This example is concerned with validating MSC.Fatigue against a widely used, in‑house code from a large aerospace company. Only the local strain (crack initiation) method is considered here. The MSC.Fatigue results are compared not only against the in‑house code but also to actual test data.

For this series of benchmark problems, the geometry used is arbitrary because stress time histories are used as the loading inputs. Since MSC.Fatigue is highly coupled to the FE model and needs to retrieve stresses (or strains) from the FE analysis, an arbitrary geometry and loading were created so as to reproduce a unit stress due to a unit load. This ensures that the actual stress time history is used exactly as input without any scaling necessary.

If you look carefully at the geometry, you will notice that the units are inches and that a 1 PSI pressure has been applied at one end, while the other end has been constrained. At Node 1 a stress of 1000 PSI results. See the definition of the time history in Step 3: Loading Histories, 1240 of this example.

The geometry is in a file called code.out with the stress results in a file called code.txt, which needs to be converted via RESTXT to code.res similar to all previous example problems. Files can be found in

<install_dir>/mscfatigue_files/examples

The material used is the same as that used by the in-house code (i.e., Aluminum 7075-T6). In order to get the exact same strain data curves into MSC.Fatigue, the strain-life and stress-strain curves used by the in-house code were digitized and curve fit to produce the material properties as shown in Table 14‑1. The MSC.Fatigue calculations assume a good machined surface finish and no surface treatment. The following keystrokes can be used to enter the materials information into the MSC.Fatigue materials database manager. Once the materials data is entered, you can inspect the corresponding strain-life and stress‑strain plots, if you wish. Use these operations to enter the data into the MSC.Fatigue materials database:

The loading histories, as mentioned before, are stress histories. Because the FE model was created such as to give a unit stress due to a unit load, the stress time histories are taken as is, with no scaling factors applied. Table 14‑9 shows the stress time history spectra for each problem. Basically, each block consists of an initial overload (and underload sometimes) followed by 999 (or 998) constant amplitude cycles for a total of 1000 cycles per block.

We will be creating these time histories by reading in ASCII versions of them which can be found in the examples directory delivered with your MSC.Fatigue system. The names of the files are coden.asc where the n refers to the time history number from Table 14‑9. Note that the values in these files are PSI and not KSI as reported in Table 14‑9. This is because we are using stress at Node 1 which was reported to be 1000 PSI which scales the values in the files to KSI.

The following keystrokes are used to set up the load spectra into PTIME. At the end of the PTIME execution, there will be files called coden.dac in your directory. The ASCII read utility of PTIME is used to read in the load spectra and convert it into a form that PTIME can understand. This session assumes you are starting in a clean, empty directory.

The files needed to set up the MSC.Fatigue job can be found in the examples directory delivered with your MSC.Fatigue system. The are called code.out and code.txt, where code.out contains the arbitrary geometry and code.txt contains the finite element results. The finite element results come from MSC.Patran FEA and therefore need to be translated into the binary results file code.res by way of the MSC.Patran utility RESTXT. The following operations setup the MSC.Fatigue job:

Before completely filling out the Materials form we need to digress a bit and create a special MSC.Patran group which contains only the node with the unit stress associated with it. We only care to do a fatigue analysis at this node since the entire geometry is actually arbitrary.

Quickly view the results.

You will notice that the answer from MSC.Fatigue is a very large number of Repeats. This number is actually meaningless without calibration. Our goal now is to calibrate MSC.Fatigue to come in line with the actual test data results of 47 Repeats.

To illustrate this, invoke the MSC.Fatigue module called FEFAT by typing the symbol fefat at the system prompt or pick Optimize from the Results form in MSC.Fatigue. First reanalyze the original problem to see that the same answer is achieved and then set the parameter optimization to do a scale factor back calculation. You will be asked for a back calculation life. This life should be the actual test value of 47.657 Repeats. Reanalyze and note the scale factor calculated. Now you can run the rest of the problems by changing the time history and using this new calibration scale factor.

A back calculated scaling factor of about 2.2 is reported. Note the exact number on your computer and use it in the rest of the exercise. Instead of re-entering the MSC.Patran system to set up each additional problem, invoke the fatigue preprocessor routine FEFAT by typing the symbol fefat from the system prompt and enter the following keystrokes.

We can now use our calibration scale factor.

Table 14‑10 tabularizes the measured and predicted lives. MSC.Fatigue used three different mean stress correction factors, STW, Morrow, and no correction. Figure 14‑8 plots the predicted vs. measured lives. Repeat the last few operations (fefat) for each different problem you wish to review.

Note that problem #1 is not listed in the results table. This is because problem #1 was used to “calibrate” MSC.Fatigue, as was done for the in-house code as well. A “K” factor was determined to make the results for the first problem fall exactly in line with the test data. This factor was then applied to the remaining six problems.

In general, MSC.Fatigue and the in-house code are in good agreement with the measured lives, falling well within the scatter band of Figure 14‑8. However, the strain-life durability is very sensitive to the effects of positive mean stress and is highly dependent of surface finish and surface treatment.