I wonder how accurate are flight simulator scenarios compared to real flying emergency situations pilots are facing and also how they capturing the data which they input into the flight simulators.

Quoting BUGYUL (Thread starter):

I guess you mean the simulators we pilots are trained on?

They are very realistic. You sit in a real cockpit with the real instrument. Of course the outside view is not as the one in real life.

The step from the simulator to the real airplane is very small. I felt very comfortable and well prepared when I did my first flights in 737, MD11 or 747 after the simulator hours.

wilco737
  

Just to piggyback on this thread: Which scenarios cannot be trained in a sim with any accuracy?

For example, for full stalls the data is lacking - we've discussed that at lengths in another thread. From another thread, I've picked up that pilots have some troubles landing the sim because they don't give a good depth perception.

How is it with rapid decompression, fire or smoke in the cockpit and things like that?



David

Quoting flyingturtle (Reply 2): Which scenarios cannot be trained in a sim with any accuracy?

There are a few: ditching is rather difficult as no data available. Partial gear up landings cannot be trained properly...

Quoting flyingturtle (Reply 2): How is it with rapid decompression, fire or smoke in the cockpit and things like that?

Rapid decompression is simulated, but of course not the fog, the bang, the cold, the pain in the ears, the lack of oxygen etc. So you basically train to put on the oxygen mask asap and then initiate the emergency descent.
Smoke in the cockpit is simulated with smoke in the cockpit. There is actually some very dense smoke in the cockpit that you cannot read the instruments properly. So you are under the oxygen mask again and try to land. This is rather good simulated as the smoke is really thick.

Landings feel a little different in the sim, but not too much. Of course a little of the outside reference is missing, but I felt very well prepared for my first landing in real life after doing many many in the sim.

wilco737
  

Quoting flyingturtle (Reply 2): Which scenarios cannot be trained in a sim with any accuracy? For example, for full stalls the data is lacking - we've discussed that at lengths in another thread. From another thread, I've picked up that pilots have some troubles landing the sim because they don't give a good depth perception.

Wilco did a great job addressing most of your questions, but we don't train full stalls in aircraft or sims; certainly the sims provide excellent fidelity (it wasn't always so) but the closer you get to the boundaries of the aircraft's performance (fully developed stalls, spins, etc.) the less accurate the fidelity will be. Of course much of that region isn't explored in airliner certification, either (e.g. spins) so we really have no idea how well the models would match reality. (Thankfully.)

Older sims suffered from depth-of-field landing issues more than newer sims. I have never had particular difficulty landing the sim or aircraft, but in the old Level B night-only sims, it was decidedly more difficult. Again I agree with Wilco737: the sim isn't the aircraft, but if you can safely land the sim you are well-prepared for flying the real thing.

Hand flying and visual approaches aren't quite the same because of many of the reasons mentioned but they are getting better on both visual and tactile fronts in regards to a similiar "handling". The newer ERJ-170 sims at FSI have electric servo motors for motion versus hydraulic systems and this is a dramatic improvement.

Visually the biggest place I notice a difference is perriferral (I more than likely spelled that wrong) visual cues. They don't exist. The projection screen stops perpendicular to the DV window in the two types I've trained in. It can actually make me a little naseous. Looking straight ahead it's quite accurate when you're not in zero visibility. In fact after a few minutes I forget it isn't the real thing.

That isn't really the point of simulator training. The point is to reinforce procedure and practice abnormalities. You can't quite address every situation but in my only aircraft-related emergency (an engine shut down) I felt like it was slow motion. I primarily attribute this to reinforcement of training and the speed at which things happen in the simulator. You're often dealing with more than one emergency in a very critical phase of flight and as soon as you have landed, you set up and start over again primarily because of time constraints. You're almost always seeing the engine failures at V1, every approach is to minimums or a missed approach and go arounds (which RARELY happens in the real world, maybe 5-10 times a year for me out of approximately 400 legs).

Quoting pilotpip (Reply 5): The newer ERJ-170 sims at FSI have electric servo motors for motion versus hydraulic systems and this is a dramatic improvement.

Electric motion is more or less standard on new sims these days. Apart from better response, they are quieter in operation and cheaper to install and operate.

Quoting BUGYUL (Thread starter): also how they capturing the data which they input into the flight simulators.

Data for flight simulators comes from aircraft flight tests, manufacturer's engineering data, wiring diagrams and maintenance manuals. There isn't much data for emergency simulation, apart from accident reports, so a degree of creativity is required by the designers. The malfunctions are designed around the Abnormal and Emergency Procedures in the Aircraft Operating Manual. From a systems point of view the modelling required simply replicates the failure concerned and then (hopefully) the secondary effects will occur naturally as a result of the simulation model. Engine and fire malfunctions often require additional modelling to drive indications as required.

Quoting BUGYUL (Thread starter): how they capturing the data which they input into the flight simulators

In the case of a Full Flight Simulator, basically you can divide it into several parts :
- flight physics
- aircraft systems
- "environment" : motion, visuals, sounds...
- packaging and interfaces (for example, how the instructor can control the simulation)

The last 2 parts are developed by the simulator manufacturer (CAE, Thales, Mechtronix for example). However the first 2 parts are the ones primarily involved in reproducing aircraft behavior

For the physics part, the aerodynamic performance/aircraft geometry/weight&balance models come from the aircraft manufacturer. They are usually the same models, or very closely related models, as those used during development. So they are based on data coming from advanced aerodynamic calculation tools, and used countless times on various engineering simulation means, regularly updated and finally validated with flight test data. In some cases the models contain the equations of the a mathematical model, which will be solved “live” while the simulation is running. This reproduces the aircraft behavior very well, but is heavy in terms of computing power and computing time (these sims have to work in real time). So the other possibility is to use tables of discrete data which is then extrapolated to the current operation conditions.

For systems, the sim manufacturer can either use real equipment (usually for displays units, and the FMS) or real-time models. As in the case of the physics models, the system models can be supplied by the aircraft manufacturer, based on the system models used for aircraft development (and so tried and tested as well). This would be the case of flight control (ATA27) models. Or the models can come directly from the equipment supplier, or the sim maker can create his own based on available documentation. This may be used for some systems like air systems (ATA21) which may just need to provide warnings. The exact mix depends on how good and especially how expensive you want the sim to be.
Actually the system simulation is by far the trickiest part of the simulator, because you have lots of licensing problems, you have to make sure real hardware and more or less detailed models work together, and so on. There's actually an ARINC standard for this (ARINC 610 ; but I have no good non-paying link with details).
Here’s a paper from the Royal Aero. Society – Flight Simulation group about those topics :
http://www.raes-fsg.org.uk/yabbattac...lator_Data_Challenges_RAeS_FSG.pdf

Now once you’ve gathered all the models to build your simulator, you have to make sure all these parts work together to reproduce the aircraft properly. For single systems you can use Acceptance Test Manuals (ATMs), but the main tool will be the Qualification Test Guide (QTG). The principle here is to establish a set of reference flight conditions from relevant points of the operations envelope, and gather data about the corresponding aircraft behavior. This can be from Flight Tests or from high-level engineering simulators. For each new sim, you then compare the sim behavior to the reference behavior (it’s a huuuge job !   ). A subjective assessment by a pilot is also carried out (actually 25 years ago the entire validation process was subjective : stick a pilot in the sim and ask him if it feels right !    )

And that's more or less how you do it !   

Here are some good links :
http://www.raes-fsg.org.uk/13/Past_Conferences/
en.wikipedia.org/wiki/Flight_Simulator (the wiki article is actually very good and detailed)

Quoting Jetlagged (Reply 6): then (hopefully) the secondary effects will occur naturally

Well hopefully...it isn't just "hopefully" that they occur or not !! 
Actually that will depend on the level of detail you need for your simulator, and so how detailed the models will be. In some cases you just need the warnings to show up on the displays, in others you might need to model how failures cascade from one system to the other so you'll need some depth in the simulation. it really depends on what you want to do with the simulation.
Which brings me to a last point, BTW, there aren't just Full Flight Sims out there, there's actually a whole range of products : simple desktop simulations, fixed-base simulators, 3D training devices. And you can use sims for flight training as well as for other training courses like maintenance ops training. The first ones will need accurate cockpit control&indicating behavior for "flight-related" systems (flight controls, hydraulics, electricity etc...), whereas the second will require detailed modeling of every system, including computer networks, central maintenance computer, and so on.
There are some good examples on the CAE or Thales websites...For example see here : http://www.cae.com/en/sim.products/cae.simfinity.asp

Many thanks to you all, especially airmagnac!

 

Quoting pilotpip (Reply 5): The newer ERJ-170 sims

Does this sim cover all EJet types? E70-75-90-95.

Also does the manufacturer load every airport in the world or only the ones the airline chooses or just a standard training set?

Quoting airmagnac (Reply 7): Well hopefully...it isn't just "hopefully" that they occur or not !! Actually that will depend on the level of detail you need for your simulator, and so how detailed the models will be. In some cases you just need the warnings to show up on the displays, in others you might need to model how failures cascade from one system to the other so you'll need some depth in the simulation. it really depends on what you want to do with the simulation.

"Hopefully" was just good old British understatement. What I meant was that if the modelling is in sufficient depth then there is no need to add additional modelling for secondary effects, they will be part of the systems modelling and the integration between different systems. In such a case, typical of an FFS or high level FTD, the malfunction flag only needs to be monitored in one place, the other systems and sub-systems react naturally to the failure. If the simulation is less deep then you have to add the secondary effects as required, each affected system checking the malfunction flag to switch the failure modelling in or out.

Quoting Jetlagged (Reply 10): "Hopefully" was just good old British understatement.

Mea culpa for taking the word of a Briton too literally.     

So we're saying the same thing : if you have sufficient detail of each system involved, and if they are sufficiently connected, then you just have to insert the fault once at the source point, and it will propagate "naturally" by itself.
If not, you have to insert the proper effects manually in each part of the plane that is involved. Which is a lot more work for the sim people to specify and validate...so any flight sim engineer would *hope* to avoid that, I guess  

Quoting airmagnac (Reply 11): Which is a lot more work for the sim people to specify and validate...so any flight sim engineer would *hope* to avoid that, I guess

Exactly. In my opinion it's usually easier to simulate something properly than take a simple "cause and effect" approach and then have to add all kinds of extra "cause and effect" stuff to make malfunctions work and correct the secondary effects of normal operations.

Indeed, I find that when you're trying to make a realistic simulation, making & integrating fully detailed models is actually much less difficult than making & integrating "simplified" ones.

Quoting wilco737 (Reply 1): The step from the simulator to the real airplane is very small. I felt very comfortable and well prepared when I did my first flights in 737, MD11 or 747 after the simulator hours.

How many hours do pilots spend in the sims before getting to tame the real beasts?   I realize this probably cannot be said in general, or am I off here?

Quoting bjorn14 (Reply 9): Also does the manufacturer load every airport in the world or only the ones the airline chooses or just a standard training set?

The 727 sims we use have a few major airports in them like MSP, JFK, and MEM. Some smaller airports are built in as well so that during LOFT sessions there are places to divert.

Quoting Harmonium (Reply 14): How many hours do pilots spend in the sims before getting to tame the real beasts? I realize this probably cannot be said in general, or am I off here?

I had somewhere between 20 and 25 hours in the sim when I received my Flight Engineer ticket. It took about the same time to upgrade to FO.

FlyMKG

Quoting bjorn14 (Reply 9): Also does the manufacturer load every airport in the world or only the ones the airline chooses or just a standard training set?

Simulated airports need two components, a custom modelled visual scene and a navigation station database. On older sims the navigation database size was limited to a few thousand ground stations. You need all the navaids around each airport plus enough enroute navaids for LOFT exercises. In practice this meant only a region could be loaded at any given time. Visuals were limited to a few customised airport scenes and one or more "generic" airports to use where no custom scene is available. Modern sims have no limitations on database size so all radio stations can be loaded for worldwide capability. Visual scenes are still limited to a number of customised models which are maintained to high accuracy plus a library of scenes which aren't necessarily up to date. Generic scenes still exist but can be modified easily by the user to get the correct runway length, positioning of buildings, etc. Modern visuals usually have a worldwide terrain database so the scenery around a non customised airport is still reasonably accurate. A far cry from the days where everything outside of the visual scene area was flat.

A full flight simulator must have at least three up to date visual scenes, which are the operator's choice. The FAA are moving towards a policy where only customised, up to date scenes will be acceptable for training, so simulator operators are forced to either keep all their library airports up to date or delete airports that aren't accurate.

Quoting Harmonium (Reply 14): How many hours do pilots spend in the sims before getting to tame the real beasts?   I realize this probably cannot be said in general, or am I off here?

When I changed to 744 it was around 60 hours in the Sim.

wilco737
  

Quoting wilco737 (Reply 17): When I changed to 744 it was around 60 hours in the Sim.

Having something like one ordinary takeoff and landing, and a bazillion engine fires, rapid decompressions etc?  

Quoting Harmonium (Reply 18): Having something like one ordinary takeoff and landing, and a bazillion engine fires, rapid decompressions etc?

No, there is one session where you do all kinds of landings, normal, crosswind etc etc. And of course manual flying. But yes, there are engine failures as well and of course many many other emergency (abnormal) conditions.

wilco737