I am just trying to get a little more detail, is the FEA critical software
to the overall process that has been validated. On a scale of low, medium high where does this software fall in terms of criticality.
I would think this software needs to be validated as per intended use, so if you are using this software for thermal, buckling, static state of stress, etc I would assume this aspect of the software needs to be validated to ensure it’s working correctly.
Is this software Part 11 compliant? does it need to be?
By critical do you mean required? There might be other more efficient ways to get the job done for instance (closed form equations, etc.). We would never take the answer from an FEA to accept any part of the design. We might use the output to guide decision making such as air flow required to meet a temperature spec. But the product is always tested to insure that the temperature requirement is met. So I would call it low, but that’s my engineering opinion not the opinion of a person trained in FDA compliance.
If I tested 10, 20 or 30 functions of the software I don’t believe that would insure that the 31st function (or the 10,000th) would work correctly. Every commercial software has bugs as we all know and many of them get discovered and reported by users. We are using software that’s been used in many industries for decades from firms that have hundreds of programmers that do nothing but V&V. They’re at the top of FEA and analysis programs. Many other engineers have insisted that to validate that they are working correctly (completely) is virtually an impossible task. So if I validate 10 functions does that mean users can’t use any of the others?
I’m a bit confused with the word “correctly” and confused if that implies accurately too. FEA is just an approximation of the real world and so accuracy is a relative term. I’ve learned the following from a paper by the ASME: The verification part is usually supplied in the form of industry accepted problems to verify that the math in the FEA program is correct. The validation (of the physics) is done on a case-by-case basis by the person doing the analysis by performing bench experiments, hand calculations, etc.
The approach I’ve taken is simple and rudimentary and I question what it proves. Does it import models from the solid modeler? Does it mesh the model? Can you define boundary conditions and material properties.? Will it solve the model? Can you view the results and extract numeric information? Can the results be saved and retrieved? If the software did not do any of this we would never have purchased it in the first place and the company would not be in business selling it.
The electronic output from the FEA is captured in a PDM program which then can (probably always should) become part of the DHF. Users are formerly trained by the vendor, which doesn’t necessarily make them skilled users. Only years of experience and mentoring will do that. And even the mentors are continually learning. There aren’t any controls to lock out inexperienced users nor is there an SOP on how to do FEA. So we might have a ways to go on part 11 compliance as I understand it given all of this. Sorry for the length of this response.