I'm a fairly new user of PFPX but very familiar with everything it does and how to correctly fly the aircraft using the data from PFPX. Still, I am finding many differences between the estimated fuel values. 

I do have a real weather engine (ASN), I do put the wind data in the FMC (the one generated by PFPX online weather database, which closely matches to my ASN winds enroute), use the same cost index from PFPX, do step climbs as per the schedule. 

When I used the biases of 100% and drag of 100%, takeoff at the exact same fuel as per the log (e.g. if my taxi fuel was 500 kgs, I spent 450-550 kgs), and my fuel value slowly goes incorrect as I progress with the flight. On ultra long flights I land with as much as 10 tons more fuel than scheduled. 

So, I decided to change the bias values with PFPX evaluator... Climbed to the optimum altitude, and recorded the values. I've came up with the following numbers - 95.6% climb, 95.6% cruise, 55,7% descend, and 100.8% drag. I am now doing a 3000nm flight with those values right now, close to top of descend, and the estimated fuel is PERFECTLY (within 0.1 tons, as low as the plane can tell) matching the flight log. But still, in my FMC on the IDENT page, the, DRAG/FF values are +0.0/+0.0, so something is wrong, either PFPX or the 777. 


Another question which bothers me are the mach/TAS/elapsed time values which are off. Not by much, but they are off. So, I have the correct weight as the flight log shows, same static air temperature,  altitude of course, yet my mach varies by about 0.005 and TAS varies within 3 knots. For instance now, the flight log shows 0.831, and my plane is doing 0.835 at this moment of writing, TAS varies by 2 knots. 

I dunno man all I notice is that not everyone will have the same results using similar bias values. For whatever reason even if the weather matches perfectly between 5 simmers, the bias values will likely be different. Just something I have observed with people using the 777. Someone from pmdg said that a bias of 104% is perfect for the 777, I don't know if it really is accurate or they were being conservative.

I can't even imagine what would happen with default 104% bias from the PFPX template. With 100%, I was as high as 10 tons off from what it calculated.

Aviation is a highly variable beast.  It deals with two hugely variable problems (and a third that's just a generic problem - not "variable"):

1. People
2. Weather
3. The Unknown

People:

People are highly variable and rather unpredictable.  How much do they actually weigh (reference to my SYSK - Trim video there)?  How closely will the crew follow the SOP?  How much hand flying will they be doing?  Will they take advantage of direct-to, when offered?  How much will the crew employ fuel saving measures?

 

That is to say that we can't plan 100% accurately because one crew will behave differently from another, even if only slightly.

 

Weather:

Weather also has the potential to be highly variable, and despite being somewhat predictable, it's an imperfect science.  Yes, we do have forecast certainty, and metrics on weather stability, but again, we don't know for sure.  There's a reason there's an Actual column on flight plan paperwork - we know that it's going to be off.  Despite what it may seem in the sim realm, upper atmosphere wind data really isn't truly understood and accurate (note the giant gaps in the altitudes for the data).  PIREPs exist because weather forecasting and observing on the ground is incomplete.  We need eyes in the sky to correct and verify that data.  In fact, one of the projects I worked on (NCR - NAS Common Reference) included a whole segment on using aircraft as observation stations in the sky to downlink weather data to make forecasts and observations more accurate and complete.

 

That is to say, even with the heaps and heaps of data we do have, there are large gaps, and it's not 100% accurate.  If it's not 100% accurate, you can't plan your fuel 100% accurately.

 

The Unknown:

When Boeing released the 777 on day 1, where do you think the data for fuel burn and performance came from?  Mathematical simulations, and the relatively limited lifespan of the aircraft in testing.  Aircraft aren't born with perfect data sets available to the manufacturer.  As with data collection of any sort, you're going to get a more accurate picture by observing over time and making the appropriate adjustment.  Why else would Boeing include a scalar for FF and Drag on the very system that they created.  If it was, in fact perfect, why would you give the end operator the ability to adjust scalars for it to be "more perfect?"

 

All of this is to say that we don't know what we don't know.  Boeing can't test the aircraft for 10 years before releasing it to the operators, so their data is not going to be as accurate as it could be after running it for years and years.  Airlines do exactly what you did: they look at the data that came with the plane, operate the plane, and then evaluate the data against the control data.  From there, they make adjustments to the FF/Drag scalars, and their own flight planning software to better match what they're seeing in the data.

 

 

 

Honestly, I'd consider it a feature...

Welcome to the life of an aircraft operator.

If it matches perfectly, then something is wrong.

 

Nothing wrong with ending up with MORE fuel at the end. It's when you end up with LESS that there is a problem.

 

If you're flying with real-world weather then all bets are off. If the actual is much different from the forecast, and you end up with stronger tailwinds/weaker headwinds then the remaining fuel will increase, as it should. Conversely, weaker tailwinds/stronger headwinds will result in less fuel, and if necessary, a divert. Experience will tell you which routes need more additional fuel. Airlines have statistical data to use as well as "route experience" in tuning their fuel requirements.

 

I tune the figures in PFPX slightly, but I'm not obsessive about it, as aircraft are not perfect, and in reality, actual weather doesn't match the forecast, especially oceanic or in remote regions. In fact, the forecasts can be so off, that pilots actually talk on discrete frequencies between aircraft when in oceanic/remote regions, especially if there is rough turbulence/bad weather, as the forecast could be 6-12 hours old by then.

 

Best regards,

Robin.

Kyle, 

First of all, thank you for the extensive reply. I fully agree with everything what you have said, although the factors you mention aren't exactly applicable in the simulator, at least in my opinion:

- People:

I would say the part regarding the trim fully applies to real life, but not in the simulator. Since the add-on should be able to know 100% precisely what are the weights which it sets, but if the PMDG made it so there's a random error same as in real life, then it does apply of course. 
I  do not remember using any directs as there was very little ATC coverage (VATSIM) enroute. I did some more hand flying than usual (e.g. turned on the autopilot at 10000 feet, turned it off at glideslope intercept), but how exactly will this significantly affect the fuel?


- Weather: 

As you have said, the weather in real life does vary, sometimes significantly from the predicted one. However, my opinion is that while we are using a weather engine such as Active Sky Next, this variance is not as pronounced, since ASN actually uses the data we have, not the real one (which is where the variance comes from)!

And, as I have mentioned, the winds were indeed matching very closely to what was predicted (+/- 10 degrees and +/- 10 knots mostly). And it is unrealistic for it to be so accurate, but I guess that it's a "simmism", as you say  .

- The unknown: 

If PMDG and PFPX use the 777 from the data provided by Boeing, even if it's wrong compared to the real life, shouldn't it be right from the simulator?

Of course, differences from the predicted values are realistic, but should they happen in the simulator? Always ending up with exactly 10 tons more is definitely unrealistic.

Once again, let me just mention that I am just stating my opinion.

 

 

I would say the part regarding the trim fully applies to real life, but not in the simulator. Since the add-on should be able to know 100% precisely what are the weights which it sets, but if the PMDG made it so there's a random error same as in real life, then it does apply of course.
I do not remember using any directs as there was very little ATC coverage (VATSIM) enroute. I did some more hand flying than usual (e.g. turned on the autopilot at 10000 feet, turned it off at glideslope intercept), but how exactly will this significantly affect the fuel?

 

While you're correct on the weight issue, I was really only using it as an identifier of the many ways in which humans are a huge variable.  My main point was that humans do not behave exactly alike.

 

As far as your point goes about hand flying affecting fuel, have a look at this (ignore what appears to be a turnback and a false origin of OKV - I was only on flight following, and FlightAware tried to make sense of partial data from two flights):

 

Despite having a G1000 in the plane, which has some pretty advanced navigation/flight aids, to include a flight path vector indication on the PFD to help negate wind drift, you can see that my hand flown track was not the best.  Being a single pilot, I had to divert my attention away from maintaining my track from time to time, which caused slight turns, and resultant corrections.  Overall, it's pretty good, but we, as humans, really aren't as precise as you'd think.  If you were to compare this to something that wasn't hand flown (particularly if it was IRS/GPS, high precision like most airliners), you'd see a lot less variance.  Less variance means less fuel.  Granted, that won't account for 10 tons, but variance is variance.

 

PFPX simply can't plan for human variance.  No planner can.  That's why we have things like Contingency Fuel (one of the reasons, anyway).  Vectored departures and arrivals can also significantly affect your plan, too, so there's a significant grey area in the plan.  Have a look at a flight out of DCA to ORD.  If it's in north ops (departing Runway 1), you're going to be on a heading up the Potomac River for a significant amount of time until you can climb above the airspace used for Dulles (IAD) before you can turn towards AML, and onto your route.  That's about 25nm versus the 11nm direct path the planner would use on the route (example).  Again, not 10 tons worth, but when you add vectoring on departure, vectoring on arrival, and human control variance, you begin to see why the numbers err on the side of caution.

 

Then combine this with weather:

 

 

 

As you have said, the weather in real life does vary, sometimes significantly from the predicted one. However, my opinion is that while we are using a weather engine such as Active Sky Next, this variance is not as pronounced, since ASN actually uses the data we have, not the real one (which is where the variance comes from)!

And, as I have mentioned, the winds were indeed matching very closely to what was predicted (+/- 10 degrees and +/- 10 knots mostly). And it is unrealistic for it to be so accurate, but I guess that it's a "simmism", as you say .

 

True.  Again, it's a lot more variable in the real world than in the sim, but also remember this:

Is PFPX using ASN for weather?  Nope.  ASN's weather model will differ from AS2012's, FSX's own internal system, REX's, and so on.  This occurs even in the real world.  NOAA claimed the DC area would get 1-3 inches of snow today, while the Capital Weather Gang was quoted at 0.5-1.5 inches.  The reason is that they're using different methods to combine existing data, model the current situation, and forecast the future situation.  The data they got was all likely the same data (NOAA data at its base, at the very least).  Similarly, each FS weather add-on weather program attempts to fill in the holes in the weather, and depending on your depiction mode (depending on the add-on, some have this feature), it will model the current situation differently even within the same program.

 

PFPX doesn't include a weather engine.  It likely fills in gaps in data through simple interpolation.  It's not trying to depict anything, or make the simulated weather believable through interpretation; rather, it's trying to figure out the current weather picture, and likely relies on external forecasts for forecasted weather (which is subject to the relevant real world agency's forecast model).  Again, there's variance between the weather program, and PFPX.  Variance there means variance in the fuel.  This won't account for 10 tons, but you see we're continuing to add variance.

 

Then add in the fact that we don't know what we don't know:

 

 

 

If PMDG and PFPX use the 777 from the data provided by Boeing, even if it's wrong compared to the real life, shouldn't it be right from the simulator?

 

No.  Remember that Boeing provides its own data, which may or may not be time-tested as it has been in the airlines.  People can get really edgy about data.  Airlines have collected all kinds of data on performance over the years and don't often share it.  It could be seen as a competitive advantage.  If my data collection has shown me Lessons A, B, and C, but the other guy's data collection might not be as good, I'm not going to give him my data because he could also learn Lessons A, B, and C.  With that knowledge, he could negate my competitive advantage.

 

Furthermore, PFPX may or may not be using Boeing's data directly.  It could be augmented by other data sets like the BADA Performance Data.  I can't say for sure because I didn't do that work.  If there are performance tables, how is the data interpolated?  It is linear, or is it somehow a line of best fit?  Is this different from PMDG's interpretation and management of the data?

 

 

 

Of course, differences from the predicted values are realistic, but should they happen in the simulator? Always ending up with exactly 10 tons more is definitely unrealistic.

 

Any time I have cases like that, I start to look inwardly first.  What am I doing that's causing this?  The most common error I find with PFPX and fuel "discrepancies" is that people are planning flights using constant Mach cruises (the default selection, last I checked), but are using Cost Indexes in the plane (you can enter a CI in the Cruise Speed box).  This could account for significant error.

 

As I've repeated several times already, it's all about consistency.

 

The picture you paint in PFPX has to match the picture you paint in FSX.

• If you plan in PFPX using a constant Mach cruise, you'd better use that constant Mach cruise in the plane
• If you plan in PFPX using a CI, you'd better you use that CI in the plane
• If you plan a certain altitude (or OPT profile using steps) in PFPX, you'd better use that in the plane

That list could go on for quite some time, and those are only the items you can control.  Again, you're going to get bad values if you give PFPX a different version of what you actually do in the sim.

 

Now, add the human, weather, and unknown factor in all of that (all of the factors you can't control) and you can get a significant amount of variance.  It's up to you to both behave in accordance with your plan, and adjust the biases for said plans to match the kind of flying you do.  Just like a real aircraft operator.

Kyle, I immensely appreciate your answer, and entirely agree with it. 

Though, what I should have stressed out more is the fact that in the end, I always end up with ~10 tons more than what was predicted (or whatever is proportionate to the length of the flight), and that's generally what bothers me the most. Then, I adjust the bias value to be more correct to the fuel, but then TAS/mach values are wrong, and always by the same amount, within a very small margin, which is covered up by many variables we discussed about.

EDIT: The reason I think these values are wrong is due to an incorrect cost index range. PFPX shows the maximum for the 777 is 999, but it's actually 9999. Could this be the reason?

 

 

EDIT: The reason I think these values are wrong is due to an incorrect cost index range. PFPX shows the maximum for the 777 is 999, but it's actually 9999. Could this be the reason?

 

Could be - like I said, it's all about the variances between expected and actual.  If it's always about 10 tons, then it's likely a variable you're in control of.

I find that PFPX fuel predictions are extremely accurate for the 777, with no bias editing necessary. For me, it usually varies no more than 300 kilos either way. I use weather data from ASN and a cost index of 35 for my fuel calculations as well as in the sim. Perhaps try this method and see if you get better results?

 

Regards.

I don't know, how long are your flights usually? I do some long hauls quite often so the value gets off more. I used the value of 10 tons for a flight from Dubai to San Francisco (17 hours), but of course with shorter flights it's not as much.

I've done ULH flights before, the variation is never as high as that. Even on those, the maximum variation is no more than 700-800 kg. My last ULH was Dubai to Seattle, using ASN weather settings. If I recall correctly, I landed within one ton of the predicted fuel to remain.

 

Regards.