I don't think you can fairly 'calibrate' the software, as that would require that you could compare it against some known standards (measurement standards - pieces of higher accuracy test equipment with certificates of calibration traceable to national standards, and so forth).
The remaining possibility would seem to be 'validation'. To validate the software, you would need known good parameters and known bad parameters to test the softwares ability to discern between pass and fail conditions.
The other possibility might be to get some sort of certification or documentation from the software vendor. If they are ISO or QS9000 certified, they may keep documentation as to how they validated the software. Those are just some preliminary ideas.
------------------