Sensitivity and Reliability Analysis of Nonlinear Frame Structures

Similar documents
Structural Reliability

Nonlinear static analysis PUSHOVER

EDEM DISCRETIZATION (Phase II) Normal Direction Structure Idealization Tangential Direction Pore spring Contact spring SPRING TYPES Inner edge Inner d

A Krylov Subspace Accelerated Newton Algorithm

Fire Analysis of Reinforced Concrete Beams with 2-D Plane Stress Concrete Model

NUMERICAL SIMULATION OF THE INELASTIC SEISMIC RESPONSE OF RC STRUCTURES WITH ENERGY DISSIPATORS

Institute for Statics und Dynamics of Structures Fuzzy probabilistic safety assessment

Finite Element Modelling with Plastic Hinges

Course in. Geometric nonlinearity. Nonlinear FEM. Computational Mechanics, AAU, Esbjerg

Force-based Element vs. Displacement-based Element

Seismic Pushover Analysis Using AASHTO Guide Specifications for LRFD Seismic Bridge Design

International Journal of Scientific & Engineering Research Volume 9, Issue 3, March-2018 ISSN

Sabah Shawkat Cabinet of Structural Engineering Walls carrying vertical loads should be designed as columns. Basically walls are designed in

ENERGY DIAGRAM w/ HYSTERETIC

Finite Element Analysis Prof. Dr. B. N. Rao Department of Civil Engineering Indian Institute of Technology, Madras. Module - 01 Lecture - 13

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

SPREAD OF PLASTICITY ANALYSIS OF R/C BUILDINGS, SUBJECTED TO MONOTONIC SEISMIC LOADING

Improved refined plastic hinge analysis accounting for strain reversal

VORONOI APPLIED ELEMENT METHOD FOR STRUCTURAL ANALYSIS: THEORY AND APPLICATION FOR LINEAR AND NON-LINEAR MATERIALS

University of Sheffield The development of finite elements for 3D structural analysis in fire

EQUIVALENT FRACTURE ENERGY CONCEPT FOR DYNAMIC RESPONSE ANALYSIS OF PROTOTYPE RC GIRDERS

A Constant Displacement Iteration Algorithm for Nonlinear Static Push-Over Analyses

EFFECTS OF CONFINED CONCRETE MODELS ON SIMULATING RC COLUMNS UNDER LOW-CYCLIC LOADING

Reduction of Random Variables in Structural Reliability Analysis

Structural Analysis of Large Caliber Hybrid Ceramic/Steel Gun Barrels

NUMERICAL SIMULATION OF STEEL CATENARY RISER

ENGN2340 Final Project: Implementation of a Euler-Bernuolli Beam Element Michael Monn

Stress analysis of a stepped bar

INFLUENCE OF FLANGE STIFFNESS ON DUCTILITY BEHAVIOUR OF PLATE GIRDER

INELASTIC SEISMIC DISPLACEMENT RESPONSE PREDICTION OF MDOF SYSTEMS BY EQUIVALENT LINEARIZATION

ESTIMATING PARK-ANG DAMAGE INDEX USING EQUIVALENT SYSTEMS

Reliability implications of advanced analysis in design of steel frames

Mechanical Engineering Ph.D. Preliminary Qualifying Examination Solid Mechanics February 25, 2002

MODELLING DAMAGE AND PROGRESSIVE COLLAPSE OF FRAMES USING A GAUSSIAN SPRINGS BASED APPLIED ELEMENT METHOD

EUROCODE EN SEISMIC DESIGN OF BRIDGES

March 24, Chapter 4. Deflection and Stiffness. Dr. Mohammad Suliman Abuhaiba, PE

Moment redistribution of continuous composite I-girders with high strength steel

Uncertainty Quantification in MEMS

The University of Melbourne Engineering Mechanics

NORMAL STRESS. The simplest form of stress is normal stress/direct stress, which is the stress perpendicular to the surface on which it acts.

. D CR Nomenclature D 1

Reduction of Random Variables in Structural Reliability Analysis

MECHANICS OF MATERIALS. Prepared by Engr. John Paul Timola

Comparison of Structural Models for Seismic Analysis of Multi-Storey Frame Buildings

component risk analysis

New Developments in Tail-Equivalent Linearization method for Nonlinear Stochastic Dynamics

Biomechanics. Soft Tissue Biomechanics

EARTHQUAKE SIMULATION TESTS OF BRIDGE COLUMN MODELS DAMAGED DURING 1995 KOBE EARTHQUAKE

A Modified Response Spectrum Analysis Procedure (MRSA) to Determine the Nonlinear Seismic Demands of Tall Buildings

Seismic Assessment of a RC Building according to FEMA 356 and Eurocode 8

A probabilistic method to predict fatigue crack initiation

CHAPTER 14 BUCKLING ANALYSIS OF 1D AND 2D STRUCTURES

Civil Engineering Design (1) Design of Reinforced Concrete Columns 2006/7

Uncertainty modelling using software FReET

Method of Finite Elements II Modeling ohysteresis

PREDICTION OF THE CYCLIC BEHAVIOR OF MOMENT RESISTANT BEAM-TO-COLUMN JOINTS OF COMPOSITE STRUCTURAL ELEMENTS

Large Thermal Deflections of a Simple Supported Beam with Temperature-Dependent Physical Properties

STATICALLY INDETERMINATE STRUCTURES

Design issues of thermal induced effects and temperature dependent material properties in Abaqus

Pushover Seismic Analysis of Bridge Structures

999 TOWN & COUNTRY ROAD ORANGE, CALIFORNIA TITLE PUSHOVER ANALYSIS EXAMPLE BY R. MATTHEWS DATE 5/21/01

Introduction to Engineering Materials ENGR2000. Dr. Coates

Non-Linear Modeling of Reinforced Concrete Structures for Seismic Applications

Slenderness Effects for Concrete Columns in Sway Frame - Moment Magnification Method (CSA A )

POST-PEAK BEHAVIOR OF FRP-JACKETED REINFORCED CONCRETE COLUMNS

Title. Author(s)DONG, Q.; OKAZAKI, T.; MIDORIKAWA, M.; RYAN, K.; SAT. Issue Date Doc URL. Type. Note. File Information BEARINGS

Unit 18 Other Issues In Buckling/Structural Instability

TORSION INCLUDING WARPING OF OPEN SECTIONS (I, C, Z, T AND L SHAPES)

Soil-Structure Interaction in Nonlinear Pushover Analysis of Frame RC Structures: Nonhomogeneous Soil Condition

Appendix G Analytical Studies of Columns

A numerical parametric study on the plate inclination angle of steel plate shear walls

Advanced aerostatic analysis of long-span suspension bridges *

ENG1001 Engineering Design 1

A NEW SIMPLIFIED AND EFFICIENT TECHNIQUE FOR FRACTURE BEHAVIOR ANALYSIS OF CONCRETE STRUCTURES

Model Calibration under Uncertainty: Matching Distribution Information

FINITE ELEMENT ANALYSIS OF TAPERED COMPOSITE PLATE GIRDER WITH A NON-LINEAR VARYING WEB DEPTH

Junya Yazawa 1 Seiya Shimada 2 and Takumi Ito 3 ABSTRACT 1. INTRODUCTION

Chapter Objectives. Design a beam to resist both bendingand shear loads

System reliability assessment with nonlinear finite element analysis

Reliable Condition Assessment of Structures Using Uncertain or Limited Field Modal Data

E1. Column subjected to bending and axial action, Non-dissipative material

SIMPLIFIED METHOD FOR PREDICTING DEFORMATIONS OF RC FRAMES DURING FIRE EXPOSURE

Influence of residual stresses in the structural behavior of. tubular columns and arches. Nuno Rocha Cima Gomes

Aim of the study Experimental determination of mechanical parameters Local buckling (wrinkling) Failure maps Optimization of sandwich panels

Centrifuge Shaking Table Tests and FEM Analyses of RC Pile Foundation and Underground Structure

Experimental Study and Numerical Simulation on Steel Plate Girders With Deep Section

and F NAME: ME rd Sample Final Exam PROBLEM 1 (25 points) Prob. 1 questions are all or nothing. PROBLEM 1A. (5 points)

Methods of Analysis. Force or Flexibility Method

ALGORITHM FOR NON-PROPORTIONAL LOADING IN SEQUENTIALLY LINEAR ANALYSIS

Post Graduate Diploma in Mechanical Engineering Computational mechanics using finite element method

Pseudo-Force Incremental Methods

Lecture-08 Gravity Load Analysis of RC Structures

S. Freitag, B. T. Cao, J. Ninić & G. Meschke

SEISMIC RELIABILITY ANALYSIS OF BASE-ISOLATED BUILDINGS

Design optimization of multi-point constraints in structures

Lecture Slides. Chapter 4. Deflection and Stiffness. The McGraw-Hill Companies 2012

MODELING OF NONLINEAR BEHAVIOR OF RC SHEAR WALLS UNDER COMBINED AXIAL, SHEAR AND FLEXURAL LOADING

Use of Simulation in Structural Reliability

STEEL JOINTS - COMPONENT METHOD APPLICATION

SEMM Mechanics PhD Preliminary Exam Spring Consider a two-dimensional rigid motion, whose displacement field is given by

Transcription:

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, 2011 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 1 / 34

Outline 1 Structural Response Sensitivity Analysis Governing Equilibrium Equations Computing Response Sensitivity Sensitivity Verification 2 Structural Reliability Analysis Problem Statement Computing Reliability Sensitivity at Design Point Cantilever Example Reinforced Concrete Frame Example M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 2 / 34

Sensitivity Overview Governing Equilibrium Equations Structural Equilibrium Balance of external loads and internal resisting forces P f (t) = P r (U f (t)) (1) P f (t) time-varying external loads (given) U f (t) time-varying nodal displacements (unknown) P r resisting forces (by selection and assembly) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 3 / 34

Sensitivity Overview Governing Equilibrium Equations Residual Equilibrium Equations Solve residual equilibrium using Newton-Raphson algorithm where R f (U f (t)) = P f (t) P r (U f (t)) = 0 (2) U f = ( ) 1 K j T Rf (U j f ) (3) K j T = P r U f (4) or any other suitable stiffness matrix according to other root finding algorithms (modified, quasi, accelerated Newton). Regardless, use increment to update displacement U j+1 f = U j f + U f (5) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 4 / 34

Sensitivity Overview Governing Equilibrium Equations Parameterized Equilibrium Equations Let θ be an uncertain parameter of the external load or the finite element model P f (t, θ) = P r (U f (t, θ), θ) (6) Nodal displacements, U f, depend on all θ, whether of the load or FE model Resisting forces depend explicitly on a θ of the FE model and implicitly on all θ via U f Load vector, P f, depends explicitly on a θ of the external load Typical properties θ represents: steel yield stress, concrete strength, member cross-section dimensions, nodal coordinates, etc. M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 5 / 34

Sensitivity Overview Computing Sensitivity Sensitivity to Uncertain Parameters Two approaches to determine U f / θ, the sensitivity of structural response to changes in θ 1 Finite Differences re-run analysis with perturbed parameter value, e.g., forward finite difference U f θ U f (θ + θ) U f (θ) θ 2 Direct Differentiation develop analytic expressions for sensitivity by differentiating governing equations of the finite element analysis (7) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 6 / 34

Sensitivity Overview Computing Sensitivity Direct Differentiation Method (DDM) Differentiate equilibrium equations wrt parameter, θ P f θ = P r U f U f θ + P r θ (8) Uf Solve for nodal displacement sensitivity, U f / θ ( U f θ = P f K 1 T θ P ) r θ Uf (9) At end of a simulation time step: Conditional derivative, P r / θ Uf, assembled from element contributions in same manner as P r Factorized K T from last Newton-Raphson iteration. Or reform and factorize K T if using Newton-like iteration. M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 7 / 34

Sensitivity Overview Computing Sensitivity Verification Compare with forward finite difference U f (θ + θ) U f (θ) lim = U f θ 0 θ θ (10) Repeat analysis with perturbed parameter If finite difference approximation converges to analytic sensitivity as θ decreases, then DDM is correct M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 8 / 34

Sensitivity Overview Sensitivity Verification Verification Example Slender cantilever with material and geometric nonlinearity Cross-Section 0.5 m Stress-Strain Bilinear 0.1 m 000000000000000000000 111111111111111111111 000000000000000000000 111111111111111111111 000000000000000000000 111111111111111111111 U(t) 5.0 m P(t) = P max sin(t) Discretize with five frame finite elements (linear curvature, constant axial deformation approximation) Corotational formulation for large displacements Numerically integrate stress-strain response over cross-section E = 2.0e8 kpa, σ y = 4.1e5 kpa, 2% kinematic strain-hardening One load cycle with P max = 1710 kn (5 times yield load) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 9 / 34

Sensitivity Overview Sensitivity Verification Load-Displacement Response 2000 Load, P (kn) 1000 0 1000 2000 4 2 0 2 4 Displacement, U (m) Material yield at about 400 kn load Tension stiffening at about 1 m displacement Elastic unloading and reverse cyclic yielding M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 10 / 34

Sensitivity Overview Sensitivity Verification Sensitivity to Yield Stress (θ σ y ) DDM verification and response envelope for strength parameter ( U/ θ)θ (m) 6 4 2 0 2 FFD, θ = 0.001θ DDM FFD Load, P (kn) 2000 1000 0 1000 U ± ( U/ θ)(0.1θ) 4 0 0.2 0.4 0.6 0.8 1 Time, t 2000 4 2 0 2 4 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 11 / 34

Sensitivity Overview Sensitivity Verification Sensitivity to Elastic Modulus (θ E) DDM verification and response envelope for stiffness parameter ( U/ θ)θ (m) 0.3 0.2 0.1 0 0.1 FFD, θ = 0.001θ DDM FFD Load, P (kn) 2000 1000 0 1000 U ± ( U/ θ)(0.1θ) 0.2 0 0.2 0.4 0.6 0.8 1 Time, t 2000 4 2 0 2 4 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 12 / 34

Sensitivity Overview Sensitivity Verification Sensitivity to Cross-Section Depth (θ d) DDM verification and response envelope for local geometric parameter ( U/ θ)θ (m) 10 5 0 5 FFD, θ = 0.001θ DDM FFD Load, P (kn) 2000 1000 0 1000 U ± ( U/ θ)(0.1θ) 10 0 0.2 0.4 0.6 0.8 1 Time, t 2000 4 2 0 2 4 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 13 / 34

Sensitivity Overview Sensitivity Verification Sensitivity to Cantilever Length (θ L) DDM verification and response envelope for global geometric parameter ( U/ θ)θ (m) 10 5 0 5 FFD, θ = 0.001θ DDM FFD Load, P (kn) 2000 1000 0 1000 U ± ( U/ θ)(0.1θ) 10 0 0.2 0.4 0.6 0.8 1 Time, t 2000 5 0 5 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 14 / 34

Problem Statement Structural Reliability Analysis Assess uncertain structural response in terms of probability of failure under prescribed events, or limit states Define scalar valued limit state function in terms of a vector of random variables, x g(x) = { 0 fail > 0 safe (11) Each random variable in x Has a prescribed PDF, CDF, mean, st.dev, etc.; Generally, correlated and non-normal; and Maps to one or more parameters, θ, of the load or FE model M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 15 / 34

Problem Statement Probability of Failure The probability that the limit state function evaluates to fail over all possible realizations of the random variables in x p f = f X (x)dx (12) g(x) 0 f X (x) is the joint PDF of all random variables Integrate over failure domain g(x) = 0 x n Contours of f X (x) SAFE FAIL x 1 g(x) = 0 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 16 / 34

Problem Statement Monte Carlo Simulation Generate N random realizations of x using joint inverse CDF N i=1 p f (g(x(i) ) 0) N Captures nonlinearities of limit state function Computationally intense: N >= 1e5 for accurate p f Difficult to obtain sensitivity of failure to random variables (13) x n g(x) = 0 SAFE FAIL x 1 g(x) = 0 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 17 / 34

Problem Statement First Order Approximation Linearize at design point, x, on failure surface closest to M x, mean value of x g(x) = 0 M x x x n SAFE FAIL x 1 g(x) = 0 x is solution to constrained optimization problem min x M x 2 (14) g(x)=0 Distance from mean to failure surface is measure of reliability Higher order approximations of failure surface also possible M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 18 / 34

Problem Statement Standard Normal Space To find design point, transform all random variables to space of uncorrelated, standard normals, u, with joint PDF ( 1 φ(u) = exp 1 ) (2π) n/2 2 ut u Find design point, u, in u-space (15) min u 2 (16) g(u)=0 Rotational symmetry of u-space: p f from inverse normal CDF 3 0.2 2 φ(u) 0.15 0.1 0.05 0 un 2 2 0 2 0 2 2 3 u n u 1 3 2 1 0 1 2 3 u 1 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 19 / 34 1 0 1

Computing Reliability HLRF Iteration Find design point, u, by Hasofer-Lind-Rackwitz-Fiessler (HLRF) iteration Start at u 0 (usually at mean point 0), then u m+1 = ( α T u m + g(x) u g where α = ug u g Converges to a point where with the reliability index Probability of failure from normal CDF ) α (17) (18) u = βα (19) β = α T u (20) p f = Φ( β) (21) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 20 / 34

Convergence of HLRF Iteration Design point, u, where Computing Reliability or magnitude is less than a specified tolerance u α T u α = 0 (22) u α T u α 2 tol (23) u n α α T u m α u m α T u m α u m u 1 g = 0 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 21 / 34

Failure Sensitivity Sensitivity Vector The vector α is unit normal to failure surface α = ug u g In addition to providing search directions for HLRF iteration, component α i indicates The nature of random variable i in u-space (α i > 0 load, α i < 0 resistance) The importance (via magnitude) of random variable i relative to other random variables in u-space Transform to original x-space in order to determine nature and importance of random variables in problem domain Depends on α and Jacobian of x-u transformation M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 22 / 34

Failure Sensitivity Computation of Sensitivity Vector All measures of sensitivity and importance depend on the gradient of the limit state function u g = x g x u x/ u Jacobian of x-u transformation (24) For a limit state function defined in x-space in terms of finite element response, U f, which depends on an uncertain parameter θ, the chain rule gives x g = g U f U f θ θ x g/ U f analytic derivative of limit state function U f / θ nodal response sensitivity of FE model θ/ x mapping of random variable to FE model parameter (25) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 23 / 34

Failure Sensitivity Direct Differentiation Analytical U f / θ in gradient of limit state function Pros: Cons: Requires as many limit state function evaluations as HLRF iterations Compute sensitivity as analysis proceeds Use factorized stiffness matrix, K T, for each parameter of FE model Not subject to round off error (beyond that of FE analysis) Difficult to implement Must be implemented for all objects present in FE analysis when path-dependency is present M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 24 / 34

Failure Sensitivity Finite Differences Approximate U f / θ in gradient of limit state function Pros: Cons: Easy to implement Applicable for free to all objects in FE analysis Requires more limit state function evaulations than HLRF iterations Re-run analysis for each parameter perturbation Reform and factorize tangent stiffness matrix, K T, for each parameter perturbation at each iteration Subject to round off error based on parameter perturbation M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 25 / 34

Cantilever Example Cantilever Reliability Determine probability that peak displacement exceeds 4.0 m g(x) = 4.0 U x 1 LN(4.1e5, 0.05) MPa maps to yield stress, σ y x 2 LN(2.0e8, 0.1) MPa maps to elastic modulus, E x 3 N(1710, 0.15) kn maps to applied load, P max x 4 N(5.0, 0.02) m maps to cantilever length, L x 5 N(0.5, 0.04) m maps to cross-section depth, d 2000 Load, P (kn) 1000 0 1000 Limit State 2000 5 0 5 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 26 / 34

Cantilever Example Reliability Results HLRF converges in 7 iterations with both approaches to U f / θ: DDM 7 evaluations of limit state function Finite Difference 42 evaluations of limit state function Reliability index, β = 1.983 Probability of failure, p f = Φ( 1.983) = 2.368% Random Variable, i µ i xi α i 1 σ y 4.1e5 kpa 3.897e5 kpa -0.2317 2 E 2.0e8 kpa 1.996e8 kpa -0.005413 3 P max 1710 kn 2032 kn 0.6393 4 L 5.0 m 5.100 m 0.5008 5 d 0.5 m 0.4787 m -0.5356 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 27 / 34

Cantilever Example Mean and Design Point Response Cantilever load-displacement response at realization of random variables in x Load, P (kn) 3000 2000 1000 0 1000 2000 Mean Design Pt Limit State 3000 5 0 5 Displacement, U (m) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 28 / 34

Reinforced Concrete Frame Limit state 2% roof drift RC Frame Example g(x) = 0.02(8.3) U 7 (26) 7 (9) 8 (10) 9 (2) (4) (6) 3.7 m 4 (7) 5 (8) 6 (1) (3) (5) 4.6 m 1 2 3 7.3 m 7.3 m Outer column sections Inner column sections Girder sections b=610 mm b=610 mm h=610 mm h=610 mm Bar area = 700 mm 2 h=690 mm b=460 mm Bar area = 1000 mm 2 M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 29 / 34

RC Frame Example Material Properties Capture axial-moment interaction of reinforced concrete using numerical integration of stress-strain response over cross-sections E=200,000 MPa f y =420 MPa α=0.05 σ f y E 1 1 αe ǫ f c =28 MPa f c =36 MPa f cu =33 MPa ǫ c =0.002 ǫ c =0.005 ǫ cu =0.06 ǫ cu =0.02 σ ǫ cu ǫ c ǫ ǫ cu ǫ c σ ǫ -f y f c f cu f c (a) (b) (c) M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 30 / 34

RC Frame Example Uncertainty Modeling Total of 142 random variables For each member: Reinforcing steel material properties (LN and 0.6 group correlation) Concrete material properties (LN and 0.6 group correlation) Cross-section width and height (N and no group correlation) Reinforcing bar area (N and no group correlation) For column members only: Cover concrete thickness (N and no group correlation) For all joints: X and Y coordinates (N and no group correlation) Deterministic gravity and lateral loads M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 31 / 34

RC Frame Example Pushover Load-Displacement At mean and design point realizations HLRF converges in 9 iterations: β = 2.76, p f = 0.29% DDM 9 limit state function evaluations Finite difference 1200+ limit state function evaluations M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 32 / 34

RC Frame Example Importance Ranking Highest importance on depth of members framing in to interior joint Steel yield stress and cover thickness also rank high Using γ importance vector instead of α as it accounts for correlation among random variables M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 33 / 34

RC Frame Example Conclusions DDM is accurate and efficient alternative to finite differences in assessing structural response sensitivity Fewer function evaluations in reliability analysis Other applications in optimization and system identification Ongoing work: Live load reliability analysis of bridge girders Particle finite element reliability analysis of fluid-structure interaction M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 34 / 34

References T. Haukaas and M. H. Scott. Shape sensitivities in the reliability analysis of nonlinear frame structures. Computers and Structures, 84(15 16):964 977, 2006. M. H. Scott and F. C. Filippou. Exact response gradients for large displacement nonlinear beam-column elements. Journal of Structural Engineering, 133(2):155 165, February 2007. M. H. Scott and T. Haukaas. Software framework for parameter updating and finite-element response sensitivity analysis. Journal of Computing in Civil Engineering, 22(5):281 291, 2008. M. H. Scott (OSU) Sensitivity and Reliability AMC Seminar, 2011 34 / 34