I’ve been working in the simulation industry for a few years now and this is a question that I’ve been asked rather frequently. Although I’ve been told it’s not good to answer a question with another question, sometimes it’s necessary. Before I answer that question I would ask: “With respect to what? With respect to input data? Real life of the product? Test data? Regulation standard? Or perhaps QA measure?” There are so many questions I could keep asking but they all are valid questions to respond to the original question if you ask me.
Another question I have come across is: “Did I make any mistakes in my study?” This is the same general idea but slightly different flavor. The answer to this question would be: “I might be able to find some mistakes but only you (as you the engineer) are in a position to catch them all.” There are some cases where users make mistakes by applying twice the load or perhaps the mistake is in the material’s properties, thickness, etc. How can someone external know if it’s right or wrong?
Let’s imagine that we are talking about literature and not FEA or CFD. Imagine that you’re a proofreader for a publishing company and you are asked to proofread a biography for someone. You are asked if the person that wrote that biography made any mistakes. What kind of mistakes? There are certain ones that you could catch such as grammar, punctuation, misspells, etc. But other factors you might not be able to catch such as wrong names, dates, events, etc. How can someone that doesn’t know the life of the person catch all the incorrect biography data?
Sometimes the question of accuracy is not based on the methodologies or program implementation and capabilities, but with a particular problem in mind. This is good, but you need to make sure that the type of simulation you run corresponds to the mode of failure you are observing or expecting in reality. A linear analysis is only good as long as the underlying linearity assumptions are satisfied. A static simulation analyzing strength does not reveal much about a fatigue failure that analyzes life under cyclic loading. If you do not know which failure mode you have to pursue, you either should simulate all different modes of failure or do some experiments in the lab to gain a better understanding of which failure modes to rule in or out. If you ask someone: “Are there any mistakes?” only you, the engineer, know what type of failure you are solving for. Perhaps only you know what the correct boundary conditions and valid assumptions are that we could make.
There are also fundamental concepts that you need to understand in order to get accurate results. I think what people often forget when they are performing a linear static analysis, is that when they go beyond material yield strength anywhere in the design, a re-distribution of stress will occur which can only be captured by a nonlinear analysis. Let’s remember that linear static analysis assumes that the relationship between loads and the induced response is linear. Meaning, if you double the magnitude of loads, the response (displacements, strains, stresses, reaction forces, etc.), will also double. However, nonlinear analysis do not behave in this manner. In the real world most of the structures behave nonlinearly in one way or another at some level of loading. In some cases, linear analysis may be adequate, as long as you understand the assumptions and the materials do not go over the yield strength. In many other cases, the linear solution can produce erroneous results because the assumptions upon which it is based are violated. Nonlinearity can be caused by the material behavior, large displacements, and contact conditions; therefore it is important that you understand the behavior of your model.
In other instances when trying to replicate experimental data or actual response in the field with simulation, it is important to remember that the source of potential discrepancies is often in the setups. Therefore, it is recommended to carefully revise all the boundary conditions, material properties, etc. prior to running the analysis. The more accurate the quantitative results you are seeking, the more careful you should be with the setup of your model and interpretation of results.
The good news is if you setup your model correctly you can be confident that the results are very close approximations of the physical or theoretical values. Within the documents included in the Autodesk Simulation FEA and CFD products, there are validation problems that compare the numerical solution to selected problems with a closed-form analytical solution, established experimental results or solutions from other resources.
Autodesk Inventor Pro Simulation
For Autodesk Inventor Pro Simulation 2015, there is a built-in Convergence Plot option that helps users determine if their solutions has converged and what that convergence rate is.
Autodesk Nastran In-CAD
For Autodesk Nastran In-CAD 2015, the software contains a Verification Manual built-in within the software.
The Verification Manual is a 264 pages long document with very robust examples that compare the FEA results to theoretical calculations and physical test results.
Autodesk Simulation Mechanical
For Autodesk Simulation Mechanical 2015, there is a link to the Accuracy Verification Manual for Autodesk Simulation from the Autodesk Knowledge Network. Even the pre-built files are included for different versions. This document has 529 pages of examples with comparison to theoretical values.
Autodesk Simulation CFD
For Autodesk Simulation CFD 2015, there is a long list of verification examples that help validate results from the software. The verification examples are built-in directly in the online help of the software.
In the Autodesk Simulation CFD online help, there is also a Theoretical Background section with fluid flow definitions, mathematical concepts, governing equations, and several other resources vital for all simulation users.