial without cross terms (PN1), quadratic polynomial with cross terms, radial basic function network (RBFN) and support vector machine (SVM). The three RSMs using the traditional sampling method, the new iterative sampling method and the new experiment design method are RSM1, RSM2 and RSM3, respectively. The accuracy and efficiency of different RSMs are illustrated through three examples. When an iterative method is used for locating sample points, the PN1-based RSM2 is proposed for its accuracy and efficiency. And when an experiment design method is used for locating sample points, the RBFN- or SVM-based RSM3 is suggested because the RBFN or SVM is suitable for global fitting.

In order to consider high-order effects on the actual limit state function, a new response surface method is proposed for structural reliability analysis by the use of high-order approximation concept in this study. Hermite polynomials are used to determine the highest orders of input random variables, and the sampling points for the determination of highest orders are located on Gaussian points of Gauss-Hermite integration. The cross terms between two random variables, only in case that their corresponding percent contributions to the total variation of limit state function are significant, will be added to the response surface function to improve the approximation accuracy. As a result, significant reduction in computational cost is achieved with this strategy. Due to the addition of cross terms, the additional sampling points, laid on two-dimensional Gaussian points off axis on the plane of two significant variables, are required to determine the coefficients of the approximated limit state function. All available sampling points are employed to construct the final response surface function. Then, Monte Carlo Simulation is carried out on the final approximation response surface function to estimate the failure probability. Due to the use of high order polynomial, the proposed method is more accurate than the traditional second-order or linear response surface method. It also provides much more efficient solutions than the available high-order response surface method with less loss in accuracy. The efficiency and the accuracy of the proposed method compared with those of various response surface methods available are illustrated by five numerical examples.

PurposeBecause of the high computational efficiency, response surface method (RSM) has been widely used in structural reliability analysis. However, for a highly nonlinear limit state function (LSF), the approximate accuracy of the failure probability mainly depends on the design point, and the result is that the response surface function composed of initial experimental points rarely fits the LSF exactly. The inaccurate design points usually cause some errors in the traditional RSM. The purpose of this paper is to present a hybrid method combining adaptive moving experimental points strategy and RSM, describing a new response surface using downhill simplex algorithm (DSA-RSM).Design/methodology/approachIn DSA-RSM, the operation mechanism principle of the basic DSA, in which local descending vectors are automatically generated, was studied. Then, the search strategy of the basic DSA was changed and the RSM approximate model was reconstructed by combining the direct search advantage of DSA with the reliability mechanism of response surface analysis.FindingsThe computational power of the proposed method is demonstrated by solving four structural reliability problems, including the actual engineering problem of a car collision. Compared to specific structural reliability analysis methods, the approach of modified DSA interpolation response surface for structural reliability has a good convergent capability and computational accuracy.Originality/valueThis paper proposes a new RSM technology based on proxy model to complete the reliability analysis. The originality of this paper is to present an improved RSM that adjusts the position of the experimental points judiciously by using the DSA principle to make the fitted response surface closer to the actual limit state surface.

In structural reliability analysis, the response surface method is widely adopted because of its numerical efficiency. It should be understood that the response function must approximate the actual limit state function accurately in the main region influencing failure probability where it is evaluated. However, the size of main region influencing failure probability was not defined clearly in current response surface methods. In this study, the concept of sub-region of interest is constructed, and an improved response surface method is proposed based on the sub-region of interest. The sub-region of interest can clearly define the size of main region influencing failure probability, so that the accuracy of the evaluation of failure probability is increased. Some examples are introduced to demonstrate the efficiency and the accuracy of the proposed method for both numerical and implicit limit state functions.

This paper presents an algorithm for structural reliability with the response surface method. For this aim, an approach with three stages is proposed named as improved response surface method. In the algorithm, firstly, a quadratic approximate function is formed and design point is determined with First Order Reliability Method. Secondly, a point close to the exact limit state function is searched using the design point. Lastly, vector projected method is used to generate the sample points and Second Order Reliability Method is performed to obtain reliability index and probability of failure. Five numerical examples are selected to illustrate the proposed algorithm. The limit state functions of three examples (cantilever beam, highly nonlinear limit state function and dynamic response of an oscillator) are defined explicitly and the others (frame and truss structures) are defined implicitly. ANSYS finite element program is utilized to obtain the response of the structures which are needed in the reliability analysis of implicit limit state functions. The results (reliability index, probability of failure and limit state function evaluations) obtained from the improved response surface are compared with those of Monte Carlo Simulation, First Order Reliability Method, Second Order Reliability Method and Classical Response Surface Method. According to the results, proposed algorithm gives better results for both reliability index and limit state function evaluations.

An important step when designing and assessing the reliability of existing structures and/or structural elements is to calculate the reliability level described by failure probability or reliability index. Since calculating the structural response of complex systems such as bridges is usually a time-consuming task, the utilization of approximation methods with a view to reducing the computational effort to an acceptable level is an appropriate solution. The paper introduces a small-sample artificial neural network-based response surface method. An artificial neural network is used as an approximation (a so-called response surface) of the original limit state function. In order to be as effective as possible with respect to computational effort, a stratified Latin hypercube sampling simulation method is utilized to properly select training set elements. Subsequently, the artificial neural network-based response surface is utilized to calculate failure probability. To increase the accuracy of the determined failure probability, the response surface can be updated close to the failure region. This is performed by finding a new anchor point, which lies close to the design point of the limit state function. The new anchor point is then used to prepare the updated training set. The efficiency of the proposed method is tested for different training set sizes using a nonlinear limit state function taken from the literature, and the reliability assessment of three concrete bridges, one with explicit and two with implicit limit state functions in the form of finite element method models.

The use of structural reliability methods with implicit limit state functions (LSFs) shows the increasing demand for efficient stochastic analysis tools, because the structural behaviour predictions are often obtained by finite element analysis. All stochastic mechanics problems can be solved by Monte Carlo simulation method, nevertheless, in most cases, at a prohibitively high computational cost. Several approximations can be achieved using first-order reliability method (FORM) and second-order reliability method and response surface methods. In this paper, a method that combines the FORM and Kriging interpolation models, as response surface, is proposed. The prediction accuracy of the Kriging response surface obtained from different sampling techniques is assessed, and the failure probability estimates calculated by the FORM using the classical second-order polynomial regression models and the Kriging interpolation models as surrogates of nonlinear LSFs are compared. The usefulness and efficiency of the reliability analysis using the Kriging response surface are demonstrated on the basis of existing results available in the literature and with an application problem of a stiffened plate structure with initial imperfections.

One of popular offshore wind turbine foundations, monopile has been largely used because of their cheapness and constructability. Monopile is subjected by large cyclic lateral loadings such as wind, wave and current. The P-y curve method which represents a relationship between lateral pile displacement and passive soil resistance along the leading face of the pile is known as a highly reasonable method to analyses lateral behaviors of pile. Performing reliability analysis of a monopile foundation using nonlinear p-y curves is difficult because the limit state functions (LSF) of pile head deformation and rotation angle of pile are in implicit forms. To solve such problem with implicit LSF, a response surface method (RSM) could be used. Basic concept of RSM is to approximate the limit state boundary using an explicit function of the random variables. In case of using RSM, reliability analysis could be very simple, but the accuracy of analysis due to approximation depends much on the linearity of the LSF and on the distance of the axial points in the failure space. This paper investigates the best combinations of the RSM techniques with a reliability analysis of monopile for an offshore wind foundation.

In order to obtain reliable structural design, it is of extreme importance to evaluate the failure probability, safety levels of structure (reliability analysis) and the effect of a change in a variable parameter on structural safety (sensitivity analysis) when uncertainties are considered. With a computationally cheaper approximation of the limit state function, various response surface methods (RSMs) have emerged as a convenient tool to solve this especially for complex problems. However, the traditional RSMs may produce large errors in some conditions especially for those highly non-linear limit state functions. Instead of the traditional least squares approximation, in the present paper, a new RSM is proposed which employs moving Kriging interpolation, based on the axial experimental points selected from the region where the most probable failure point (MPFP) is likely to exist to construct the substitute response surface. The proposed method is illustrated by comparing with the results obtained from first order reliability method (FORM) and other conventional RSMs with reference to specific structural reliability analysis problems. The results show that the proposed method improves the accuracy of the reliability analysis with a reasonable computational cost, and could more quickly approach the exact solution.

Response surface method (RSM) is widely used in structural reliability analysis with implicit performance function (PF) which requires formidable computational effort. The ill conditioned coefficient matrix of normal equation in classical RSM prevents it from being used in high order conditions. The stochastic response surface method (SRSM), deriving from classical RSM, offers one alternative to solve this problem. Yet the regression method of conventional SRSM is based on normal least square method which ignores the different significance of each sample point through which the response surface function (RSF) is formed. To yield RSF close to the limit state which leads to better estimation of probability of failure, this paper introduces the weighted regression into SRSM and several examples with hypothetic explicit PF are given to test the performance of SRSM. In addition, we use this method in the fatigue reliability analysis of crankshaft with implicit PF. All these examples demonstrate the advantages of the proposed method.

To obtain the reliability index or failure probability of a slope, several values of limit state function (LSF) need to be computed. For complex slopes, the LSF of slope stability analysis is usually an implicit function, which is hard to calculate. This study proposes an improved response surface method (RSM) where only one response surface function (RSF) is generated. The Least squares support vector regression (LSSVR), Radial basis function neural network (RBFNN), and Gaussian process regression (GPR) are used as the RSF of the actual LSF, and a method for locating the sampling centre for generating random samples using Latin Hypercube Sampling is presented. Benefiting from the improved sampling method and the flexibility and global fitting properties of the LSSVR, RBFNN, and GPR, the RSF needs to be generated only once and it can simulate the actual LSF very well even for highly nonlinear LSF. Based on the RSF using LSSVR, RBFNN, or GPR, the Monte Carlo simulation method (MCSM) is adopted to perform reliability analysis to assure the accuracy of reliability index. The excellent performance of the proposed RSM is verified by four examples, including a problem with highly nonlinear LSF and three slope examples with increasing complexities.