Transcription

NESSUS Users’ ManualDavid RihaBarron BichonJohn McFarlandAugust 11, 2015Copyright c 1998–2013 bySouthwest Research InstituteSimeon Fitch

Contents1 Overview42 Getting Started73 Problem Definition3.1 Define Fault Tree . . . . . . . . . . . . . . .3.2 Problem Statement . . . . . . . . . . . . . .3.3 Random Variables . . . . . . . . . . . . . .3.3.1 Correlations . . . . . . . . . . . . . .3.3.2 Confidence Bounds . . . . . . . . . .3.4 Response Models . . . . . . . . . . . . . . .3.4.1 Regression . . . . . . . . . . . . . . .Polynomial Regression Models . . .Gaussian Process Regression Model3.4.2 Dynamically Linked . . . . . . . . .3.4.3 Predefined . . . . . . . . . . . . . . .3.4.4 Numerical . . . . . . . . . . . . . . .Overview . . . . . . . . . . . . . . .Numerical Model Usage in NESSUSCreate Mappings . . . . . . . . . . .Variable Mapping . . . . . . . . . .Response Selection . . . . . . . . . .8991315181920202122232323242729324 Deterministic Analysis344.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.2 Importing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 Probabilistic Analysis5.1 Analysis Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.2 Analysis Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.2.1 Specified Probability Levels . . . . . . . . . . . . . . . . . . . . . .5.2.2 Specified Performance Levels . . . . . . . . . . . . . . . . . . . . .5.2.3 Full Cumulative Distribution . . . . . . . . . . . . . . . . . . . . .5.2.4 Global Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . .5.3 Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.3.1 Monte Carlo method (MONTE) . . . . . . . . . . . . . . . . . . .5.3.2 Latin Hypercube Simulation (LHS) . . . . . . . . . . . . . . . . . .Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.3.3 First- and Second-Order Reliability Methods (FORM and SORM)Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.3.4 Mean Value methods (MV, AMV, AMV ) . . . . . . . . . . . . .Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1.404040404343434344454646464647

.3.14Curvature-based adaptive importance sampling with advanced mean value MPP search(AMV AIS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Importance sampling w/radius reduction factor (ISAMF) . . . . . . . . . . . . . . . .Importance sampling w/user-defined radius (ISAMR) . . . . . . . . . . . . . . . . . .Importance sampling at user-defined MPP (ISMPP) . . . . . . . . . . . . . . . . . . .Plane-based adaptive importance sampling (AIS1) . . . . . . . . . . . . . . . . . . . .Curvature-based adaptive importance sampling (AIS2) . . . . . . . . . . . . . . . . . .Efficient Global Reliability Analysis (EGRA) . . . . . . . . . . . . . . . . . . . . . . .Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Response Surface Method (RSM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Gaussian Process Response Surface Method (RSM GP) . . . . . . . . . . . . . . . . .Variance Decomposition (VARDCMP) . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Results Visualization6.1 Overview . . . . . . . . . . .6.2 Deterministic Results . . . . .6.3 Probabilistic Analysis Results6.4 Global Sensitivity Results . .6.5 Chart Formatting Controls .6.6 Viewing Analysis Files . . . .4749495050505152525353.555555575858587 System Requirements7.1 Installation . . . . . . . . . . . . . . . . . . . . .7.1.1 Unix Platforms . . . . . . . . . . . . . . .7.1.2 Microsoft Windows . . . . . . . . . . . . .7.2 Distribution Structure . . . . . . . . . . . . . . .7.3 External Analysis Packages . . . . . . . . . . . .7.3.1 ABAQUS (cf. Section B.1) . . . . . . . .7.3.2 ANSYS (cf. Section B.2) . . . . . . . . . .7.3.3 LS-DYNA (cf. Section B.3) . . . . . . . .7.3.4 MSC/NASTRAN (cf. Section B.4) . . . .7.3.5 MATLAB . . . . . . . . . . . . . . . . . .7.3.6 USER DEFINED (cf. Section B.6) . . . .7.4 Updating External Interfaces Configuration File .7.4.1 Example Case . . . . . . . . . . . . . . . .6161616161626262626262636363. . . . . . . . . . . . . . .Visualization. . . . . . . . . . . . . . . . . . . . . .Appendices64A Command Line Execution64B InterfacesB.1 ABAQUS . . . . . . . . . . . . . . . . . . . .B.1.1 Introduction . . . . . . . . . . . . . .B.1.2 System Requirements . . . . . . . . .B.1.3 Execution Command . . . . . . . . . .B.1.4 Responses . . . . . . . . . . . . . . . .B.1.5 Examples . . . . . . . . . . . . . . . .B.2 ANSYS . . . . . . . . . . . . . . . . . . . . .B.2.1 Introduction . . . . . . . . . . . . . .B.2.2 System Requirements . . . . . . . . .B.2.3 Execution Command . . . . . . . . . .B.2.4 Responses . . . . . . . . . . . . . . . .B.2.5 Examples . . . . . . . . . . . . . . . .B.2.6 User Defined Response Extraction . .B.2.7 ANSYS Defined Response Extraction2.656565666666696969707070727273

B.3 LS-DYNA . . . . . . . . . . . . .B.3.1 Introduction . . . . . . .B.3.2 System Requirements . .B.3.3 Execution Command . . .B.3.4 Responses . . . . . . . . .B.3.5 Examples . . . . . . . . .B.4 MSC.NASTRAN . . . . . . . . .B.4.1 Introduction . . . . . . .B.4.2 System Requirements . .B.4.3 Execution Command . . .B.4.4 Input Card RequirementsB.4.5 Responses . . . . . . . . .B.4.6 Examples . . . . . . . . .B.5 NASGRO . . . . . . . . . . . . .B.5.1 Introduction . . . . . . .B.5.2 System Requirements . .B.5.3 Execution Command . . .B.5.4 Variable Mapping . . . .B.5.5 Responses . . . . . . . . .B.6 User Defined External Models . .B.6.1 Introduction . . . . . . .B.6.2 System Requirements . .B.6.3 Execution Command . . .B.6.4 Responses . . . . . . . . .B.6.5 Examples . . . . . . . . .B.6.6 UPOST Subroutine . . .B.7 MATLAB . . . . . . . . . . . . .B.7.1 Windows . . . . . . . . .B.7.2 Linux . . . . . . . . . . .B.7.3 Mac OS X . . . . . . . . .C Problem Statement BNF 84868688889090909192D NESSUS Tutorials93D.1 Design of Experiment (DOE) Tutorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93D.1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93D.1.2 Example Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 933

Chapter 1OverviewThe purpose of the NESSUS GUI is to aid the development of NESSUS analysis input files, and the visualization of the generated results. Although a NESSUS analysis input file may be crafted “by hand,” theGUI provides a more efficient and intuitive means of working with the input files, guiding one through theanalysis process and detecting potential problems.The NESSUS GUI employs the traditional document data paradigm, whereby one or more NESSUS“documents” may be created, opened, and modified. The contents of the analysis are summarized by adisplayed outline, which changes as the analysis is developed. The outline (Figure 1.1) guides you throughthe steps of developing an analysis, ending with the initiation of a NESSUS job run and the visualization ofthe computed results.You navigate through the contents of the analysis by selecting the nodes of the outline. The area tothe right of the outline changes depending on which outline node is selected. Although the outline helpsguide you through the analysis development in a linear fashion, you have the freedom to navigate throughthe outline in any order you wish. However, certain portions of the outline do not appear until the datarequired for them to appear has been entered.It is important to note that an analysis is developed in a top down approach, starting with defining aproblem statement and random variables, followed by model definitions, analysis type selection, analysisexecution, and ending with results visualization. Once an analysis has been defined, you can go back andmodify inputs as needed and re-perform the analysis.A powerful feature of NESSUS is ability to link models together in a sequential fashion. In the problemstatement window, each model is defined only in terms of input and output. This improves readability,conveys the essential flow of the analysis, and allows complex reliability assessments to be defined whenmore than one model is required, for example, a stress computed from a finite element model can be passedinto an analytical equation that computes fatigue life. In the example shown here, a response model (σmax SigmaMax) is linked to a failure model (g Paris crack growth law), i.e., 1 n/21 n/22 af aig n1.16 10 9 (2 n) (Y σmax n)σmax 3P S2B 2or in Fortran syntax,g aMax*sqrt(Pi))**n)SigmaMax 3*P*S/2/B**2The multiple model capability is also useful for breaking a complex model into smaller more manageableparts. For example, a complicated analytical response model can be defined using two or more expressions,which when executed in sequence, results in the original analytical response model.4

Figure 1.1: NESSUS outline5

You may save and restore an analysis at any point in its development. The complete state of the analysisis saved in the “jobname.dat” file, which can be read using any text editor.6

Chapter 2Getting StartedAfter the GUI has been started, you must either create a new analysis file, or open an existing one. Thiscan be done via one of the first two buttons of the tool bar, or from the File menu.The outline (Figure 1.1 will appear after starting new or resuming from a previously saved session. Theoutline is intended to logically guide the user through the problem setup, analysis and visualization of results.7

Chapter 3Problem DefinitionThe definition of the problem to be solved takes place in the Problem Definition outline node, as shown inFigure 3.1. This includes the definition of the functional relationship between the model inputs and outputs(Section 3.2) and the defintion of certain model inputs as random variables (Section 3.3).Figure 3.1: Problem Definition outline node8

Note that textual information about the project can also be stored using the Edit Project Informationoutline node. This screen allows you to enter a title and description for your analysis, which will appearat the top of the saved analysis file. The title and description are purely informational, and are ignored byNESSUS when a job is invoked. The job name field is a read-only display of the jobname that NESSUS willuse for saving analysis data and results.3.1Define Fault TreeThe first item in the problem statement window is the Define Fault Tree button. When performing ananalysis that considers only one failure mode, this step is not necssary, and the user should continue withthe problem statement definition, as described in Section 3.2.System reliability analysis involves multiple g-functions. NESSUS uses a fault tree structure to definethe system failure. A fault tree has three major characteristics; bottom events, combination gates and theconnectivity between the bottom events and gates. NESSUS is presently limited to AND and OR gates.Conditional gates can be simulated using the AND gates with appropriately defined conditional performancefunctions.NESSUS uses a graphical approach to define the fault tree as shown in the figure below. First the pfblock is selected to begin definition of the fault tree. Next gates and bottom events are added to define thefault tree.The following dialog box appears when a bottom event is defined. The variable name is limited to 8characters. Operator and Value are currently fixed. The limit state for each bottom event is defined inthe problem statement (cf. Section 3.2) window and must be formulated such that failure occurs when theperformance is less than or equal to zero (e.g., g STRENGTH - STRESS). The probability statementshown in the above figure will be inserted in the problem statement window.Probabilistic analysis methods for system problems are restricted to Monte Carlo simulation (MONTE (cf.Section 5.3.1)) and importance sampling methods (AIS2 (cf. Section 5.3.10) and ISAMF (cf. Section 5.3.6)).The Analysis Type must be Specified performance levels (cf. Section 5.2.2) with a Z Value of 0.0.3.2Problem StatementThe problem statement editor is the central part of the user interface. Through this screen the performancemodel is defined in direct, easy to understand mathematical notation. This notation (see Appendix C9

Figure 3.2: Probabilistic Fault Tree Definition in NESSUS10

for grammar definition) is displayed with a syntax highlighting feature to highlight errors and improvereadability. Each line represents a separate evaluation where variables are assigned constant values, analyticalexpression results, or model responses. The apply button must be clicked for changes in the window to becomeactive.Sequential models are evaluated from the bottom up with the top equation defining the performance. Theexample below computes the fatigue life of a three point bend specimen. The performance is life computedusing a Paris equation with input of stress from the second equation.Variable and function names are restricted to 8 characters. Analytical expressions can contain independent and dependent variables as well as constants. Analytical expressions use standard Fortran syntax asfollows: Addition: Subtraction: Multiplication: * Division: / Exponents: ** (ab a**b) Common intrinsic functions: sqrt, exp, log, log10, sin, cos, tan, cotan, atan, asin, acos,sinh, cosh, tanh, abs, int11

The argument for trigonometric intrinsic functions use radiansSyntax highlighting is as follows: blue – functions green – variables black – operators, groupings, and numbers orange – intrinsics red – error (typically more than 8 characters for a variable or function name, non matching parentheses,equation error)Response models are declared as functions as shown in the following example. This example computesthe fatigue life of a three point bend specimen. The equations are evaluated from the bottom up with thefinal performance defined by the life variable. This example uses the stress computed from the “fe” functionas input to the Paris equation. The “fe” function is defined in the “Define Response Models” section of theoutline in the GUI. See the Response Models (cf. Section 3.4) section for information on how to define theresponse model.When defining functions the form of the equation cannot contain any mathematical operators or intrinsicfunctions. Note that parameters can also be included in the problem statement by equating the variable toa value.Multiple response variables can be equated to a function using the format in the following figure. Eachof these response variables will be defined in the "Define Response Models" section in the GUI.The GUI differentiates between dependent variables and independent random variables. After a problemstatement is entered or modified, and the Apply button is clicked, the GUI parses the equation and determineswhether a variable is dependent or independent. Although default inputs are provided, independent variablesmust be defined by specifying a distribution, mean and standard deviation. Variables can be changed fromindependent to dependent and back by changing the problem statement definition.Below the problem statement editor is a table displaying the variables detected when the Apply buttonis pressed. The inputs for the independent variables can be edited in this table. A detailed description ofthis table can be found in the Random Variables (cf. Section 3.3) section.Reminders for defining the problem statement:12

Equations are evaluated from the bottom up Variable and function names are limited to 8 characters Press the "Apply" button to apply changes3.3Random VariablesAfter a problem statement is entered and applied, the GUI processes all variables to determine whetherthey are dependent variables or independent random variables. Random variables are assigned a defaultdistribution that can then be edited in several ways.The primary mechanism for editing random variables is the “Edit Random Variables” node, as shownin Figure 3.3. The variable table lists all variables defined in the problem statement. Editable fields in thetable are shown in white. The properties of a particular random variable can be edited by double-clicking onthe editing icon under the “Distribution” column. This brings up an editing window, as shown in Figure 3.4,where the distribution type and distribution parameters can be specified.Figure 3.3: Edit Random Variables node.The following distribution types are supported: Beta*13

Figure 3.4: Random variable editing screen14

Chi-square Exponential* Frechet* GEVDmax (Generalized Extreme Value for maxima)* GEVDmin (Generalized Extreme Value for minima)* Gamma* Gumbel Lognormal Normal Pareto* Triangular* Truncated Normal Truncated Weibull Uniform WeibullThe starred distributions are only supported by following probabilistic analysis methods (see Section 5.3):MONTE, LHS, ISMPP, EGRA, and RSM GP.The parameters governing most random variables can be specified using one of two schemes: “Naturalparameters” or “Moments.” All random variables c