Gill fuel flow sensor failure modes

[Edis]

I have a question to the more knowledgeable.
I haven't seen and I don't know the internal design of an F1 car fuel system. But I assume they should have some kind of fuel recirculation, like on ordinary cars. Excess fuel from the feeder pump to the rail is returned back and one of the reasons is to prevent fuel getting too hot.
On the other hand the rules postulate that the fuel flow sensor is in the fuel tank. Up to now I thought the fuel sensor is somewhere after the high pressure pump.
If there is a return then how do they account for it?
Is the HP pump in the fuel tank too?

A lot of "ordinary cars" lack a fuel return; returnless systems have been around for quite some time and can either use a fuel pressure regulator in the tank (in this case before the fuel flow meter) or use a on demand electric pump under control by the ECU.

The high pressure pump is most likely a "flow on demand" piston type pump placed on the cam covers. The piston is actuated by a lobe on the camshaft, and the flow from the pump is controlled by a solenoid valve in the pump. The pump and its placement has been clearly visible on some of the engine photos released by the engine manufacturers and appear to be very similar to regular production units. On the Renault engine it seems to be driven by the intake camshafts, and there is one pump per bank.

The fuel pressure supplied to the high pressure pump is probably not more than a few bar.

I'd love to see evidence of an ECU calculating the fuel flow over a short enough period of time to be considered instantaneous with an accuracy of 1% as is the one stated by Gill and without any correction factor programmed into it. It boils down to the question: if you put a 1ms pulse over a 10cc/second injector, do you get exactly 0.01cc of fuel each time?

What I bet is that, all the engine manufacturers have correction factors on the simple math that is SUM(injector_pulse_width * injector_size) which yield, for that engine a more accurate fuel rate value. The FIA needed the same measure tool for all, thus the the flow sensor. If every team gets 10 of them, tests them and uses the one which they like most (reads the lowest) then it's a level playing field.

No, you will not get exactly 0,01 cc of fuel every time from such an injector, and if you double the pulse to 2 ms you would most likely get more than twice the flow than at 1 ms.

For an injector to actually deliver it's full flow, the injector needle have to reach full lift, and the shorter the pulse, the less time spent at full lift. How fast the injector reach full lift will also vary.

Among some production diesels, the ECU is actually calibrated for the individual injectors used on the engine, just to give a good enough calculated of injected fuel mass and injector timing.

It is direct injection so 1 ms is a long time, the whole compression stroke at 10krpm is 12ms
the flow through the injector will depend on the differential pressure, so will opening and closing times and it direct
injection so it isn't just fuel pressure vs. ~1bar like manifold injection it is fuel pressure vs. cylinder pressure
opening time will also depends on the voltage driving the injectors.

And I don't see why the teams, other than for the flow restriction, need an absolute flow flow number, they need
injections to be consistent so the engine can be tuned to the correct mixture

At higher loads you typically start the injection when the exhaust valve have closed, the fuel injection is then stopped at about BTC. So roughly 180 crankshaft degrees.

At lower loads you typically reduce fuel rail pressure.

I don't know if any of the current F1 engines have made use of stratified injection mode, but any such mode (and the corresponding late and short fuel injections) are limited to low loads and most likely lower speeds.

[Tommy Cookers]

At higher loads you typically start the injection when the exhaust valve have closed, the fuel injection is then stopped at about BTC. So roughly 180 crankshaft degrees.
At lower loads you typically reduce fuel rail pressure.
I don't know if any of the current F1 engines have made use of stratified injection mode, but any such mode (and the corresponding late and short fuel injections) are limited to low loads and most likely lower speeds.

isn't this saying that 2014 F1 engines have all the CR they want with all the boost they want .....
without needing this to be helped by their 400-500 bar piezoelectric injectors injecting the fuel in around 1 millisec or less ?
(ie starting injection near the end of compression to allow higher CR without detonation)

we seem to have pump-on-demand before our fuel meter
and we seem to have a (very) high pressure pump-on-demand supply for the injectors
this can cause brief excursions up or down from the nominal 100 kg/hr fuel rate though the flowmetered rate never exceeds 100
eg so that the engine can run through the gears 10500-12300 rpm without changing mixture strength and/or boost
the injected rate eg being exactly 100 kg/hr only as the rpm swings through 11400
acceleration being a race-efficient way to use the engine, surely they want to do this
typically only about +- 1.5 gm fuel 'float' is needed
and the rules 'they' wrote seem not to prevent this (innocent/inevitable 'accumulation' effect in the line after the flowmeter)
though dedicated accumulation devices are banned of course

[langwadt]

I have a question to the more knowledgeable.
I haven't seen and I don't know the internal design of an F1 car fuel system. But I assume they should have some kind of fuel recirculation, like on ordinary cars. Excess fuel from the feeder pump to the rail is returned back and one of the reasons is to prevent fuel getting too hot.
On the other hand the rules postulate that the fuel flow sensor is in the fuel tank. Up to now I thought the fuel sensor is somewhere after the high pressure pump.
If there is a return then how do they account for it?
Is the HP pump in the fuel tank too?

A lot of "ordinary cars" lack a fuel return; returnless systems have been around for quite some time and can either use a fuel pressure regulator in the tank (in this case before the fuel flow meter) or use a on demand electric pump under control by the ECU.

The high pressure pump is most likely a "flow on demand" piston type pump placed on the cam covers. The piston is actuated by a lobe on the camshaft, and the flow from the pump is controlled by a solenoid valve in the pump. The pump and its placement has been clearly visible on some of the engine photos released by the engine manufacturers and appear to be very similar to regular production units. On the Renault engine it seems to be driven by the intake camshafts, and there is one pump per bank.

The fuel pressure supplied to the high pressure pump is probably not more than a few bar.

I'd love to see evidence of an ECU calculating the fuel flow over a short enough period of time to be considered instantaneous with an accuracy of 1% as is the one stated by Gill and without any correction factor programmed into it. It boils down to the question: if you put a 1ms pulse over a 10cc/second injector, do you get exactly 0.01cc of fuel each time?

What I bet is that, all the engine manufacturers have correction factors on the simple math that is SUM(injector_pulse_width * injector_size) which yield, for that engine a more accurate fuel rate value. The FIA needed the same measure tool for all, thus the the flow sensor. If every team gets 10 of them, tests them and uses the one which they like most (reads the lowest) then it's a level playing field.

No, you will not get exactly 0,01 cc of fuel every time from such an injector, and if you double the pulse to 2 ms you would most likely get more than twice the flow than at 1 ms.

For an injector to actually deliver it's full flow, the injector needle have to reach full lift, and the shorter the pulse, the less time spent at full lift. How fast the injector reach full lift will also vary.

Among some production diesels, the ECU is actually calibrated for the individual injectors used on the engine, just to give a good enough calculated of injected fuel mass and injector timing.

It is direct injection so 1 ms is a long time, the whole compression stroke at 10krpm is 12ms
the flow through the injector will depend on the differential pressure, so will opening and closing times and it direct
injection so it isn't just fuel pressure vs. ~1bar like manifold injection it is fuel pressure vs. cylinder pressure
opening time will also depends on the voltage driving the injectors.

And I don't see why the teams, other than for the flow restriction, need an absolute flow flow number, they need
injections to be consistent so the engine can be tuned to the correct mixture

At higher loads you typically start the injection when the exhaust valve have closed, the fuel injection is then stopped at about BTC. So roughly 180 crankshaft degrees.

At lower loads you typically reduce fuel rail pressure.

I don't know if any of the current F1 engines have made use of stratified injection mode, but any such mode (and the corresponding late and short fuel injections) are limited to low loads and most likely lower speeds.

I don't know how they do it but F1 engines doesn't spend much time at low loads or low speed, so why bother with 500bar
DI if they were going to inject pretty much like if the injectors were in the manifold?

[jz11]

The one thing I don't understand is the people who say " He finished on 100kg , the race was more than an hour so he was under 100kg/h. Do you people not get the idea of an instantaneous flow rate? The sampling rate is 5 hz, so, every . 2 seconds they take an average flow rate for that .2 seconds and that average flow must be under 100kg/h. Got it?

Hate to be splitting hairs but this is a common misunderstanding. The sample rate isnt 5hz. Its sampled at 1kHz and then output onto the CAD bus at 100Hz. The FIA then apply a 5Hz filter on the data. This is absolutely not the same as a sampling rate of 5Hz. A sampling rate of 5Hz means that no data above 2.5Hz can be in the signal.

The actual situation is that the data is a 100Hz, so it can contain frequencies of upto 50Hz. It is then filtered from 5Hz up - but filtering is not a perfect cut off. There will still be frequencies greater than 5Hz in the signal, they will just be heavily attenuated.

took this from the other topic since it is too technical to be discussed there

bear with me on this, my technical English is not the best, so it might be difficult to understand what I meant, but I did bounce these ideas off some very smart people in physics and electronic I know (means nothing here, but I was simply checking if I didn't misjudge the whole situation) and the conclusion was the same - you should not use ultrasound for this type of application

all this is speculation, since I don't know how the actual sensor does it sampling, not even which type of US flow sensor it really is (might be hybrid even), error checking, averaging and so on, but here goes

my thoughts about the sample rate and bandwidth you mention don't quite come together, since sample rate may have little to do with the actual sample length, both transient and doppler type US sensors need to send short pulses and listen for them, determine either time difference between them or measure received frequency change, there is no flood-signal (I'm not sure how to call it English)

each sample, which is taken either each 1ms or 0,5ms, may consist of a pattern of short pulses of some frequency (and this time it is ultrasonic range sound), and this pattern is what lets the sensor distinguish its own signals from the noise in the medium, if the received pattern is too distorted - the sample is discarded, if it matches pattern and matches nearest neighboring samples in the value - it may be passed on (to CAN BUS, now at the rate 1 sample each 10ms)

this is where pressure shock waves and other "noise" from fuel pumps and injectors opening and closing come in, they generate some sort of sound waves in that volume, and they don't go away, because the flow meter is mounted in a closed loop, so those waves are going back and forth inside the fluid, and each time engine revolutions change - different frequency sound waves may be "generated" inside that same volume, and if the waves just happen to match, you might get some very high amplitude spikes in that volume, and this may very well happen with this sensor, and because of the nature of the ever changing noise in this volume, you cannot really adjust the filtering logic in the sensor itself

so if you just happen to have some resonance is the fluid slower speed than the sample length, then this wave will distort the reading you get from that particular sample sent by the flow meter - if you get on the positive slope - the resonance will be combined with the flow itself - thus getting a high reading of the flow, and if you happen to sample when the wave is on down slope - it will make your reading say that the flow was slower than it really was

these are the problems related to using ultrasound in this type of environment for high resolution sampling, the solution might be actually ok for a very long monitoring session (lets say - control the amount of fuel used during the race), because the average will eventually even out somewhat, but in my opinion, it cannot be used to accurately monitor "instant" fuel flow, and instant is quotes because there is no definition of this instant in the rule book

you can solve this issue on the engine dyno stand when you map the engine, you can weight the fuel you supply to the engine and figure out your fuel flow to pretty high degree, and you can simulate any lap on any circuit for any engine, and have a model of your fuel flow made - this is what I think of when I say - injection time based fuel flow measurement, lock those engine maps and be done with this whole "issue"

[chip engineer]

my thoughts about the sample rate and bandwidth you mention don't quite come together, since sample rate may have little to do with the actual sample length, both transient and doppler type US sensors need to send short pulses and listen for them, determine either time difference between them or measure received frequency change, there is no flood-signal (I'm not sure how to call it English)

each sample, which is taken either each 1ms or 0,5ms, may consist of a pattern of short pulses of some frequency (and this time it is ultrasonic range sound), and this pattern is what lets the sensor distinguish its own signals from the noise in the medium, if the received pattern is too distorted - the sample is discarded, if it matches pattern and matches nearest neighboring samples in the value - it may be passed on (to CAN BUS, now at the rate 1 sample each 10ms)

this is where pressure shock waves and other "noise" from fuel pumps and injectors opening and closing come in, they generate some sort of sound waves in that volume, and they don't go away, because the flow meter is mounted in a closed loop, so those waves are going back and forth inside the fluid, and each time engine revolutions change - different frequency sound waves may be "generated" inside that same volume, and if the waves just happen to match, you might get some very high amplitude spikes in that volume, and this may very well happen with this sensor, and because of the nature of the ever changing noise in this volume, you cannot really adjust the filtering logic in the sensor itself

so if you just happen to have some resonance is the fluid slower speed than the sample length, then this wave will distort the reading you get from that particular sample sent by the flow meter - if you get on the positive slope - the resonance will be combined with the flow itself - thus getting a high reading of the flow, and if you happen to sample when the wave is on down slope - it will make your reading say that the flow was slower than it really was
...

What you say might happen if the sensor is very poorly implemented for its intended use, but I hope F1 can do better than that. The sensor is inside the fuel tank which should give fair isolation from injectors and other external sources of noise. Unless someone purposely makes their fuel pump extremely noisy at ultrasonic frequencies, that seems unlikely to over-power a sensor that should be generating fairly high internal ultrasonic power at its operating frequency.

Even if there is an occasional sample (say one out of 1000 each second) that is corrupted, the software should be capable of eliminating a few outliers. So it seems to me that an ultrasonic sensor should be capable of reasonably determining flow rate at least averaged over a few seconds (which should be short enough to prevent unfair fuel use advantages).

[jz11]

sensor and pumps can be in the tank, that doesn't change the fact that the fluid is being measured and the fluid itself is resonating, that is the "noise" I'm talking about, and with short enough sample lengths you are measuring those oscillations of top of the fluid flow itself - thus resulting in "weird" readings

fix would be to to extend the sample length, but that is limited by sound propagation speed in the fluid and dimensions of the sensor itself

that's my guess anyway

edit, fuel flow meter is mounted most likely in between a low pressure and high pressure pumps, thus creating that closed loop where the resonance has a chance to build up

[sevgi]

@bill shoe, I agree with you on certain points that this is a point FIA has left open for debate and since there is no clearly writing set of rules governing this, these issues will continue to rise and RedBull was to blame in this case because they exceeded the tolerable limits. However because is no hard-line by FIA on this particular issue, that’s what prompted RedBull to try doing this to see if they can get away with this in the first place. Having limits on fuel-rate are essential in order to continue with F1 racing in future so there has to be a certain formula governing the fuel-rate based on temperature, humidity and altitude of the tracks.