1. Postprocessing Chapter Four

2. Training Manual October 15,, 2001 Inventory # 001565 4-2 4. Postprocessing What is different about nonlinear postprocessing? The same fundamental objectives apply to both linear and nonlinear analyses. –Plot and list your results in ways that make them useful for engineering decision-making. Will my part fail? Is it over designed? How can I improve it? Present results to other decision makers (clients, customers, etc.). –Verify the validity of your results. Are the results correct? Nonlinear postprocessing differs in that: –You review an expanded set of results items, over a range of “time.” –Verification includes checking for convergence. –Results cannot be scaled or superimposed.

3. Training Manual October 15,, 2001 Inventory # 001565 4-3 …Postprocessing In this chapter, we will present a brief introduction to the basics of nonlinear postprocessing via the following topics: A. Checking for convergence B. The general postprocessor C.The results viewer D. The time-history postprocessor E. Verification The purpose is to give you an understanding of special postprocessing techniques for nonlinear analyses.

4. Training Manual October 15,, 2001 Inventory # 001565 4-4 Postprocessing A. Checking for convergence Obviously, engineering decisions cannot be based on unconverged results. Before continuing with postprocessing, you must verify that you have a converged solution. –If unconverged, find out what went wrong. –If converged, gain insight into how to make a similar analysis more efficient next time. There are many ways to verify convergence. –Review the output, error, and monitor files. –List results summary in the general postprocessor. –Graph time-history results.

5. Training Manual October 15,, 2001 Inventory # 001565 4-5 Postprocessing … Checking for convergence Review the output file. –If converged – convergence messages displayed for last substep, at expected “time” at end of last load step. –If unconverged – restart information is given.

6. Training Manual October 15,, 2001 Inventory # 001565 4-6 Postprocessing … Checking for convergence Review the error file. Review the monitor file. –Make sure a solution was completed at the expected “time” at the end of the last load step.

7. Training Manual October 15,, 2001 Inventory # 001565 4-7 Postprocessing … Checking for convergence List the results summary in the General Postprocessor. –Unconverged results are written for substep 999999. –Don’t postprocess step 999999, except to debug nonconvergence.

8. Training Manual October 15,, 2001 Inventory # 001565 4-8 Postprocessing … Checking for convergence Graph time-history results. –A solution that has converged for several substeps, but that terminates with an unconverged substep, will usually display a characteristic straight-line “jump” in time-history graphs. “Jump” Smooth curve Last converged substep Substep 999999 (unconverged)

9. Training Manual October 15,, 2001 Inventory # 001565 4-9 Postprocessing … Checking for convergence Nonconvergence is sometimes expected. –Structures that become physically unstable: Nonlinear buckling. Complete plastic section. Last converged solution will have engineering significance F u Convergence at full load will never be achieved

10. Training Manual October 15,, 2001 Inventory # 001565 4-10 Postprocessing B. The general postprocessor The general postprocessor allows you to review results for the entire model, at a single point in time. You can also use the general postprocessor to animate results over a range of time or results sets. SEQV plot at time = 2.0 Animated response history

11. Training Manual October 15,, 2001 Inventory # 001565 4-11 Postprocessing … The general postprocessor To plot or list results at a given point in time: –Make sure the database is in memory. (RESUME jobname.db if necessary.) –Choose the desired results set using the results summary list. General Postproc > Results Summary –Avoid using General Postproc > -Read Results- By Time/Freq Can lead to unnecessary interpolation errors. Reading results by Time Results summary list

12. Training Manual October 15,, 2001 Inventory # 001565 4-12 Plot or list results at a given point in time (cont’d): Use most of the same plot and list operations as for linear postprocessing. Note that all von Mises (equivalent) strains (elastic, plastic, creep, etc.) will be calculated using effective Poisson’s ratio. By default: –EFFNU = user input for EPEL and EPTH –EFFNU = 0.5 for EPPL and EPCR –EFFNU = 0.5 if reference material is Hyperelastic –EFFNU = 0 for line elements (includes beam, link, and pipe elements, as well as discrete elements) Note: Default EFFNU can be over-ridden in POST1 with the AVPRIN command if necessary (I.e. EFFNU = 0.495 for nearly incompressible material ) Postprocessing … The general postprocessor

13. Training Manual October 15,, 2001 Inventory # 001565 4-13 Postprocessing …The general postprocessor Animation allows you to review the response history of the entire model. Be sure to use the correct animation option: –PlotCtrls > Animate > Over Time –PlotCtrls > Animate > Over Results –No other animation options are appropriate for nonlinear results! Click the desired button Specify a “time” range Animate Over Time

14. Training Manual October 15,, 2001 Inventory # 001565 4-14 Postprocessing...The general postprocessor Important: Because nonlinear response is not proportional, the principle of superposition is invalid. –Load case combinations (General Postproc > Load Case) are invalid. –Results scaling is invalid. F1F1 u1u1 F2F2 u2u2 F 1 +F 2 u 1 +u 2 + Superimposed results are wrong! 

15. Training Manual October 15,, 2001 Inventory # 001565 4-15 Postprocessing C. Results Viewer Post-processing of large models with many time steps may slow your system performance (memory limitations, large amounts of disk I/O). The Results Viewer may provide a 10x post-processing performance improvement. It is a specialized postprocessing menu and graphic system. –You can quickly access and compare results sets for different substeps. –It provides an easy to use menu system for quick results viewing. –It uses the PGR file (a specialized graphics object).

16. Training Manual October 15,, 2001 Inventory # 001565 4-16 The PGR file can be created two different ways … –Use the POUTRES command before solving to write a jobname.pgr file during solution (greatest speed advantage). –Solution > Output Cntrls > PGR file … Postprocessing … Results Viewer Hold CTRL key for multiple selection

17. Training Manual October 15,, 2001 Inventory # 001565 4-17 Postprocessing … Results Viewer –Or use the PGWRITE command after solution to write a jobname.pgr file. –General Postproc > Write PGR File…

18. Training Manual October 15,, 2001 Inventory # 001565 4-18 Postprocessing … Results Viewer Open the Results Viewer from the General Post Processor.

19. Training Manual October 15,, 2001 Inventory # 001565 4-19 Element Plot Nodal/element/vector/trace results plots Query Results Animate using PNG files List Results Capture/print image Results Set Locator Postprocessing... Results Viewer

20. Training Manual October 15,, 2001 Inventory # 001565 4-20 The graphics window becomes “Context Sensitive”. Postprocessing... Results Viewer Right Click on Model Right Click on Contour Legend Right Click on Color Contour Bar

21. Training Manual October 15,, 2001 Inventory # 001565 4-21 Postprocessing D. The time-history postprocessor The time-history postprocessor allows you to display analysis results at specific points in the model in X-Y response history graphs. You can graph: –Deflection vs. time. –Stress vs. strain. –Load vs. deflection. –And many other items.

22. Training Manual October 15,, 2001 Inventory # 001565 4-22 Postprocessing … The time-history postprocessor To create a time-history graph, you must first store results items as temporary variables. –The program identifies these variables by index numbers. –The program automatically defines “Time” as variable 1. “Time” is used for the X-axis by default. –You can assign other results items to variable 2, etc. –You can specify a different variable for the X-axis. Open Results File: TimeHist Postpro> Variables Viewer

23. Training Manual October 15,, 2001 Inventory # 001565 4-23 Postprocessing … The time-history postprocessor Add Variable to Viewer Table Assign unique variable name –in this case “DISPLACEMENT” for UY DOF solution

24. Training Manual October 15,, 2001 Inventory # 001565 4-24 Postprocessing … The time-history postprocessor Select the applicable node by graphical picking or via command input window Pick node (in element or node plot) Repeat for other variables, as desired.

25. Training Manual October 15,, 2001 Inventory # 001565 4-25 Postprocessing … The time-history postprocessor Each variable represents one results item at one node. Total force at node 1 UY displacement at node 1 Calculator allows user to perform many mathematical operations on variables

26. Training Manual October 15,, 2001 Inventory # 001565 4-26 Postprocessing … The time-history postprocessor Graph the force vs deflection curve –By default, the X-Axis will be variable 1, or “Time.” You can specify a different variable by changing the selection in the X-Axis column 1. Specify the DISPL variable for the X-axis 2. Highlight the FORCE variable 3. Select the graph button

27. Training Manual October 15,, 2001 Inventory # 001565 4-27 Postprocessing … The time-history postprocessor With variables defined, and the desired X-Axis variable specified, you can plot the X-Y graph by selecting the graph button on the variable viewer menu Load-Deflection graph Let’s see how to improve the quality of this graph.

28. Training Manual October 15,, 2001 Inventory # 001565 4-28 Postprocessing … The time-history postprocessor You can modify many aspects of the graph axes. PlotCtrls > Style > Graphs > Modify Axes –Axis labels. –X and Y ranges. –Font size. –Reverse X and Y axes. –And more. Labels X range Y range Font size Reverse Y-axis

29. Training Manual October 15,, 2001 Inventory # 001565 4-29 Postprocessing … The time-history postprocessor You can add a grid background to the graph. PlotCtrls > Style > Graphs > Modify Grid

30. Training Manual October 15,, 2001 Inventory # 001565 4-30 Postprocessing … The time-history postprocessor You can also perform math operations on variables using the variable viewer calculator. –Take absolute value –Differentiate. –Integrate. –And more. However, be careful not to scale or superimpose results! –Remember, for nonlinear analyses, the principle of superposition is invalid.

31. Training Manual October 15,, 2001 Inventory # 001565 4-31 Postprocessing … The time-history postprocessor You can also change graph colors. PlotCtrls > Style > Colors > Reverse Video PlotCtrls > Style > Colors > Graph Colors –Curve color, etc. Improved Load-Deflection graph

32. Training Manual October 15,, 2001 Inventory # 001565 4-32 Postprocessing E. Verification You cannot simply assume that an analysis that converged with no error messages will give valid results. Various modeling problems can cause a seemingly well- behaved solution to be incorrect. –Mesh discretization (too coarse a mesh). –Mesh distortion (poor element shapes after first iteration). –Errors in material characterization. –Failure to identify potential contact regions. –Improper boundary conditions. You must always verify that your results are correct.

33. Training Manual October 15,, 2001 Inventory # 001565 4-33 Postprocessing … Verification Engineering judgment is your first and best verification tool. –Are your results “reasonable?” –Do they conform to your expectations? Good judgment is based on: –Experience. –Benchmark studies. –Physical testing or known behavior of prototype parts. –A skeptical, vigilant frame of mind. Good practice always requires that you reconcile analytical results with rational expectations. Do these results make sense?

34. Training Manual October 15,, 2001 Inventory # 001565 4-34 Postprocessing … Verification Element contour plots can help reveal mesh discretization error. –Discontinuities and missing contours are common indicators of a mesh that is too coarse. Path plots can also be a useful way to check the mesh discretization density. –A smooth path plot is usually expected. Jagged plots may indicate that the mesh is too coarse.

35. Training Manual October 15,, 2001 Inventory # 001565 4-35 Postprocessing … Verification Deformed shape plots can reveal highly distorted elements. Poor element shapes may yield inaccurate results. –Most shape checking is active only at the start of solution. –In large-strain analyses, elements might distort into unviable shapes after the first iteration. –Plot the displaced shape to check distorted element shapes. Originally quad-shaped element deforms to 180 0 corner angle - results are questionable.

36. Training Manual October 15,, 2001 Inventory # 001565 4-36 Postprocessing … Verification Stress-strain plots can help verify the material response. –Does the known stress-strain data extend beyond the range of strain experienced in the model? –Do stress-strain results match the input stress-strain data? Input dataTime-history graph

37. Training Manual October 15,, 2001 Inventory # 001565 4-37 Postprocessing … Verification Deformed shape plots will reveal if any contact surfaces have excessive over-penetration. Animating the deformed shape over time will show up any overlapping regions where you failed to identify potential contact surfaces.

38. Training Manual October 15,, 2001 Inventory # 001565 4-38 Postprocessing … Verification Time-history plots should generally be smooth. –A jagged time-history plot may have a valid physical explanation. Such as, windshield wiper blade “chatter.” –Without a good physical explanation, jagged results are questionable. Might indicate that time step size is too large. Animations should similarly exhibit smooth motion.

39. Training Manual October 15,, 2001 Inventory # 001565 4-39 Postprocessing … Verification Unrealistic boundary conditions will ruin the validity of your analysis. –Both over- and under-constrained conditions will give incorrect displacements and erroneous stresses. Careful review of deformed shape plots, stress contour plots, and other postprocessing data will help you answer the question, “Do these results make sense?”

40. Training Manual October 15,, 2001 Inventory # 001565 4-40 Postprocessing F. Workshop Please refer to your Workshop Supplement for instructions on: W7. Postprocessing – Tensile Specimen W8. Postprocessing - Results Viewer