Adaptive probability analysis using an enhanced hybrid mean value method

Size: px
Start display at page:

Download "Adaptive probability analysis using an enhanced hybrid mean value method"

Transcription

1 Research Paper Struct Multidisc Optim 28, ) DOI /s Adaptive probability analysis using an enhanced hybrid mean value method B.D. Youn, K.K. Choi, L. Du Abstract This paper proposes an adaptive probability analysis method that can effectively generate the probability distribution of the output performance function by identifying the propagation of input uncertainty to output uncertainty. The method is based on an enhanced hybrid mean value HMV+) analysis in the performance measure approach PMA) for numerical stability and efficiency in search of the most probable point MPP). The HMV+ method improves numerical stability and efficiency especially for highly nonlinear output performance functions by providing steady convergent behavior in the MPP search. The proposed adaptive probability analysis method approximates the MPP locus, and then adaptively refines this locus using an a posteriori error estimator. Using the fact that probability levels can be easily set a priori in PMA, the MPP locus is approximated using the interpolated moving least-squares method. For refinement of the approximated MPP locus, additional probability levels are adaptively determined through an a posteriori error estimator. The adaptive probability analysis method will determine the minimum number of necessary probability levels, while ensuring accuracy of the approximated MPP locus. Several examples are used to show the effectiveness of the proposed adaptive probability analysis method using the enhanced HMV+ method. Nomenclature X Random parameter; X =[X 1,X 2,...,X n ] T x Realization of X; x =[x 1,x 2,...,x n ] T U Independent and standard normal random parameter; U =[U 1,U 2,...,U n ] T u Realization of U; u =[u 1,u 2,...,u n ] T β Probability level GX) Performance function; the design is considered a fail if GX) < 0 ĜX) Approximate performance function of GX) t Parametric coordinate along the arc of an n- dimensional sphere u Most probable point n Normalized steepest-descent direction of the performance function U j Approximate MPP locus h Basis vector for MPP locus approximation a Coefficient vector for MPP locus approximation M Full moment matrix in approximating MPP locus B Half moment matrix in approximating MPP locus W Weight matrix in approximating MPP locus A posteriori error estimator of MPP locus ε MPPL Key words TS a Received: Revised manuscript received: April 2004 Published online: 2004 Springer-Verlag 2004 B.D. Youn, K.K. Choi,L.Du Center for Computer-Aided Design and Department of Mechanical Engineering, College of Engineering, The University of Iowa, Iowa City, IA ybd@ccad.uiowa.edu, kkchoi@ccad.uiowa.edu, liudu@ccad.uiowa.edu 1 Introduction The fact that the decision-making process must be made under various uncertainties underscores the need for more effective output probability analysis methods that can identify the propagation of input uncertainty to output uncertainty. To identify the uncertainty propagation of an engineering system, various tools for output probability analysis have been developed: sampling methods, moment estimation methods, most probable point MPP)- based methods, etc. Due to its generality and simplicity, Monte Carlo simulation Rubinstein 1981) has been extensively used TS a Please add keywords.

2 2 as one of sampling methods for output probability analysis. However, given the need for a large sample size to obtain the probability distribution of the output performance function, this method requires intensive computation and is expensive. Several alternative sampling methods, so called variance reduction methods, have been proposed: an importance sampling method Melchers 1989), an adaptive sampling method Bucher 1988), an adaptive importance sampling method Wu 1994), etc. The basic idea behind these methods is to reduce the variance in the estimate during the simulation procedure by using the probability density function of an importance sampling. Thus, the probability density function must be properly selected to yield accurate estimates for high probability levels with possibly a small number of samples Melchers 1989; Bucher 1988; Wu 1994). It has been pointed out Du and Chen 2001) that the adaptive sampling method should choose a proper probability density function, which is the original density f X x) conditional on the failure domain. However, even though the variance reduction method requires a smaller number of samples than Monte Carlo simulation, the computational requirement is still quite significant for large-scale problems. To alleviate such a high computational burden, several moment estimation methods were proposed using a quadrature formula for numerical integration Evans 1972; Evans and Falkenburg 1976), or using a numerical integration scheme and a design of experiment DOE) D Errico and Zaino 1988; Seo and Kwak 2003). The former approximates statistical moments of the output performance function by directly integrating statistical moments of its Taylor series approximations using the quadrature formula. The latter estimates those moments by employing numerical integration with integration points and their corresponding weights, and performing DOE at those integration points. Even though these methods are accurate for high-order statistical moments, they are still expensive for large-scale applications with a sizeable number of uncertain parameters D Errico and Zaino 1988; Seo and Kwak 2003). Two MPP-based methods were developed Wu and Wirsching 1987; Du and Chen 2001) based on the advanced first-order second-moment method that was introduced by using a rotationally invariant reliability measure Hasofer and Lind 1974): the reliability index approach or G-level approach) and the performance measure approach or P-level approach). In this paper, the performance measure approach PMA) *Tu and Choi 1999; Youn et al. 2001, 2003) or fixed norm approach Lee and Kwak )) is used for output probability analysis, since it is difficult in the reliability index approach RIA) to set proper output performance function levels a priori, whereas the probability levels β i [ 6, 6] can be easily set a priori in PMA. For MPP search, the advanced mean value method Wu and Wirsching 1987) is shown to be either inefficient or unstable for a concave performance function Youn et al. 2001, 2003; Youn and Choi 2003). Thus, a hybrid mean value HMV) method was developed in Youn et al. 2001, 2003) to handle both convex and concave performance functions more effectively. However, it is shown in this paper that the HMV method could fail to converge for highly nonlinear output performance functions. To improve numerical stability and efficiency, an enhanced HMV HMV+) method is proposed in this paper. The HMV+ method is shown to improve numerical stability and efficiency for highly nonlinear output performance functions by providing steady convergent behavior in the MPP search. A probability analysis method was proposed in Du and Chen 2001) by approximating the most probable point MPP) locus, where the MPP locus is approximated using the extrapolated least-squares method, and the number of probability levels are predetermined. An adaptive probability analysis method is proposed in this paper by employing an a posteriori error estimator. That is, the MPP locus is refined by adaptively determining additional probability levels using an a posteriori errorestimator. Thus, this method allows us to use a minimum number of necessary probability levels, while controlling accuracy even for highly nonlinear performance functions. For approximation of the MPP locus of probabilistic structural responses, using the fact that probability levels β i [ 6, 6] can be easily set a priori in PMA, the MPP locus is approximated using interpolated least squares and moving least-squares methods Lancaster and Salkauskas 1986). The comparison between leastsquares and moving least-squares methods is discussed in terms of the rate of convergence. The proposed method is demonstrated using numerical examples. Comparisons are also made to other probability analysis methods available in the literature. 2 Enhanced HMV HMV+) method for output probability analysis Probability analysis provides a complete probability distribution of output performance function by identifying the propagation of input uncertainty to output uncertainty. In this paper, the MPP-based method is developed to carry out the output probabilistic structural analysis, which employs a series of MPP search carried out using the proposed HMV+ method. 2.1 Output probability analysis When an engineering system is subject to a variety of uncertainties, an output probability analysis provides a valuable probability distribution or density) function, or statistical moments of the output performance function for a decision-based design Chen et al. 2000). The probability distribution function is obtained by identifying the propagation of input uncertainty to output

3 3 uncertainty as Madsen et al. 1986; Haldar and Mahadevan 2000) F G g)= f X x)dx 1...dx n 1) where the uncertainty F X x) orf X x)) of input X is propagated to the uncertainty F G g) of output GX). In this paper, the probability distribution function is approximated by employing PMA, in which HMV+ searches the MPP x i corresponding to a given probability F G g i )=β i where gx i )=g i. Using the MPP-based method, the output probability analysis in 1) can be redefined as F G g i )= f X x)dx 1...dx n = GX) g i f X x)dx 1...dx n =Φ β i ), GX) g i 0 i =1,...,NPL 2) where Φ ) is the standard normal distribution and NPL is the number of probability levels. Taking an inverse transformation of F G for PMA, the corresponding output performance function level is computed using g i = F 1 G Φ β i)) = gx i ), i=1,...,npl 3) The MPPs x i are obtained at a series of probability levels using HMV+. Once the MPP locus is obtained, the probability distribution or density) function of output is approximately constructed by associating the probability levels and their output performance functions, as shown in Fig. 1. TS b For many engineering applications, it is desirable to obtain accurate tail end of probability distribution especially for high levels of probability, i.e. 3σ,whichcanbe estimated using an advanced first-order second-moment method Youn and Choi 2003). The advanced first-order second-moment method Hasofer and Lind 1974) is used in this paper with an emphasis on improving numerical efficiency, while maintaining a desirable level of accuracy. A transformationt: X U is introduced to map the original random space X to a standard, uncorrelated normal random space U. If the random variables X i are mutually independent, the transformation is Rackwitz and Fiessler 1978) T : U i =Φ 1 F Xi x i ) ), i=1 n 4) where F Xi is the probability distribution function of X i. To simplify the probability integration in 2), the limit state surface G g i = 0 can be expressed as a first-order approximation. This is referred to as the first-order reliability method FORM) Hasofer and Lind 1974; Madsen et al. 1986; Haldar and Mahadevan 2000), as shown in Fig. 2. UsingFORM,2)or3)canbesolvedbyformulating an optimization with one equality constraint Tu and Choi 1999; Youn et al. 2001, 2003; Lee and Kwak ) to obtain g i = gu i ), if β i > 0, maximize GU) subject to U = β i, for i =1,...,NPL if β i < 0, minimize GU) subject to U = β i, for i =1,...,NPL 5) Fig. 1 MPP-based method for output probability analysis TS b Please check quality of all your figures send higher resolution versions of Figs. 1, 3 5 and 7 10 if possible.

4 4 Fig. 2 First- and second-order reliability methods Once the MPP u i = T x i ) is obtained for a given probability level, the corresponding output performance function level is computed using g i = gu i ). The MPP-based method using PMA is very effective for the output probability analysis since it allows one to set probability levels β i [ 6, 6] a priori, without knowing the output performance function levels. 2.2 Enhanced hybrid mean value HMV+) method Even though the HMV method Youn et al. 2001, 2003) performs well for convex or concave performance functions, it could fail to converge for highly nonlinear output performance functions. To improve numerical stability and efficiency, an enhanced HMV HMV+) method is proposed by revising the previous HMV algorithm. In this new HMV+ algorithm, if the value of the performance function is decreased at the next search point, the performance function is approximated along the arc with constant probability level β ι ) between the current point and next search point to find a new search point where the approximated performance function has the maximum value. Step 1. Set the iteration counter k = 0 and probability level β i with the convergence parameter ε =10 4.Letu 0) HMV+ = 0. Step 2. Calculate the performance function gu k) HMV+ ) and its sensitivity gu k) HMV+ ). Step 3. Check the Karush Kuhn Tucker condition: sgnβ u k) HMV+ i) u k) HMV+ nk) HMV+ 1 ε for k 2, where n is the normalized steepest ascent direction of GU); and sgnβ ι ) is the signum function, such that it is +1 if β ι > 0and 1 ifβ ι < 0. If it is satisfied, then stop. [ ] Step 4. If k 2 and sgnβ i ) gu k) HMV+ ) guk 1) HMV+ ) < 0, interpolate the performance function along the arc region between these two search points and obtain a new search point u k+1) HMV+ by maximizing the approximate performance function. Otherwise, use the HMV method to obtain a new search point u k+1) HMV+.Letk = k +1 and go to Step 2. In the HMV+ method, the arc-interpolation method is employed when the performance function value is decreased at the next search point for β i > 0orincreased at the next search point for β i < 0. Performance function and its sensitivity values at two search points,, are used to interpolate the performance function along the arc region between these two search points. For the interpolation, a parametric coordinate t is introduced as u k) HMV+ and uk 1) HMV+ U = st)u k 1) + tu k) and U = β t, s,t 0 6) where s + t)= tuk 1) u k) + tu k 1) u k)) 2 +1 t2 )β 4 t tu s t)= tuk 1) u k) k 1) u k)) 2 +1 t2 )βt 4 βt 2 7) The sensitivity of the performance function with respect to a parametric coordinate t can be obtained using a chain rule as β 2 t

5 dg dt = G U i U i t = G ) ds U i dt uk 1) + u k) 8) value for β i > 0 or smaller performance function value for β i < 0, needs to be selected. 5 In 8), the sensitivities of performance function are evaluated at two search points u k 1) and u k) as dg u k 1)) = dt g u k 1)) ds ± ) t) dt u k 1) + u k) = t=0 g u k 1)) u k 1) u k) ) βt 2 u k 1) + u k) dg u k)) = g u k)) ds ± ) t) dt dt u k 1) + u k) = t=1 g u k 1)) β 2 ) t u k 1) u k) uk 1) + u k) 2.3 Numerical examples of MPP search for probability level β ι Three numerical examples are used to demonstrate the effectiveness of the HMV+ method for MPP search at probability level β ι : two for numerical efficiency and one for numerical stability. Example 1: nonlinear performance function 1 The first example is from Youn et al. 2001, 2003), which is given as Then, gu k 1) ),gu k) dgu ), k 1) ) dt, and dguk) ) dt are used to interpolate the performance function using a cubic polynomial as Gt)=a 0 + a 1 t + a 2 t 2 + a 3 t 3 9) The next search point u k+1) is obtained where the approximated performance function G is maximum, as u k+1) = st )u k 1) + t u k) at t = t Gt) is maximum, for β i > 0 where Gt) is minimum, for β i < 0 10) Note that st) may not be unique for one value of t, if βt t>1s, t > 0; s, t 4 βt 4 uk 1) u k) ) 2 ), as shown in 7). This could be true when the angle composed of three points u k 1), u 0), and u k) is more than 90,whichcan be expressed mathematically as u k 1) u k) < 0. One of the two s values, that yields greater performance function Fig. 3 Performance function contour and MPP search iterations using the HMV+ method for Example 1 Table 1 Results of MPP search for Example 1 Iteration X 1 X 2 g x k)) dg k 1)/ dt dg k)/ dt s t MPP

6 6 Table 2 Comparison for numerical efficiency of the HMV+ method Method gx ) x NFE/NSE HMV , 3.161) 17/17 Proposed HMV , 3.156) 8/8 MFD , 3.156) 40/6 SLP , 3.155) 18/12 SQP , 3.157) 18/8 GX)= 0.3X 2 1 X 2 + X 2 0.8X ) where X i N1.2, 0.42), X i N1.0, 0.42), and β t =6.0. The performance function contour and MPP search iterations, using the HMV+ method, are shown in Fig. 3. As shown in Fig. 3, the response is highly nonlinear and contains both saddle point and regions with high curvature. For numerical efficiency, the HMV+ method is compared to some general optimization algorithms: a modified feasible direction MFD), a sequential linear programming SLP), and a sequential quadratic programming SQP). It is shown in Table 1 that the proposed HMV+ method converges in seven iterations. Three interpolations are employed at 2 nd,4 th,and6 th iterations. On the other hand, the HMV method and the other general optimization algorithms converge slowly, as shown in Table 2. The last column in Table 2 shows the number of total function evaluations NFE) and sensitivity evaluations NSE). The total number of function and sensitivity evaluations is the least for the HMV+ method. Example 2: nonlinear performance function 2 Another highly nonlinear performance function is from Du et al. 2003), which is given as GX)= 4+X ) 2 X ) 3 X ) 4 + X 2 12) where X 1 N0.0, 1.0), X 2 N0.0, 1.0), and β t =3.0. The performance function contour and MPP search iterations, using the HMV+ method, are shown in Fig. 4. As Fig. 4 Performance function contour and MPP search iterations using the HMV+ method for Example 2 shown in Fig. 4, the response is highly nonlinear with high curvature contours. Table 3 illustrates that the proposed HMV+ method converges in five iterations with two interpolations at the second and fourth iterations. On the other hand, as shown in Table 4, the HMV method, SLP, and SQP fail to find the MPP, while MFD converged very slowly. Thus, the HMV+ method is shown to be numerically stable and efficient in the MPP search. The result of the most probable point of inverse reliability MPPIR) method Du et al. 2003) for the same problem is shown in Table 4, requiring 20 function evaluations. Example 3: nonlinear performance function 3 A vehicle side-impact is employed to perform MPP search for a high-dimensional nonlinear response. The problem isdescribedindetailinyounet al. 2004). The velocity Table 3 Results of MPP search for Example 2 Iteration X 1 X 2 g x k)) dg k 1)/ dt dg k)/ dt s t MPP

7 7 of the door is used as the performance function, which is expressed as GX)= X 3 X X 5 X X 9 X X 9 X X ) Table 6 Comparison for numerical efficiency of the HMV+ method Method gx ) NFEA HMV /59 Proposed HMV /9 MFD /7 SLP /43 SQP /12 where β t =3.0, X i N1.0, 0.05) for i =1to7, X i N0.3, 0.006) for i =8, 9, and X i N0.0, 10.0) for i =10, 11. Table 5 provides the MPP search results of the proposed HMV+ method in X-space. Throughout the MPP search, the variables X 1, X 2, X 4,andX 8 remain unchanged. The HMV+ method converges in eight iterations along with three interpolations at the the second, fourth and seventh iterations. Other methods shown in Table 6 converge to the same MPP, but are not as efficient as the MPP+ method. Figure 5 shows the performance function contour and MPP search iterations, using the HMV+ method, on the X 10 X 11 hyperplane by setting X i = x i for i =1to9. Fig. 5 Performance function contour and MPP search iterations using the HMV+ method for Example 3 Table 4 Comparison for numerical efficiency of the HMV+ method Method gx ) x NFEA HMV Diverged Proposed HMV , 2.679) 6/6 MPPIR , 2.679) 20/0 MFD , 2.679) 55/5 SLP , 2.996) 16/11 SQP , 2.996) 12/4 3 Adaptive probability analysis An adaptive probability analysis method is proposed in this paper, by taking advantage of the proposed HMV+ method, to approximate the probability distribution of the output performance function. To this end, two methods are integrated in this paper: an interpolated instead of extrapolated) MPP locus approximation using the moving least-squares method; and an adaptive method to choose probability levels effectively depending on the nonlinearity of probabilistic output performance functions. In the interpolated method, the approxima- Table 5 Results of MPP search for Example 3 Iter X 3 X 5 X 6 X 7 X 9 X 10 X 11 g x k)) dg k 1 /dt dg k /dt MPP

8 8 Fig. 6 Flow chart of adaptive probability analysis tion is made in the region surrounded by known samples, whereas in the extrapolated method, the approximation is also made for outside this region. The interpolated MPP locus approximation is studied by comparing the least-squares and moving least-squares methods. For the adaptive method, an a posteriori error estimator is defined as the difference between the initial search point on the approximate MPP locus and the MPP point obtained using the HMV+ method at the selected probability level β ι. The MPP locus is refined by adaptively adding more probability levels until the a posteriori error is small enough at all selected probability levels. Thus, the adaptive method is a closed-loop probability analysis approach, as shown in Fig. 6, unlike the open-loop probability analysis approach where all probability levels β i [ 6, 6], i=1 n, needs to be predetermined. Detailed discussions of this proposed method are presented in the following sections. 3.1 MPP locus approximation In the open-loop probability analysis approach, it is difficult to set a proper number of probability levels without knowing the degree of nonlinearity of the performance function. That is, too many probability levels will be expensive, while too few levels may not yield accurate probability analysis. Furthermore, the MPP search at one probability level is performed independently from the other probability level. Therefore, much of the valuable information generated during the MPP search at one probability level is ignored at other probability levels. These shortcomings can be overcome using the interpolated approximation of the MPP locus, which provides good initial search points for MPP search for the next probability level. Error analyses of the MPP locus approximation are carried out for the least-squares and moving least-squares methods Least squares method The MPP locus approximation using the least-squares method can be formulated as NB U j β)= h i β)a ij = h T a j, forj =1,...,NRV 14) i=1 where NB is the number of basis monomials, NRV is the number of random parameters, a j is the coefficient vector for the j th random parameter, and h is the basis vector. Mutually independent monomials are used as basis functions. In this paper, a quadratic polynomial basis is used to approximate the MPP locus. Higher-order polynomial bases are avoided due to incorrect level of oscillation near the boundaries. To obtain the coefficient vector, a residual E LS can be defined as NPL E j LS = NPL I=1 where I=1 U j β I ) U j β I )) 2 = h T a j U j β I ) ) 2 =Haj U j ) T Ha j U j ) 15)

9 9 h 1 β 1 ) h NB β 1 ) H =..... h 1 β NPL ) h NB β NPL ) U j = [ U j β 1 ) U j β NPL ) ] 16) Here, NPL represents the number of probability levels and the subscript LS denotes the least-squares method. The residual E j LS is a positive definite quadratic form and the necessary condition to minimize E j LS is aj E j LS =0 17) By solving 17), the coefficient vector for the j th random parameter is obtained as a j = H T H ) 1 H T U j, j =1,...,NRV 18) Thus, the MPP locus approximation using the leastsquares method can be expressed as U j β)=h T H T H ) 1 H T U j, j =1,...,NRV 19) Moving least-squares method Lancaster and Salkauskas 1986) The MPP locus approximation using the moving leastsquares method is NB U j β)= h i β)a ij β)=h T β)a j β) i=1 for j =1,...,NRV 20) where h is the basis vector and a j β)=[a 1j β),a 2j β),...,a NBj β)] T is the j th coefficient vector, which is a function of the probability level β. Mutually independent monomials are used as basis functions. In this paper, a quadratic polynomial basis is used to approximate the MPP locus, which is found to provide a sufficiently accurate result, since the MPP locus is not highly oscillatory even if the output performance function is highly nonlinear. To obtain the coefficient vector, the residual E MLS can be defined as NPL E j MLS β)= NPL I=1 I=1 [ 2 wβ β I ) U j β I ) U j β I )] = wβ β I ) [ h T a j U j β I ) ] 2 = [Ha j β) U j ] T Wβ)[Ha j β) U j ] 21) where h 1 β 1 ) h NB β 1 ) H =..... h 1 β NPL ) h NB β NPL ) U j = [ U j β 1 ) U j β NPL ) ] wβ β 1 ) wβ β 2 ) 0 Wβ)= ) 0 0 wβ β NPL ) Here, NPL is the number of probability levels, and the subscript MLS denotes the moving least-squares method. Note that the approximation U j β I ) draws closer to U j β I ) at those points β I with a relatively large weight wβ β I ) by introducing a weight inversely proportional to the distance between the sample and interpolation points. The necessary condition to minimize E j MLS is aj E j MLS =0 23) By solving 23) the coefficient vector for the j th random parameter is obtained as a j β)=m 1 β)bβ)u j, j =1,...,NRV 24) where Mβ) is referred to as the moment matrix given by Mβ)=H T Wβ)H and Bβ)=H T Wβ) 25) Thus, MPP locus approximation using the moving leastsquares method is U j β)=h T β)m 1 β)bβ)u j, j =1,...,NRV 26) Since the coefficient vector is a function of the probability level, 24) must be solved at different probability levels of interest, unlike the least-squares method in 18). If the weight matrix is set as the identity matrix, the moving least-squares approximation in 26) becomes the leastsquares approximation in 19). 3.2 Adaptive set of probability levels Using either the least-squares or moving least-squares method, the approximate MPP locus can be refined as the number of probability levels are increased. The adaptive method to add more probability levels is determined using an a posteriori error estimator, defined as ε MPPL = u 0 k u k β2 k ) where u 0 k and u k are the initial search point and MPP, respectively, at the current probability level β k,andε MPPL is the error measure of the MPP locus. CE c Please complete this sentence

10 10 Fig. 7 Adaptive set of probability levels In the proposed probability analysis method, an a posteriori error estimator is used to decide on the appropriateness of the number of probability levels for accuracy of the approximated MPP locus. That is, the MPP locus and its a posteriori error estimator make it possible to add more probability levels adaptively using the bisection method. The adaptive procedure continues to refine the MPP locus by adding number of probability levels until the a posteriori error is small for all probability levels. Detailed procedures involved in the adap- tive probability analysis are illustrated in Fig. 7. First, the MPP search is carried out at the probability levels ±1/2βH, where βh is predetermined, say βh = 6.0. The HMV+ method is used since it can efficiently and accurately perform MPP search for even a high probability level, e.g. βh = 6.0 shown in Example 1. Next, the MPP locus is approximated based on information obtained at probability levels, β = 0, ±3, ±6, as shown in Fig. 7a). Using the bisection method, the next probability levels, β = ±1.5, ±4.5, are selected and the cor-

11 11 responding MPP searches are carried out starting from the initial search point on the approximate MPP locus to find the MPPs at these probability levels. The a posteriori error estimator defined in 27) is then used to check the CE c. This estimator shows that the error of the MPP locus is unacceptable when β< 3.0, as illustrated in Fig. 7a). Next, the second approximated MPP locus is constructed using information at the probability levels β =0, ±1.5, ±3, ±4.5, ±6, as shown in Fig. 7b). At the same time, using information from the approximated MPP locus, MPP searches are carried out at the selected probability levels β = 3.75, 5.25) where the MPP locus is found inaccurate. The adaptive probability analysis constructed the third approximated MPP locus as shown in Fig. 7c). The accuracy is verified by the a posteriori error at all probability levels and the probability analysis is stopped when the accuracy is satisfied. Finally, using the MPP locus, the cumulative distribution function CDF), probability density function PDF), and statistical moments of the response are generated, as shown in Fig. 7d). The adaptive probability analysis method will accurately estimate the tail end of the probability distribution, since the it will tend to add more probability levels at the tail end of probability distribution rather than near its mean due to the a posteriori error. 3.3 Numerical procedure of adaptive probability analysis A numerical procedure illustrated in Fig. 7 for the proposed adaptive probability analysis is presented in this section. Step 1. Set the MPP locus counter to k = 0 and select the highest probability level β H. Select the convergence parameter ε REL for the MPP search, and ε MPPL for MPP locus approximation. Carry out MPP search at ±1/2β H and ±β H. Step 2. Select probability levels using the bisection method, where the initial search point is not close to the MPP, i.e. ε 1 >ε MPPL, and approximate the k +1 th MPP locus. Step 3. Obtain initial search points on the approximate MPP locus at the selected probability levels. Step 4. Carry out MPP searches starting from the initial search points at the selected probability levels. Step 5. Check to see if the a posteriori error estimator satisfies the error criteria ε>ε MPPL.Ifthe convergence is achieved at all probability levels, then stop. Otherwise, go to Step 2. 4 Numerical results and discussion Example 4: side-impact crashworthiness for MPP locus approximation The side-impact crashworthiness used in Example 3 is employed here for MPP locus approximation in the probability analysis. Two responses are used: a mildly nonlinear response abdomen load) and a highly nonlinear response velocity of door). This study also presents a comparison between the least-squares and moving least-squares methods for the MPP locus approximation. As the number of probability levels is increased, the a posteriori error ε MPPL in 27) is measured to observe the rate of convergence in approximating the MPP locus. From Fig. 8, it can be seen that the moving least-squares method shows a faster rate of convergence than the least-squares method in MPP locus approximation. The moving least-squares method is better in reproducing both local and global behaviors in the MPP locus than the least-squares method. As expected, Table 7 Numerical results of adaptive probability analysis No. of levels, I Probability level β i G No. of analyses Total number of analyses 28

12 12 a larger error in the MPP locus is observed for the highly Abdomen Load Vel. of Door nonlinear response, ε <ε MPPL MPPL. Example 5: mathematical example for adaptive probability analysis A performance function for output probability analysis is defined as GX)=X 1 0.3X 1 X 2 0.1X ) where two random parameters are defined as X i N0.0, 0.5), i =1,2,andβ H =6.0. The process of the output probability analysis is depicted in Fig. 7 of Sect. 3. It is observed in Fig. 7 that the adaptive probability analysis refines the MPP locus by adaptively adding probability levels only in the region where the a posteriori error is greater than the allowable amount. The adaptive probability analysis requires ten MPP searches excluding β i = 0) and 28 analyses, as shown in Table 7. For comparison, the exact MPP locus in Fig. 7 is constructed using 32 probability levels. Compared to Monte Carlo simulation with samples, the adaptive probability analysis is shown to yield accurate CDF, PDF, and statistical moments, as shown in Fig. 7d). Numerical efficiency of the adaptive probability analysis is compared to those of Monte Carlo simulation and NESSUS Southwest Research Institute 1996), as shown in Table 8. NESSUS fails to complete the output probability analysis, due to the failure of the MPP search for β<0. For the adaptive probability analysis, different MPP search methods are employed, where the AMV method diverged. On the contrary, the adaptive probability analyses with HMV and HMV+ converged, where the proposed HMV+ method enhances numerical efficiency and stability, compared to all other methods. Example 6: side-impact crashworthiness for adaptive probability analysis The same side-impact application used for Example 4 is considered for both abdomen load and door velocity. β H is set to 4.0 for both responses. The result of output probability analysis for abdomen load is shown in Table 9. Eight probability levels excluding β i =0) are required to complete the analysis. In addition to the probability level at β i = 0, the initial probability levels are selected at β i = ±2 and±4, and MPP searches are carried out starting from the initial point at the origin in U-space. When the next probability levels are added at β i = ±1 and±3, MPP searches are carried out starting from the initial points obtained from the approximated MPP locus. Thus, the MPP searches for β i = ±1 and±3 require fewer analyses than the ones for probability levels β i = ±2and±4, as shown in Table 9. Fig. 8 Errors in MPP locus, ε MPPL Table 8 Results of probability analysis for Example 5 Monte Carlo Adaptive Probability Analysis Method NESSUS simulation AMV HMV HMV+ No. of analyses Diverged Diverged 41 28

13 13 Fig. 9 CDF and statistical moments of abdomen load Fig. 10 CDF and statistical moments of velocity of door Table 9 Numerical results of adaptive probability analysis for abdomen load No. of levels, I Probability level β i G No. of analyses Total number of analyses 27 Table 10 Numerical results of adaptive probability analysis for velocity of door No. of levels, I Probability level β i G No. of analyses Total number of analyses 44 Table 11 Results of probability analysis for Example 6 Monte Carlo Adaptive Probability Analysis Method NESSUS simulation AMV HMV HMV+ Abdomen load Velocity of door Diverged Diverged Diverged 44

14 14 To verify numerical accuracy, CDF and statistical moments obtained using the adaptive probability analysis are compared to those obtained using Monte Carlo simulation with samples, as shown in Fig. 9. Note that a relatively larger error of the adaptive probability analysis occurs near 10% and 90%, which is mainly due to the error of FORM. The second-order reliability method SORM) might be able to reduce these errors. The result of output probability analysis for door velocity is shown in Table 10. Eleven probability levels excluding β i = 0) are required to complete the output probability analysis. The adaptive probability analysis of door velocity requires more probability levels than the abdomen load, since the former response is more nonlinear than the latter. For numerical accuracy the adaptive probability analysis is again compared to Monte Carlo simulation with samples, as shown in Fig. 10. Note that a relatively larger amount of error of the adaptive probability analysis occurs between 50% and 95%, which is again mainly due to the error of FORM. For a given performance function value, the adaptive probability analysis estimates smaller probability than Monte Carlo simulation, since the response is a concave function in this example. As observed in Table 11, the adaptive probability analysis with HMV+ for abdomen load is carried out efficiently, while the same eight probability levels are employed for NESSUS with less efficient result. In general, NESUSS is unable to set the minimum necessary number of probability levels. Moreover, it is found that the HMV+ method makes the adaptive probability analysis more efficient. For the highly nonlinear response of the door velocity, NESSUS and adaptive probability analyses methods with AMV or HMV fail to complete output probability analysis due to the failure of the MPP searches, whereas the proposed adaptive probability analysis method with HMV+ has carried out output probability analysis efficiently. 5 Conclusions This paper proposes an adaptive output probability analysis method that aids in the decision-based design process by efficiently and accurately identifying the propagation of the input uncertainty to the output uncertainty. In addition, the enhanced hybrid mean value HMV+) method is proposed to improve numerical stability and efficiency in the MPP search. Three examples are used to show numerical efficiency and accuracy of the HMV+ method, as compared to the original HMV method and general optimization algorithms. By using MPP locus approximation and an a posteriori error estimator, it is found that the adaptive probability analysis uses the least number of necessary probability levels adaptively, and to perform output probability analysis efficiently. A comparison study between the least-squares and moving least-squares methods shows that the latter converges faster in approximating the MPP locus. Two numerical examples are used to demonstrate the effectiveness of the proposed adaptive probability analysis using the HMV+ method, in terms of numerical efficiency and stability. It has also been found that numerical efficiency in adaptive probability analysis does not depend on the number of random parameters, but on the degree of nonlinearity of the performance function. Acknowledgements Research is partially supported by the Automotive Research Center sponsored by the U.S. Army TARDEC. References Bucher, C.G. 1988: Adaptive Sampling An Iterative Fast Monte Carlo Procedure. Struct Saf 5, Chen, W.; Lewis, K.; Schmidt, L. 2000: Decision-Based Design: An Emerging Design Perspective. Eng Valuation Cost Anal. special edn. on Decision-Based Design: Status & Promise 31 2), Du, X.; Chen, W. 2001: A Most Probable Point-Based Method for Efficient Uncertainty Analysis. Des Manuf 41), Du, X.; Sudjianto, A.; Chen, W. 2003: An Integrated Framework for Optimization Using Inverse Reliability Strategy. DETC-DAC48706, ASME Design Engineering Technical Conferences, Chicago, IL, September 2003 D Errico, J.R.; Zaino, N.A. 1988: Statistical Tolerancing Using a Modification of Taguchi s Method. Technometrics 304), Evans, D.H. 1972: An Application of Numerical Integration Techniques to Statistical Tolerancing, III-General Distributions. Technometrics 141), Evans, D.H.; Falkenburg, D.R. 1976: Computer Programs for the Quadrature Approximation for Statistical Tolerancing. J Qual Technol 82), Haldar, A.; Mahadevan, S. 2000: Probability. Reliability and Statistical Methods in Engineering Design. New York: Wiley Hasofer, A.M.; Lind, N.C. 1974: Exact and Invariant Second Moment Code Format. J Eng Mech Division ASCE 100, Lancaster, P.; Salkauskas, K. 1986: Curve and Surface Fitting; An Introduction. London: Academic Lee, T.W.; Kwak, B.M : A Reliability-Based Optimal Design Using Advanced First Order Second Moment Method. Mech Struct Mach 154), Madsen, H.O., Krenk, S.; Lind, N.C. 1986: Methods of Structural Safety. Englewood Cliffs, NJ: Prentice-Hall Melchers, R.E. 1989: Importance Sampling in Structural Systems. Struct Saf 6, 6 10 Rackwitz, R.; Fiessler, B. 1978: Structural Reliability under Combined Random Load Sequences. Comput Struct 9,

15 15 Rubinstein, R.Y. 1981: Simulation and Monte Carlo Method. New York: Wiley Seo, H.S.; Kwak, B.M. 2003: An Improved Reliability Analysis Using Design of Experiments and an Application to Tolerance Design. The 5 th World Congress of Structural and Multidisciplinary Optimization. Lido di Jesolo, Italy, May Southwest Research Institute 1996: NESSUS/FPI User s Manual, Version 2.3, San Antonio, TX Tu, J.; Choi, K.K. 1999: A New Study on Reliability Based Design Optimization. JMechDesASME1214), Wu, Y.-T. 1994: Computational Methods for Efficient Structural Reliability and Reliability Sensitivity Analysis AIAA J 328) Wu, Y.-T.; Wirsching, P.H. 1987: New Algorithm for Structural Reliability Estimation. J Eng Mech ASCE 1139), Youn, B.D.; Choi, K.K. 2003: Selecting Probabilistic Approaches for Reliability-Based Design Optimization. AIAA J 421), Youn, B.D.; Choi, K.K.; Gu, L.; Yang, R.-J. 2004: Reliability- Based Design Optimization for Crashworthiness of Side Impact. J Struct Multidisc Optim 273), Youn, B.D., Choi, K.K.; Park, Y.H. 2001, 2003: Hybrid Analysis Method for Reliability-Based Design Optimization. J Mech Des, ASME 1252), , 2003; Proceedings of 2001 ASME Design Engineering Technical Conferences: 27 th Design Automation Conference. Pittsburgh, PA

AN INVESTIGATION OF NONLINEARITY OF RELIABILITY-BASED DESIGN OPTIMIZATION APPROACHES. β Target reliability index. t β s Safety reliability index; s

AN INVESTIGATION OF NONLINEARITY OF RELIABILITY-BASED DESIGN OPTIMIZATION APPROACHES. β Target reliability index. t β s Safety reliability index; s Proceedings of DETC 0 ASME 00 Design Engineering Technical Conferences and Computers and Information in Engineering Conference Montreal, CANADA, September 9-October, 00 DETC00/DAC-XXXXX AN INVESTIGATION

More information

EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES

EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES 9 th ASCE Specialty Conference on Probabilistic Mechanics and Structural Reliability EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES Nam H. Kim and Haoyu Wang University

More information

A New Robust Concept in Possibility Theory for Possibility-Based Robust Design Optimization

A New Robust Concept in Possibility Theory for Possibility-Based Robust Design Optimization 7 th World Congresses of Structural and Multidisciplinary Optimization COEX Seoul, May 5 May 007, Korea A New Robust Concept in Possibility Theory for Possibility-Based Robust Design Optimization K.K.

More information

Structural Reliability

Structural Reliability Structural Reliability Thuong Van DANG May 28, 2018 1 / 41 2 / 41 Introduction to Structural Reliability Concept of Limit State and Reliability Review of Probability Theory First Order Second Moment Method

More information

An Adaptive Sequential Linear Programming Algorithm for Optimal Design Problems With Probabilistic Constraints

An Adaptive Sequential Linear Programming Algorithm for Optimal Design Problems With Probabilistic Constraints Kuei-Yuan Chan Assistant Professor Department of Mechanical Engineering, National Cheng Kung University, Tainan, Taiwan e-mail: chanky@mail.ncku.edu.tw Steven J. Skerlos e-mail: skerlos@umich.edu Panos

More information

A general procedure for rst/second-order reliability method (FORM/SORM)

A general procedure for rst/second-order reliability method (FORM/SORM) Structural Safety 21 (1999) 95±112 www.elsevier.nl/locate/strusafe A general procedure for rst/second-order reliability method (FORM/SORM) Yan-Gang Zhao*, Tetsuro Ono Department of Architecture, Nagoya

More information

Created by Erik Kostandyan, v4 January 15, 2017

Created by Erik Kostandyan, v4 January 15, 2017 MATLAB Functions for the First, Second and Inverse First Order Reliability Methods Copyrighted by Erik Kostandyan, Contact: erik.kostandyan.reliability@gmail.com Contents Description... References:...

More information

Basics of Uncertainty Analysis

Basics of Uncertainty Analysis Basics of Uncertainty Analysis Chapter Six Basics of Uncertainty Analysis 6.1 Introduction As shown in Fig. 6.1, analysis models are used to predict the performances or behaviors of a product under design.

More information

A Simple Third-Moment Method for Structural Reliability

A Simple Third-Moment Method for Structural Reliability A Simple Third-Moment Method for Structural Reliability Yan-Gang Zhao* 1, Zhao-Hui Lu 2 and Tetsuro Ono 3 1 Associate Professor, Nagoya Institute of Technology, Japan 2 Graduate Student, Nagoya Institute

More information

Parameter Estimation Method Using Bayesian Statistics Considering Uncertainty of Information for RBDO

Parameter Estimation Method Using Bayesian Statistics Considering Uncertainty of Information for RBDO th World Congress on Structural and Multidisciplinary Optimization 7 th - 2 th, June 205, Sydney Australia Parameter Estimation Method Using Bayesian Statistics Considering Uncertainty of Information for

More information

Techniques for Estimating Uncertainty Propagation in Probabilistic Design of Multilevel Systems

Techniques for Estimating Uncertainty Propagation in Probabilistic Design of Multilevel Systems 0th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conerence 0 August - September 004, Albany, New Yor AIAA 004-4470 Techniques or Estimating Uncertainty Propagation in Probabilistic Design o Multilevel

More information

A Structural Reliability Analysis Method Based on Radial Basis Function

A Structural Reliability Analysis Method Based on Radial Basis Function Copyright 2012 Tech Science Press CMC, vol.27, no.2, pp.128-142, 2012 A Structural Reliability Analysis Method Based on Radial Basis Function M. Q. Chau 1,2, X. Han 1, Y. C. Bai 1 and C. Jiang 1 Abstract:

More information

Reliability-based design optimization of problems with correlated input variables using a Gaussian Copula

Reliability-based design optimization of problems with correlated input variables using a Gaussian Copula Struct Multidisc Optim DOI 0.007/s0058-008-077-9 RESEARCH PAPER Reliability-based design optimization of problems with correlated input variables using a Gaussian Copula Yoojeong Noh K. K. Choi Liu Du

More information

System Reliability Analysis Using Tail Modeling

System Reliability Analysis Using Tail Modeling System Reliability Analysis Using Tail Modeling Palaniappn Ramu 1, Nam H. Kim 2 and Raphael T. Haftka 3 University of Florida, Gainesville, Florida, 32611 and Nestor V. Queipo 4 University of Zulia, Maracaibo,

More information

Safety Envelope for Load Tolerance and Its Application to Fatigue Reliability Design

Safety Envelope for Load Tolerance and Its Application to Fatigue Reliability Design Safety Envelope for Load Tolerance and Its Application to Fatigue Reliability Design Haoyu Wang * and Nam H. Kim University of Florida, Gainesville, FL 32611 Yoon-Jun Kim Caterpillar Inc., Peoria, IL 61656

More information

THIRD-MOMENT STANDARDIZATION FOR STRUCTURAL RELIABILITY ANALYSIS

THIRD-MOMENT STANDARDIZATION FOR STRUCTURAL RELIABILITY ANALYSIS THIRD-MOMENT STANDARDIZATION FOR STRUCTURAL RELIABILITY ANALYSIS By Yan-Gang Zhao and Tetsuro Ono ABSTRACT: First- and second-order reliability methods are generally considered to be among the most useful

More information

Technical Briefs. 1 Introduction. 876 / Vol. 129, AUGUST 2007 Copyright 2007 by ASME Transactions of the ASME

Technical Briefs. 1 Introduction. 876 / Vol. 129, AUGUST 2007 Copyright 2007 by ASME Transactions of the ASME Journal of Mechanical Design Technical Briefs Integration of Possibility-Based Optimization and Robust Design for Epistemic Uncertainty Byeng D. Youn 1 Assistant Professor Department of Mechanical Engineering

More information

Towards a Better Understanding of Modeling Feasibility Robustness in Engineering Design

Towards a Better Understanding of Modeling Feasibility Robustness in Engineering Design Xiaoping Du Research Associate Wei Chen* Assistant Professor Integrated Design Automation Laboratory (IDAL), Department of Mechanical Engineering, University of Illinois at Chicago, Chicago, IL 60607-7022

More information

However, reliability analysis is not limited to calculation of the probability of failure.

However, reliability analysis is not limited to calculation of the probability of failure. Probabilistic Analysis probabilistic analysis methods, including the first and second-order reliability methods, Monte Carlo simulation, Importance sampling, Latin Hypercube sampling, and stochastic expansions

More information

A Moving Kriging Interpolation Response Surface Method for Structural Reliability Analysis

A Moving Kriging Interpolation Response Surface Method for Structural Reliability Analysis Copyright 2013 Tech Science Press CMES, vol.93, no.6, pp.469-488, 2013 A Moving Kriging Interpolation Response Surface Method for Structural Reliability Analysis W. Zhao 1,2, J.K. Liu 3, X.Y. Li 2, Q.W.

More information

A single loop reliability-based design optimization using EPM and MPP-based PSO

A single loop reliability-based design optimization using EPM and MPP-based PSO 826 A single loop reliability-based design optimization using EPM and MPP-based PSO Abstract A reliability-based design optimization (RBDO) incorporates a probabilistic analysis with an optimization technique

More information

Assessment of Probabilistic Methods for Mistuned Bladed Disk Vibration

Assessment of Probabilistic Methods for Mistuned Bladed Disk Vibration 46th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics & Materials Conference 18-21 April 2005, Austin, Texas AIAA 2005-1990 Assessment of Probabilistic Methods for Mistuned Bladed Disk Vibration

More information

Reduction of Random Variables in Structural Reliability Analysis

Reduction of Random Variables in Structural Reliability Analysis Reduction of Random Variables in Structural Reliability Analysis S. Adhikari and R. S. Langley Department of Engineering University of Cambridge Trumpington Street Cambridge CB2 1PZ (U.K.) February 21,

More information

Uncertainty Quantification in Remaining Useful Life Prediction using First-Order Reliability Methods

Uncertainty Quantification in Remaining Useful Life Prediction using First-Order Reliability Methods ACCEPTED FOR PUBLICATION IN IEEE TRANSACTIONS ON RELIABILITY 1 Uncertainty Quantification in Remaining Useful Life Prediction using First-Order Reliability Methods Shankar Sankararaman*, Member, IEEE,

More information

A Saddlepoint Approximation Based Simulation Method for Uncertainty Analysis

A Saddlepoint Approximation Based Simulation Method for Uncertainty Analysis International Journal of Reliability and Safety Volume 1, Issue 1-2 DOI: 10.1504/IJRS.2006.010698 A Saddlepoint Approximation Based Simulation Method for Uncertainty Analysis Beiqing Huang Graduate Research

More information

Sensitivity and Reliability Analysis of Nonlinear Frame Structures

Sensitivity and Reliability Analysis of Nonlinear Frame Structures Sensitivity and Reliability Analysis of Nonlinear Frame Structures Michael H. Scott Associate Professor School of Civil and Construction Engineering Applied Mathematics and Computation Seminar April 8,

More information

Methods of Reliability Analysis in the context of RDO. Lectures. Christian Bucher

Methods of Reliability Analysis in the context of RDO. Lectures. Christian Bucher Lectures Methods of Reliability Analysis in the context of RDO Christian Bucher presented at the Weimar Optimization and Stochastic Days 2011 Source: www.dynardo.de/en/library Methods of Reliability Analysis

More information

IN MANY reliability-based design optimization (RBDO)

IN MANY reliability-based design optimization (RBDO) AIAA JOURNAL Vol 47, No 4, April 009 Reduction of Ordering Effect in Reliability-Based Design Optimization Using Dimension Reduction Method Yoojeong Noh, K K Choi, and Ikjin Lee University of Iowa, Iowa

More information

Optimal Multilevel System Design under Uncertainty

Optimal Multilevel System Design under Uncertainty Optimal Multilevel System Design under Uncertainty 1 M. Kokkolaras (mk@umich.edu) Department of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan Z.P. Mourelatos (mourelat@oakland.edu)

More information

Model Calibration under Uncertainty: Matching Distribution Information

Model Calibration under Uncertainty: Matching Distribution Information Model Calibration under Uncertainty: Matching Distribution Information Laura P. Swiler, Brian M. Adams, and Michael S. Eldred September 11, 008 AIAA Multidisciplinary Analysis and Optimization Conference

More information

Stochastic optimization - how to improve computational efficiency?

Stochastic optimization - how to improve computational efficiency? Stochastic optimization - how to improve computational efficiency? Christian Bucher Center of Mechanics and Structural Dynamics Vienna University of Technology & DYNARDO GmbH, Vienna Presentation at Czech

More information

Reliability Based Design Optimization of Systems with. Dynamic Failure Probabilities of Components. Arun Bala Subramaniyan

Reliability Based Design Optimization of Systems with. Dynamic Failure Probabilities of Components. Arun Bala Subramaniyan Reliability Based Design Optimization of Systems with Dynamic Failure Probabilities of Components by Arun Bala Subramaniyan A Thesis Presented in Partial Fulfillment of the Requirements for the Degree

More information

Reliability Analysis Methods

Reliability Analysis Methods Reliability Analysis Methods Emilio Bastidas-Arteaga, Abdel-Hamid Soubra To cite this version: Emilio Bastidas-Arteaga, Abdel-Hamid Soubra. Reliability Analysis Methods. Michael A. Hicks; Cristina Jommi.

More information

Reliability-based Design Optimization of Nonlinear Energy Sinks

Reliability-based Design Optimization of Nonlinear Energy Sinks th World Congress on Structural and Multidisciplinary Optimization 7 th - th, June, Sydney Australia Reliability-based Design Optimization of Nonlinear Energy Sinks Ethan Boroson and Samy Missoum University

More information

EFFICIENT MODELS FOR WIND TURBINE EXTREME LOADS USING INVERSE RELIABILITY

EFFICIENT MODELS FOR WIND TURBINE EXTREME LOADS USING INVERSE RELIABILITY Published in Proceedings of the L00 (Response of Structures to Extreme Loading) Conference, Toronto, August 00. EFFICIENT MODELS FOR WIND TURBINE ETREME LOADS USING INVERSE RELIABILITY K. Saranyasoontorn

More information

A Robust Design Method Using Variable Transformation and Gauss-Hermite Integration

A Robust Design Method Using Variable Transformation and Gauss-Hermite Integration International Journal for Numerical Methods in Engineering, 66(), pp. 84 858 A Robust Design Method Using Variable Transformation and Gauss-Hermite Integration Beiqing Huang Graduate Research Assistant,

More information

Fourth-Moment Standardization for Structural Reliability Assessment

Fourth-Moment Standardization for Structural Reliability Assessment Fourth-Moment Standardization for Structural Reliability Assessment Yan-Gang Zhao, M.ASCE 1 ; and Zhao-Hui Lu Abstract: In structural reliability analysis, the uncertainties related to resistance and load

More information

Optimization Methods

Optimization Methods Optimization Methods Decision making Examples: determining which ingredients and in what quantities to add to a mixture being made so that it will meet specifications on its composition allocating available

More information

RELIABILITY ANALYSIS AND DESIGN CONSIDERING DISJOINT ACTIVE FAILURE REGIONS

RELIABILITY ANALYSIS AND DESIGN CONSIDERING DISJOINT ACTIVE FAILURE REGIONS RELIABILITY ANALYSIS AND DESIGN CONSIDERING DISJOINT ACTIVE FAILURE REGIONS A Thesis by Xiaolong Cui Master of Science, Wichita State University, 2016 Bachelor of Science, Wichita State University, 2013

More information

PRACTICAL FIRST-ORDER RELIABILITY COMPUTATIONS USING SPREADSHEET

PRACTICAL FIRST-ORDER RELIABILITY COMPUTATIONS USING SPREADSHEET PRACTICAL FIRST-ORDER RELIABILITY COMPUTATIONS USING SPREADSHEET B. K. Low Associate Professor Geotechnical Research Centre School of Civil & Environ. Engineering Nanyang Technological University Republic

More information

RESPONSE SURFACE METHODS FOR STOCHASTIC STRUCTURAL OPTIMIZATION

RESPONSE SURFACE METHODS FOR STOCHASTIC STRUCTURAL OPTIMIZATION Meccanica dei Materiali e delle Strutture Vol. VI (2016), no.1, pp. 99-106 ISSN: 2035-679X Dipartimento di Ingegneria Civile, Ambientale, Aerospaziale, Dei Materiali DICAM RESPONSE SURFACE METHODS FOR

More information

CHAPTER 2: QUADRATIC PROGRAMMING

CHAPTER 2: QUADRATIC PROGRAMMING CHAPTER 2: QUADRATIC PROGRAMMING Overview Quadratic programming (QP) problems are characterized by objective functions that are quadratic in the design variables, and linear constraints. In this sense,

More information

Nonlinear Optimization: What s important?

Nonlinear Optimization: What s important? Nonlinear Optimization: What s important? Julian Hall 10th May 2012 Convexity: convex problems A local minimizer is a global minimizer A solution of f (x) = 0 (stationary point) is a minimizer A global

More information

Reliability assessment of cutting tools life based on advanced approximation methods

Reliability assessment of cutting tools life based on advanced approximation methods Reliability assessment of cutting tools life based on advanced approximation methods K. Salonitis 1*, A. Kolios 2 1 Cranfield University, Manufacturing and Materials Department 2 Cranfield University,

More information

Riccati difference equations to non linear extended Kalman filter constraints

Riccati difference equations to non linear extended Kalman filter constraints International Journal of Scientific & Engineering Research Volume 3, Issue 12, December-2012 1 Riccati difference equations to non linear extended Kalman filter constraints Abstract Elizabeth.S 1 & Jothilakshmi.R

More information

Dynamic System Identification using HDMR-Bayesian Technique

Dynamic System Identification using HDMR-Bayesian Technique Dynamic System Identification using HDMR-Bayesian Technique *Shereena O A 1) and Dr. B N Rao 2) 1), 2) Department of Civil Engineering, IIT Madras, Chennai 600036, Tamil Nadu, India 1) ce14d020@smail.iitm.ac.in

More information

component risk analysis

component risk analysis 273: Urban Systems Modeling Lec. 3 component risk analysis instructor: Matteo Pozzi 273: Urban Systems Modeling Lec. 3 component reliability outline risk analysis for components uncertain demand and uncertain

More information

Reliability analysis of geotechnical risks

Reliability analysis of geotechnical risks Reliability analysis of geotechnical risks Lazhar Belabed*, Hacene Benyaghla* * Department of Civil Engineering and Hydraulics, University of Guelma, Algeria Abstract The evaluation of safety or reliability

More information

Reliability-Based Microstructural Topology Design with Respect to Vibro-Acoustic Criteria

Reliability-Based Microstructural Topology Design with Respect to Vibro-Acoustic Criteria 11 th World Congress on Structural and Multidisciplinary Optimisation 07 th -12 th, June 2015, Sydney Australia Reliability-Based Microstructural Topology Design with Respect to Vibro-Acoustic Criteria

More information

AN ADAPTIVE RESPONSE SURFACE METHOD UTILIZING ERROR ESTIMATES

AN ADAPTIVE RESPONSE SURFACE METHOD UTILIZING ERROR ESTIMATES 8 th ASCE Specialty Conference on Probabilistic Mechanics and Structural Reliability PMC2000-068 Abstract AN ADAPTIVE RESPONSE SURFACE METHOD UTILIZING ERROR ESTIMATES Michael Macke, Dirk Roos and Jörg

More information

Lecture XI. Approximating the Invariant Distribution

Lecture XI. Approximating the Invariant Distribution Lecture XI Approximating the Invariant Distribution Gianluca Violante New York University Quantitative Macroeconomics G. Violante, Invariant Distribution p. 1 /24 SS Equilibrium in the Aiyagari model G.

More information

Application of bootstrap method in conservative estimation of reliability with limited samples

Application of bootstrap method in conservative estimation of reliability with limited samples Struct Multidisc Optim (21) 41:25 217 DOI 1.17/s158-9-419-8 RESEARCH PAPER Application of bootstrap method in conservative estimation of reliability with limited samples Victor Picheny Nam Ho Kim Raphael

More information

Reliability sensitivity analysis using axis orthogonal importance Latin hypercube sampling method

Reliability sensitivity analysis using axis orthogonal importance Latin hypercube sampling method Reliability Analysis and Design Optimization of Mechanical Systems under Various Uncertainties - Research Article Reliability sensitivity analysis using axis orthogonal importance Latin hypercube sampling

More information

Overview of Structural Reliability Analysis Methods Part I: Local Reliability Methods

Overview of Structural Reliability Analysis Methods Part I: Local Reliability Methods Overview of Structural Reliability Analysis Methods Part I: Local Reliability Methods ChangWu HUANG, *, Abdelkhalak El Hami, Bouchaïb Radi Normandie Université, INSA Rouen, LOFIMS, 76000 Rouen, France.

More information

Structural reliability analysis with implicit limit state functions

Structural reliability analysis with implicit limit state functions J. Miranda Structural reliability analysis with implicit limit state functions 1 Structural reliability analysis with implicit limit state functions Jorge Miranda Instituto Superior Técnico, University

More information

Computational Intelligence Winter Term 2017/18

Computational Intelligence Winter Term 2017/18 Computational Intelligence Winter Term 207/8 Prof. Dr. Günter Rudolph Lehrstuhl für Algorithm Engineering (LS ) Fakultät für Informatik TU Dortmund Plan for Today Single-Layer Perceptron Accelerated Learning

More information

Guideline for Offshore Structural Reliability Analysis - General 3. RELIABILITY ANALYSIS 38

Guideline for Offshore Structural Reliability Analysis - General 3. RELIABILITY ANALYSIS 38 FEBRUARY 20, 1995 3. RELIABILITY ANALYSIS 38 3.1 General 38 3.1.1 Variables 38 3.1.2 Events 39 3.1.3 Event Probability 41 3.1.4 The Reliability Index 41 3.1.5 The Design Point 42 3.1.6 Transformation of

More information

While using the input and output data fu(t)g and fy(t)g, by the methods in system identification, we can get a black-box model like (In the case where

While using the input and output data fu(t)g and fy(t)g, by the methods in system identification, we can get a black-box model like (In the case where ESTIMATE PHYSICAL PARAMETERS BY BLACK-BOX MODELING Liang-Liang Xie Λ;1 and Lennart Ljung ΛΛ Λ Institute of Systems Science, Chinese Academy of Sciences, 100080, Beijing, China ΛΛ Department of Electrical

More information

Structural Analysis of Large Caliber Hybrid Ceramic/Steel Gun Barrels

Structural Analysis of Large Caliber Hybrid Ceramic/Steel Gun Barrels Structural Analysis of Large Caliber Hybrid Ceramic/Steel Gun Barrels MS Thesis Jon DeLong Department of Mechanical Engineering Clemson University OUTLINE Merger of ceramics into the conventional steel

More information

Probabilistic analysis of off-center cracks in cylindrical structures

Probabilistic analysis of off-center cracks in cylindrical structures International Journal of Pressure Vessels and Piping 77 (2000) 3 16 www.elsevier.com/locate/ijpvp Probabilistic analysis of off-center cracks in cylindrical structures S. Rahman*, G. Chen a, R. Firmature

More information

Recent Advances in Reliability Estimation of Time-Dependent Problems Using the Concept of Composite Limit State

Recent Advances in Reliability Estimation of Time-Dependent Problems Using the Concept of Composite Limit State Automotive Research Center A U.S. Army Center of Excellence for Modeling and Simulation of Ground Vehicles Recent Advances in Reliability Estimation of Time-Dependent Problems Using the Concept of Composite

More information

Reliability Estimation of Complex Numerical Problems Using Modified Conditional Expectation Method

Reliability Estimation of Complex Numerical Problems Using Modified Conditional Expectation Method Wayne State University Civil and Environmental Engineering Faculty Research Publications Civil and Environmental Engineering 10-8-2010 Reliability Estimation of Complex Numerical Problems Using Modified

More information

Fig. 1: Example of Scallops

Fig. 1: Example of Scallops UNCERTAINTY ANALYSIS IN LASER DEPOSITION FINISH MACHINING OPERATIONS ABSTRACT The Laser Aided Manufacturing Process (LAMP) from Missouri S&T is a laser based metals rapid manufacturing process that uses

More information

Computational Intelligence

Computational Intelligence Plan for Today Single-Layer Perceptron Computational Intelligence Winter Term 00/ Prof. Dr. Günter Rudolph Lehrstuhl für Algorithm Engineering (LS ) Fakultät für Informatik TU Dortmund Accelerated Learning

More information

ENHANCING WEIGHTED UNIFORM SIMULATION FOR STRUCTURAL RELIABILITY ANALYSIS

ENHANCING WEIGHTED UNIFORM SIMULATION FOR STRUCTURAL RELIABILITY ANALYSIS INTERNATIONAL JOURNAL OF OPTIMIZATION IN CIVIL ENGINEERING Int. J. Optim. Civil Eng., 2013; 3(4):635-651 ENHANCING WEIGHTED UNIFORM SIMULATION FOR STRUCTURAL RELIABILITY ANALYSIS H. Ghohani Arab, M.R.

More information

Monotonicity Analysis, Evolutionary Multi-Objective Optimization, and Discovery of Design Principles

Monotonicity Analysis, Evolutionary Multi-Objective Optimization, and Discovery of Design Principles Monotonicity Analysis, Evolutionary Multi-Objective Optimization, and Discovery of Design Principles Kalyanmoy Deb and Aravind Srinivasan Kanpur Genetic Algorithms Laboratory (KanGAL) Indian Institute

More information

Study on Optimal Design of Automotive Body Structure Crashworthiness

Study on Optimal Design of Automotive Body Structure Crashworthiness 7 th International LS-DYNA Users Conference Simulation Technology () Study on Optimal Design of Automotive Body Structure Crashworthiness Wang Hailiang Lin Zhongqin Jin Xianlong Institute for Automotive

More information

Gradient Descent. Dr. Xiaowei Huang

Gradient Descent. Dr. Xiaowei Huang Gradient Descent Dr. Xiaowei Huang https://cgi.csc.liv.ac.uk/~xiaowei/ Up to now, Three machine learning algorithms: decision tree learning k-nn linear regression only optimization objectives are discussed,

More information

Auxiliary signal design for failure detection in uncertain systems

Auxiliary signal design for failure detection in uncertain systems Auxiliary signal design for failure detection in uncertain systems R. Nikoukhah, S. L. Campbell and F. Delebecque Abstract An auxiliary signal is an input signal that enhances the identifiability of a

More information

Robust and Reliability Based Design Optimization

Robust and Reliability Based Design Optimization Robust and Reliability Based Design Optimization Frederico Afonso Luís Amândio André Marta Afzal Suleman CCTAE, IDMEC, LAETA Instituto Superior Técnico Univerdade de Lisboa Lisboa, Portugal May 22, 2015

More information

Finite Element Structural Analysis using Imprecise Probabilities Based on P-Box Representation

Finite Element Structural Analysis using Imprecise Probabilities Based on P-Box Representation Finite Element Structural Analysis using Imprecise Probabilities Based on P-Box Representation Hao Zhang 1, R. L. Mullen 2, and R. L. Muhanna 3 1 University of Sydney, Sydney 2006, Australia, haozhang@usyd.edu.au

More information

Risk Assessment of Highway Bridges: A Reliability-based Approach

Risk Assessment of Highway Bridges: A Reliability-based Approach Risk Assessment of Highway Bridges: A Reliability-based Approach by Reynaldo M. Jr., PhD Indiana University-Purdue University Fort Wayne pablor@ipfw.edu Abstract: Many countries are currently experiencing

More information

RELIABILITY MODELING OF IMPACTED COMPOSITE MATERIALS FOR RAILWAYS

RELIABILITY MODELING OF IMPACTED COMPOSITE MATERIALS FOR RAILWAYS 16 TH INTERNATIONAL CONFERENCE ON COMPOSITE MATERIALS RELIABILITY MODELING OF IMPACTED COMPOSITE MATERIALS FOR RAILWAYS Guillaumat L*, Dau F*, Cocheteux F**, Chauvin T** *LAMEFIP ENSAM Esplanade des Arts

More information

Model Bias Characterization in the Design Space under Uncertainty

Model Bias Characterization in the Design Space under Uncertainty International Journal of Performability Engineering, Vol. 9, No. 4, July, 03, pp.433-444. RAMS Consultants Printed in India Model Bias Characterization in the Design Space under Uncertainty ZHIMIN XI *,

More information

A probabilistic method to predict fatigue crack initiation

A probabilistic method to predict fatigue crack initiation International Journal of Fracture (2006) 137:9 17 DOI 10.1007/s10704-005-3074-0 Springer 2006 A probabilistic method to predict fatigue crack initiation SALIL. S. KULKARNI, L. SUN, B. MORAN, S. KRISHNASWAMY

More information

Use of Simulation in Structural Reliability

Use of Simulation in Structural Reliability Structures 008: Crossing Borders 008 ASCE Use of Simulation in Structural Reliability Author: abio Biondini, Department of Structural Engineering, Politecnico di Milano, P.za L. Da Vinci 3, 033 Milan,

More information

Modelling Under Risk and Uncertainty

Modelling Under Risk and Uncertainty Modelling Under Risk and Uncertainty An Introduction to Statistical, Phenomenological and Computational Methods Etienne de Rocquigny Ecole Centrale Paris, Universite Paris-Saclay, France WILEY A John Wiley

More information

A Novel and Precise Sixth-Order Method for Solving Nonlinear Equations

A Novel and Precise Sixth-Order Method for Solving Nonlinear Equations A Novel and Precise Sixth-Order Method for Solving Nonlinear Equations F. Soleymani Department of Mathematics, Islamic Azad University, Zahedan Branch, Zahedan, Iran E-mail: fazl_soley_bsb@yahoo.com; Tel:

More information

Optimal Polynomial Control for Discrete-Time Systems

Optimal Polynomial Control for Discrete-Time Systems 1 Optimal Polynomial Control for Discrete-Time Systems Prof Guy Beale Electrical and Computer Engineering Department George Mason University Fairfax, Virginia Correspondence concerning this paper should

More information

COUPLED SYSTEMS DESIGN IN PROBABILISTIC ENVIRONMENTS

COUPLED SYSTEMS DESIGN IN PROBABILISTIC ENVIRONMENTS COUPLED SYSTEMS DESIGN IN PROBABILISTIC ENVIRONMENTS Tom Halecki NASA Graduate Researcher, Student Member AIAA Department of Mechanical and Aerospace Engineering University at Buffalo, SUNY Buffalo, NY

More information

Local Approximation of the Efficient Frontier in Robust Design

Local Approximation of the Efficient Frontier in Robust Design Local Approximation of the Efficient Frontier in Robust Design Jinhuan Zhang, Graduate Assistant Department of Mechanical Engineering Clemson University Margaret M. Wiecek, Associate Professor Department

More information

YIELD curves represent the relationship between market

YIELD curves represent the relationship between market Smoothing Yield Curve Data by Least Squares and Concavity or Convexity I. P. Kandylas and I. C. Demetriou Abstract A yield curve represents the relationship between market interest rates and time to maturity

More information

Full terms and conditions of use:

Full terms and conditions of use: This article was downloaded by:[rollins, Derrick] [Rollins, Derrick] On: 26 March 2007 Access Details: [subscription number 770393152] Publisher: Taylor & Francis Informa Ltd Registered in England and

More information

CHAPTER 5 ROBUSTNESS ANALYSIS OF THE CONTROLLER

CHAPTER 5 ROBUSTNESS ANALYSIS OF THE CONTROLLER 114 CHAPTER 5 ROBUSTNESS ANALYSIS OF THE CONTROLLER 5.1 INTRODUCTION Robust control is a branch of control theory that explicitly deals with uncertainty in its approach to controller design. It also refers

More information

Adaptive Optimal Sampling Methodology for Reliability Prediction of Series Systems

Adaptive Optimal Sampling Methodology for Reliability Prediction of Series Systems AIAA JOURNAL Vol. 44, No. 3, March 2006 Adaptive Optimal Sampling Methodology for Reliability Prediction of Series Systems Michael P. Enright Southwest Research Institute, San Antonio, Texas 78238 Harry

More information

Uniform Random Number Generators

Uniform Random Number Generators JHU 553.633/433: Monte Carlo Methods J. C. Spall 25 September 2017 CHAPTER 2 RANDOM NUMBER GENERATION Motivation and criteria for generators Linear generators (e.g., linear congruential generators) Multiple

More information

NONLINEAR. (Hillier & Lieberman Introduction to Operations Research, 8 th edition)

NONLINEAR. (Hillier & Lieberman Introduction to Operations Research, 8 th edition) NONLINEAR PROGRAMMING (Hillier & Lieberman Introduction to Operations Research, 8 th edition) Nonlinear Programming g Linear programming has a fundamental role in OR. In linear programming all its functions

More information

Uncertainty quantification of world population growth: A self-similar PDF model

Uncertainty quantification of world population growth: A self-similar PDF model DOI 10.1515/mcma-2014-0005 Monte Carlo Methods Appl. 2014; 20 (4):261 277 Research Article Stefan Heinz Uncertainty quantification of world population growth: A self-similar PDF model Abstract: The uncertainty

More information

Applied Mathematics Letters. Combined bracketing methods for solving nonlinear equations

Applied Mathematics Letters. Combined bracketing methods for solving nonlinear equations Applied Mathematics Letters 5 (01) 1755 1760 Contents lists available at SciVerse ScienceDirect Applied Mathematics Letters journal homepage: www.elsevier.com/locate/aml Combined bracketing methods for

More information

Online publication date: 01 March 2010 PLEASE SCROLL DOWN FOR ARTICLE

Online publication date: 01 March 2010 PLEASE SCROLL DOWN FOR ARTICLE This article was downloaded by: [2007-2008-2009 Pohang University of Science and Technology (POSTECH)] On: 2 March 2010 Access details: Access Details: [subscription number 907486221] Publisher Taylor

More information

Investigation of Plate Structure Design under Stochastic Blast Loading

Investigation of Plate Structure Design under Stochastic Blast Loading 10 th World Congress on Structural and Multidisciplinary Optimization May 19 24, 2013, Orlando, Florida, USA Investigation of Plate Structure Design under Stochastic Blast Loading Joshua J. Israel, Andrés

More information

ProbReach: Probabilistic Bounded Reachability for Uncertain Hybrid Systems

ProbReach: Probabilistic Bounded Reachability for Uncertain Hybrid Systems ProbReach: Probabilistic Bounded Reachability for Uncertain Hybrid Systems Fedor Shmarov, Paolo Zuliani School of Computing Science, Newcastle University, UK 1 / 41 Introduction ProbReach tool for probabilistic

More information

Review of Classical Optimization

Review of Classical Optimization Part II Review of Classical Optimization Multidisciplinary Design Optimization of Aircrafts 51 2 Deterministic Methods 2.1 One-Dimensional Unconstrained Minimization 2.1.1 Motivation Most practical optimization

More information

Robust optimal design of a magnetizer to reduce the harmonic components of cogging torque in a HDD spindle motor

Robust optimal design of a magnetizer to reduce the harmonic components of cogging torque in a HDD spindle motor Microsyst Technol (2014) 20:1497 1504 DOI 10.1007/s00542-014-2153-4 Technical Paper Robust optimal design of a magnetizer to reduce the harmonic components of cogging torque in a HDD spindle motor Changjin

More information

A new nonmonotone Newton s modification for unconstrained Optimization

A new nonmonotone Newton s modification for unconstrained Optimization A new nonmonotone Newton s modification for unconstrained Optimization Aristotelis E. Kostopoulos a George S. Androulakis b a a Department of Mathematics, University of Patras, GR-265.04, Rio, Greece b

More information

Iterative Methods for Solving A x = b

Iterative Methods for Solving A x = b Iterative Methods for Solving A x = b A good (free) online source for iterative methods for solving A x = b is given in the description of a set of iterative solvers called templates found at netlib: http

More information

Design for Reliability and Robustness through Probabilistic Methods in COMSOL Multiphysics with OptiY

Design for Reliability and Robustness through Probabilistic Methods in COMSOL Multiphysics with OptiY Excerpt from the Proceedings of the COMSOL Conference 2008 Hannover Design for Reliability and Robustness through Probabilistic Methods in COMSOL Multiphysics with OptiY The-Quan Pham * 1, Holger Neubert

More information

Minimax MMSE Estimator for Sparse System

Minimax MMSE Estimator for Sparse System Proceedings of the World Congress on Engineering and Computer Science 22 Vol I WCE 22, October 24-26, 22, San Francisco, USA Minimax MMSE Estimator for Sparse System Hongqing Liu, Mandar Chitre Abstract

More information

A conditional simulation of non-normal velocity/pressure fields

A conditional simulation of non-normal velocity/pressure fields A conditional simulation of non-normal velocity/ fields K. R. urley, A. Kareem Department of Civil Engineering University of Florida, ainesville, FL, 36 Department of Civil Engineering and eological Sciences

More information

Optimization. Escuela de Ingeniería Informática de Oviedo. (Dpto. de Matemáticas-UniOvi) Numerical Computation Optimization 1 / 30

Optimization. Escuela de Ingeniería Informática de Oviedo. (Dpto. de Matemáticas-UniOvi) Numerical Computation Optimization 1 / 30 Optimization Escuela de Ingeniería Informática de Oviedo (Dpto. de Matemáticas-UniOvi) Numerical Computation Optimization 1 / 30 Unconstrained optimization Outline 1 Unconstrained optimization 2 Constrained

More information