Originally Posted by
AnthonyGA
How much of the flight envelope is usually covered by hard test data, and how much has to be simulated through extrapolation?
I've heard about simulators sometimes not being very good at extrapolating to flight regimes that weren't part of the tests. Wasn't there an accident involving extreme use of rudder controls or something some years ago that was traced in part to a difference between the way the sim behaved and the way the actual aircraft behaved.
The minimal requirements for test data validation for a simulator are relatively sparse;
FAA Advisory Circular AC120-40B "Airplane Simulator Qualification" is one example. (The regulations are pretty standard these days).
However, in practice there's a lot more than the bare minimum done. The simulator has to pass qualitative as well as quantitative tests, so the data has to be reasonable across the entire envelope, not just the specifics in the regs. And in practice many training simulator flight models start out as engineering simulation models in the respective OEMs, being used for design studies and for certification analysis, and so subject to further validation and oversight by the authorities. (A good example would be the use of simulation models for VMC extrapolation, or to provide loads data for extreme high speed manoeuvres; in both cases the certifying authority will expect a certain amount of validation specific to the conditions).
Similar considerations apply to systems simulation ; where systems are simulated (rather than "stimulated", which is when actuakl a/c hardware is used) the system models will trace their development back to various systems rigs and design tools.
The bottom line is that if you do something with a fully qualified simulator which is something previously done during the development or certification testing, or which is pretty similar, then the sim should be pretty well representative. Go outside those parameters, though - whether by exceeding the tested flight envelope or introducing combinations of failures never considered in design - and the sim becomes progressively less reliable as a guide to actual aircraft behaviour.
The case you are alluding to in the second para is I believe the use of the simulators in the AA upset training, which was discussed as part of the AA587 accident investigation. IIRC, one issue was that, in order to train the upset conditions, the sim was programmed - arbitrarily, IIRC - to be unresponsive to rudder inputs in certain flight conditions. One conclusion was that this may inadvertently have led the pilot to believe that the rudder was not just ineffective, but also benign.