PFPX profile for the 747-8

[downscc 1,327]

2 hours ago, Skyrock said:

I don't need PFPX for this to see that the fuel burn is the same. Don't reduce it only to this. PFPX only provides the bias, but the fuel burn my test results is always the same.

That got my attention so I loaded up the 777-200LR and set up a cruise burn test at FL300.  I used ASP4 to set up a standard atmosphere day where at the geopotential altitude of 30000 ft the standard temperature is -47.8F / -44.4C but I rounded the standard to -44 C.  Then setting up payload and fuel such that aircraft started the test period at 700.0 klbs GW I then measured fuel burn for 5 min at 0, +10 and -10C deviation.

The results show some variation, of course there were corresponding changes in N1 and TAS but I kept her at a constant Mach of 0.832 where at 0C deviation I measured 20,484 lb/hr, +10C 20,472 lb/hr and -10C 20,592 lb/hr.  I think this illustrates that fuel burn varies with SAT (and we're ignoring the change in TAS and impact on fuel burn).

However, note that the differences are generally less than a percentage point and this highlights the problem with sampling that Gauss discovered when he was surveying.  I don't like using 5 min, I think it increases variance too much and you were using just 1 minute.  The best way to measure an hourly rate of fuel burn is to measure the actual burn for an hour, and then measure it multiple times and then take the RMS value of all the measurements. Then one has to calculate the confidence level of the results and that might mean your carefully RMS value is much less certain than hoped for.  I went down this rabbit hole last year when I wanted to improved the PFPX data but I simply ran out of patience, it was taking too long to acquire the data.

Looking back on my results I think it may have been better to conduct the tests at a much higher altitude, where the logarithmic decline of air density would have amplified the results and made it clearer... but again, I've run out of time and I need to go fix lunch now.

[Skyrock 4]

Finally an answer I can work with.

I don't think a longer interval will end up in better results since the weight will also decrease by 7-8t which in turn has an influence on the fuel burn. Besides the weight, there is no other influence like changing winds, temperatures etc. So if you don't touch anything, I don't see a reason for increased variance other than a pretty nervous A/T, but that is a different story and that issue has improved quite a bit in the 747 compared to the 777.

Basically you proved my point that there is hardly a change in the fuel burn and I doubt that this is as it should be. And again, with the adjusted profile, it has never been as accurate as before I got rid of the ISA deviations. At least there is a correlation.

Maybe, as I already mentioned, this is a general limitation of P3D/FSX and not related to PMDG aircraft only. Even if it's only an approximation, we need to tell PFPX to not take ISA deviations into account, but that's hardly possible with .per files.

[downscc 1,327]

Granted there was a small change in fuel burn, and it could have been a wash given the imprecision of our methods.  The CF6-80C2B1F for which I have some data uses a 2% change in N1 for every 10 deg change in SAT with a CI of  100.  That isn't going to translate into much of a fuel burn difference.  In fact, the effect of temperature is primarily seen in the change in TAS for a given Mach speed.  This translates into time and that affects total fuel burn.

There are lots of issue with the simulator's simple approach to engine dynamics.  This is why all of that is done by PMDG outside of the platform in their own dll module, and they have been taking this approach for a long time and are good at it.

I come back to our measurement methods, which I maintain are a primary source of error.  For example, your B77L burning 20,484 lbs/hr is going through 5.69 lb/sec and when you time something for 60 sec the odds of you being able to time exactly 60.0 sec are nil (our reaction times are longer than a second). This is why I measure over a long period, sure there is a change in weight but I take the average of the beginning and ending weight and presume that the fuel consumption curve compared to weight over this very small span is linear.  When I was trying to set up a set of performance tables for PFPX last year I ran a number of trials and found that the standard error between my measurements were as high as 3%.  That's when I became frustrated with the effort and moved on to something else.

I actually don't see anything wrong with using std temperature for long range flight planning.  I notice none of the tables in the Performance Dispatch section of the FCOM include SAT deviation. 

[VHOJT 156]

It is interesting to see that you both have found ISA deviation has had basically no effect on the aircraft's fuel burn.

I primarily fly with the RR 744 - the profile for PFPX was created by Doug Snow with accurate data from his airline's performance dispatch system.  I get extremely precise results as long as I cancel out any corrections for ISA deviation that are applied by PFPX.  i.e. negate the 3% bias per 10 degrees +/- ISA (and of course corrent for the PMDG's ISA 0 baseline bias, which is negative bias).  

We seem to agree that something in the P3D and/or PMDG causes the aircraft not to respond in fuel burn to different ISA deviations?

A word from someone on the dev team or who really knows about how the fuel burn modelling works might clear it up 🙂