2025-06-15T15:05:00
permalink Post: 11902536
Flightradar24 (I know, I know) has a short blog on the (very minimal) ADS-B data available. There's only around 4s of useful data available from 21ft o 71ft altitude (last packet received 0.8s later), But: it's odd seeing the speed DROPPING shortly after takeoff. Even if you calculate total energy (kinetic + potential) it's falling, i.e. the engines aren't producing thrust. (In fairness reported speed doesn't match my calculated speeds, but even with mine I don't see power). Also: if you assumed no thrust from 71ft AGL @ 172kt you'd reach 250ft at 160kt. Isn't that roughly where they ended up? Noisy data, but this suggests the engines stopped producing power almost as soon as the wheels left the ground. (If someone could download a CSV of another similar flight and send to me I can do a compare and contrast of Total Energy)
6 users liked this post.
2025-06-15T16:46:00
permalink Post: 11902623
Flightradar24 (I know, I know) has a short blog on the (very minimal) ADS-B data available. There's only around 4s of useful data available from 21ft o 71ft altitude (last packet received 0.8s later), But: it's odd seeing the speed DROPPING shortly after takeoff. Even if you calculate total energy (kinetic + potential) it's falling, i.e. the engines aren't producing thrust. (In fairness reported speed doesn't match my calculated speeds, but even with mine I don't see power). Also: if you assumed no thrust from 71ft AGL @ 172kt you'd reach 250ft at 160kt. Isn't that roughly where they ended up? Noisy data, but this suggests the engines stopped producing power almost as soon as the wheels left the ground. (If someone could download a CSV of another similar flight and send to me I can do a compare and contrast of Total Energy)
1. You might mathematically start at TE=0 at start of takeoff-roll, and treat drag as minimal until rotation.
2. Typically rotation will be to say 15deg nose up, but flight path will be much less (5deg? for heavy hot 787). Once that rotation is complete, aircraft will stop accelerating. Therefore engine thrust (energy gain) equals gain in PE - drag x time.
3. This might give a better insight into where energy gain/loss became unusual?
Looking at the raw data in your post, and given the speeds are likely IAS based i.e. can be affected by wind, I don't see the IAS loss as equating to dual engine failure i.e. zero thrust - but could be wrong. Once a heavy airliner gets to lift off the acceleration is reduced (drag) and the decays to zero as PE gain kicks in.
Ditto a time / distance to the crash site might give some energy info? Looks like the crash site is 50' (?) below the airfield (Google Earth will give more).
I think you are doing an interesting study on the absence of other info
2025-06-15T18:32:00
permalink Post: 11902696
Flightradar24 (I know, I know) has a short blog on the (very minimal) ADS-B data available. There's only around 4s of useful data available from 21ft o 71ft altitude (last packet received 0.8s later), But: it's odd seeing the speed DROPPING shortly after takeoff. Even if you calculate total energy (kinetic + potential) it's falling, i.e. the engines aren't producing thrust. (In fairness reported speed doesn't match my calculated speeds, but even with mine I don't see power). Also: if you assumed no thrust from 71ft AGL @ 172kt you'd reach 250ft at 160kt. Isn't that roughly where they ended up? Noisy data, but this suggests the engines stopped producing power almost as soon as the wheels left the ground. (If someone could download a CSV of another similar flight and send to me I can do a compare and contrast of Total Energy)
For instance, what about induced drag (admittedly much complicated, I imagine, by varying ground effect) once rotation begins? A comparison with another 787-8 flight from the same runway and under similar conditions (meteorological, load, etc.) might be ideal.
Are there not tools accessible to ordinary users for making detailed such simulations, rather than a back of the envelope calculation? I daresay Boeing has made such simulations already, and have a pretty good idea of whether and when thrust loss might have occurred.
2025-06-20T10:01:00
permalink Post: 11906794
Engineer not a pilot. Experience in analog front ends, A2D and R2D conversion and embedded systems generally but no specific knowledge of the 787 or GEnx.
I like everyone else have no evidence that TMCA played a role but given that it is one of the few systems with the ability to cut fuel to the engines, here are some thoughts on how signal processing could have extended the window of when TMCA could bite. In particular, I'm looking at the time immediately after the nose lifts up when something may have physically shifted onboard.
I'll phrase it as a number of questions but realise that the few people who can answer may not be able to for now.
Thanks to tdracer's explanation on TMCA (albeit 747 not 787), we know that TMCA is a logic block within the FADEC whose only external inputs are a logic signal fron the aircraft that indicates whether it is on the ground or not and throttle position as determined by two independent resolvers per throttle side.
The logic would seem to be something of the form
If (G AND (N2>A OR N2>B)) Then CutOffFuel()
where G is true when the aircraft is on the ground,
A is an envelope defined by throttle resolver channel A and
B is an envelope defined by throttle resolver channel B
Q1: Am I correct in that assumption that when on the ground, overspeed with respect to EITHER resolver A OR resolver B can trigger TMCA?
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
Q2: When the nosewheel lifts off but the MLG is still on the ground and RadALT is close to ground, will G still be true?
Next, it is common when data fusing multiple inputs that there is a desire to clean up a signal before it is sampled digitally. This can remove effects such as switch bounce. The inclusion of low pass filters or hysteresis will generally add a propogation delay.
Q3: Is there a slow filter (Tc>=1s) in the ground/air logic which could have caused a slight delay before G became false after takeoff further extending the opportunity of TMCA to activate?
Q4: Does TMCA act almost instantly or does it wait for the fault condition to stay asserted for a period of time before acting?
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
It would still need a mechanism for at least one throttle input to each FADEC to misbehave at the same time. Resolvers are fed with an excitation signal to the rotor and take back two orthogonal signals (Cos and Sin) from stator windings. Usually, the excitation comes directly from the resolver-to-digital (R2D) circuit but sometimes an external signal source is used. I would hope that in an aircraft system, each channel would be kept independent of everything else.
Q6: Does the excitation signal for the 4 throttle resolvers (2 per side) come from 4 independent (internal) sources?
My last thought for a single point of failure between both throttles would be a short between two wires or connection points carrying resolver signals, one from each side. Whether this could be caused by swarf wearing within a wiring loom, a foreign object moving about, crushed wires or even stretching of adjacent wires, I have absolutely no idea.
Q7: Do resolver signals from left or right, either channel A or B, run next to each other in a loom at any point?
I like everyone else have no evidence that TMCA played a role but given that it is one of the few systems with the ability to cut fuel to the engines, here are some thoughts on how signal processing could have extended the window of when TMCA could bite. In particular, I'm looking at the time immediately after the nose lifts up when something may have physically shifted onboard.
I'll phrase it as a number of questions but realise that the few people who can answer may not be able to for now.
Thanks to tdracer's explanation on TMCA (albeit 747 not 787), we know that TMCA is a logic block within the FADEC whose only external inputs are a logic signal fron the aircraft that indicates whether it is on the ground or not and throttle position as determined by two independent resolvers per throttle side.
The logic would seem to be something of the form
If (G AND (N2>A OR N2>B)) Then CutOffFuel()
where G is true when the aircraft is on the ground,
A is an envelope defined by throttle resolver channel A and
B is an envelope defined by throttle resolver channel B
Q1: Am I correct in that assumption that when on the ground, overspeed with respect to EITHER resolver A OR resolver B can trigger TMCA?
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
Q2: When the nosewheel lifts off but the MLG is still on the ground and RadALT is close to ground, will G still be true?
Next, it is common when data fusing multiple inputs that there is a desire to clean up a signal before it is sampled digitally. This can remove effects such as switch bounce. The inclusion of low pass filters or hysteresis will generally add a propogation delay.
Q3: Is there a slow filter (Tc>=1s) in the ground/air logic which could have caused a slight delay before G became false after takeoff further extending the opportunity of TMCA to activate?
Q4: Does TMCA act almost instantly or does it wait for the fault condition to stay asserted for a period of time before acting?
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
It would still need a mechanism for at least one throttle input to each FADEC to misbehave at the same time. Resolvers are fed with an excitation signal to the rotor and take back two orthogonal signals (Cos and Sin) from stator windings. Usually, the excitation comes directly from the resolver-to-digital (R2D) circuit but sometimes an external signal source is used. I would hope that in an aircraft system, each channel would be kept independent of everything else.
Q6: Does the excitation signal for the 4 throttle resolvers (2 per side) come from 4 independent (internal) sources?
My last thought for a single point of failure between both throttles would be a short between two wires or connection points carrying resolver signals, one from each side. Whether this could be caused by swarf wearing within a wiring loom, a foreign object moving about, crushed wires or even stretching of adjacent wires, I have absolutely no idea.
Q7: Do resolver signals from left or right, either channel A or B, run next to each other in a loom at any point?
4 users liked this post.
2025-06-20T10:56:00
permalink Post: 11906831
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
2025-06-20T11:02:00
permalink Post: 11906835
Engineer not a pilot. Experience in analog front ends, A2D and R2D conversion and embedded systems generally but no specific knowledge of the 787 or GEnx.
I like everyone else have no evidence that TMCA played a role but given that it is one of the few systems with the ability to cut fuel to the engines, here are some thoughts on how signal processing could have extended the window of when TMCA could bite. In particular, I'm looking at the time immediately after the nose lifts up when something may have physically shifted onboard.
I'll phrase it as a number of questions but realise that the few people who can answer may not be able to for now.
Thanks to tdracer's explanation on TMCA (albeit 747 not 787), we know that TMCA is a logic block within the FADEC whose only external inputs are a logic signal fron the aircraft that indicates whether it is on the ground or not and throttle position as determined by two independent resolvers per throttle side.
The logic would seem to be something of the form
If (G AND (N2>A OR N2>B)) Then CutOffFuel()
where G is true when the aircraft is on the ground,
A is an envelope defined by throttle resolver channel A and
B is an envelope defined by throttle resolver channel B
Q1: Am I correct in that assumption that when on the ground, overspeed with respect to EITHER resolver A OR resolver B can trigger TMCA?
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
Q2: When the nosewheel lifts off but the MLG is still on the ground and RadALT is close to ground, will G still be true?
Next, it is common when data fusing multiple inputs that there is a desire to clean up a signal before it is sampled digitally. This can remove effects such as switch bounce. The inclusion of low pass filters or hysteresis will generally add a propogation delay.
Q3: Is there a slow filter (Tc>=1s) in the ground/air logic which could have caused a slight delay before G became false after takeoff further extending the opportunity of TMCA to activate?
Q4: Does TMCA act almost instantly or does it wait for the fault condition to stay asserted for a period of time before acting?
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
It would still need a mechanism for at least one throttle input to each FADEC to misbehave at the same time. Resolvers are fed with an excitation signal to the rotor and take back two orthogonal signals (Cos and Sin) from stator windings. Usually, the excitation comes directly from the resolver-to-digital (R2D) circuit but sometimes an external signal source is used. I would hope that in an aircraft system, each channel would be kept independent of everything else.
Q6: Does the excitation signal for the 4 throttle resolvers (2 per side) come from 4 independent (internal) sources?
My last thought for a single point of failure between both throttles would be a short between two wires or connection points carrying resolver signals, one from each side. Whether this could be caused by swarf wearing within a wiring loom, a foreign object moving about, crushed wires or even stretching of adjacent wires, I have absolutely no idea.
Q7: Do resolver signals from left or right, either channel A or B, run next to each other in a loom at any point?
I like everyone else have no evidence that TMCA played a role but given that it is one of the few systems with the ability to cut fuel to the engines, here are some thoughts on how signal processing could have extended the window of when TMCA could bite. In particular, I'm looking at the time immediately after the nose lifts up when something may have physically shifted onboard.
I'll phrase it as a number of questions but realise that the few people who can answer may not be able to for now.
Thanks to tdracer's explanation on TMCA (albeit 747 not 787), we know that TMCA is a logic block within the FADEC whose only external inputs are a logic signal fron the aircraft that indicates whether it is on the ground or not and throttle position as determined by two independent resolvers per throttle side.
The logic would seem to be something of the form
If (G AND (N2>A OR N2>B)) Then CutOffFuel()
where G is true when the aircraft is on the ground,
A is an envelope defined by throttle resolver channel A and
B is an envelope defined by throttle resolver channel B
Q1: Am I correct in that assumption that when on the ground, overspeed with respect to EITHER resolver A OR resolver B can trigger TMCA?
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
Q2: When the nosewheel lifts off but the MLG is still on the ground and RadALT is close to ground, will G still be true?
Next, it is common when data fusing multiple inputs that there is a desire to clean up a signal before it is sampled digitally. This can remove effects such as switch bounce. The inclusion of low pass filters or hysteresis will generally add a propogation delay.
Q3: Is there a slow filter (Tc>=1s) in the ground/air logic which could have caused a slight delay before G became false after takeoff further extending the opportunity of TMCA to activate?
Q4: Does TMCA act almost instantly or does it wait for the fault condition to stay asserted for a period of time before acting?
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
It would still need a mechanism for at least one throttle input to each FADEC to misbehave at the same time. Resolvers are fed with an excitation signal to the rotor and take back two orthogonal signals (Cos and Sin) from stator windings. Usually, the excitation comes directly from the resolver-to-digital (R2D) circuit but sometimes an external signal source is used. I would hope that in an aircraft system, each channel would be kept independent of everything else.
Q6: Does the excitation signal for the 4 throttle resolvers (2 per side) come from 4 independent (internal) sources?
My last thought for a single point of failure between both throttles would be a short between two wires or connection points carrying resolver signals, one from each side. Whether this could be caused by swarf wearing within a wiring loom, a foreign object moving about, crushed wires or even stretching of adjacent wires, I have absolutely no idea.
Q7: Do resolver signals from left or right, either channel A or B, run next to each other in a loom at any point?
What happens when the 2 disparate processes that form TCMA disagree?
2025-06-20T11:04:00
permalink Post: 11906838
Let me try to answer the questions about which I have some knowledge, as an aerospace engineer:
(I am not sufficiently informed to answer Q4,6 and 7, at the moment)
Yes, overspeed on either resolver channel A or B alone will trip the TMCA fuel-cut logic
G is a single Boolean that FADEC derives from Weight-On-Wheels (MLG and NW) and radio-altimeter. Iit stays \x93ground\x94 until all WOW sensors go inactive (i.e.
every
wheel is off the runway) and the RadALT exceeds its airborne threshold.
That's very unlikely. Ground/air logic uses small hysteresis (tens to a few hundred ms), but not in the multi-second range
Let's do the math very quickly:
Kinetic energy with a weight of 200,000kg, at Vr = 150kn = 77m/s: E_kin = 600MJ
Rotational energy of a GEnX engine is hard to calculate as I don't find reliable values for the rotary inertia. I found some for a GE90 and could roughly estimate 100MJ of rotational energy for each engine. However, I seriously doubt that this energy could be effectively used to gain thrust, as the thrust will drop very quicjkly after the fuel is cut off.
the required potential energy for a 100ft climb of a 200,000kg 787 is around 70MJ.
This ignores aerodynamic drag, still, 100 ft of climb remains energetically feasible.
However, it as been pointed out several times that the actual climb was higher than 100ft. Already for 200ft I would doubt the validity of my statement above.
(I am not sufficiently informed to answer Q4,6 and 7, at the moment)
Q1: Am I correct in that assumption that when on the ground, overspeed with respect to EITHER resolver A OR resolver B can trigger TMCA?
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
We have been told that the logic (ie true or false) signal G is determined from the Weight-on-wheels sensors and the RadALT. It is reasonable to suppose that the designers still wanted TMCA to function after a hard landing where some landing gear components had failed.
At that point, the total energy of the system would have comprised of the kinetic energy of the aircraft travelling at Vr, the rotational inertia of the engines and the potential energy of whatever fuel is beyond the cutoff valves.
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
Q5: Would this total energy have been sufficient to get the aircraft 100ft into the air?
Kinetic energy with a weight of 200,000kg, at Vr = 150kn = 77m/s: E_kin = 600MJ
Rotational energy of a GEnX engine is hard to calculate as I don't find reliable values for the rotary inertia. I found some for a GE90 and could roughly estimate 100MJ of rotational energy for each engine. However, I seriously doubt that this energy could be effectively used to gain thrust, as the thrust will drop very quicjkly after the fuel is cut off.
the required potential energy for a 100ft climb of a 200,000kg 787 is around 70MJ.
This ignores aerodynamic drag, still, 100 ft of climb remains energetically feasible.
However, it as been pointed out several times that the actual climb was higher than 100ft. Already for 200ft I would doubt the validity of my statement above.
2 users liked this post.
2025-06-20T16:04:00
permalink Post: 11907092
In the CSV data set that can be downloaded from that link the first point with altitude data is 1630 ft short of the departure threshold. That point is 575. The highest alt recorded in the data set is 625. All the points with altitude data overlay the departure runway. I do not understand how anyone is using this data set to determine the maximum altitude which was way past the departure end.
Edit to add - I have made no attempt to correct the raw ADS-B altitude data. There is no need to make any correction to see altitude gain.
Edit to add - I have made no attempt to correct the raw ADS-B altitude data. There is no need to make any correction to see altitude gain.
I've had a bit more time to analyse now.
Those ADS-B data points (and particularly the rate of deceleration) are EXACTLY what you would expect to see from a total engine failure at or very shortly after TAKE-OFF
(it implies a 13:1 L/D which must be pretty close for gear down and flaps 5).
It places takeoff at 700m before the runway end @ ~185kt
Based on those, max altitude was c250ft @ 140kt (or the equivalent total energy equivalent), 500m after the end of the runway.
13:1 L/D would also get you groundspeed on impact of 120kt
Do those numbers make sense?
2025-06-20T16:39:00
permalink Post: 11907121
I posted my first-cut analysis in the earlier thread.
I've had a bit more time to analyse now.
Those ADS-B data points (and particularly the rate of deceleration) are EXACTLY what you would expect to see from a total engine failure at or very shortly after TAKE-OFF
(it implies a 13:1 L/D which must be pretty close for gear down and flaps 5).
It places takeoff at 700m before the runway end @ ~185kt
Based on those, max altitude was c250ft @ 140kt (or the equivalent total energy equivalent), 500m after the end of the runway.
13:1 L/D would also get you groundspeed on impact of 120kt
Do those numbers make sense?
I've had a bit more time to analyse now.
Those ADS-B data points (and particularly the rate of deceleration) are EXACTLY what you would expect to see from a total engine failure at or very shortly after TAKE-OFF
(it implies a 13:1 L/D which must be pretty close for gear down and flaps 5).
It places takeoff at 700m before the runway end @ ~185kt
Based on those, max altitude was c250ft @ 140kt (or the equivalent total energy equivalent), 500m after the end of the runway.
13:1 L/D would also get you groundspeed on impact of 120kt
Do those numbers make sense?
Exact location of house, Approx distance of 1.5 km from end of runway to crash site .
Coordinates of the camera : 23\xb003'42.3"N 72\xb037'03.5"E
The Approx Camera location of the Balcony is the Red Mark . Can the speed be calculated . Does the speed line up with the ADS B data , Does it Gain Any speed after this Balcony point ?
Co-incidently Another Witness is the Grand Mother of the Balcony Teen, she was closer to the airport as per her . she is saying that the engine was silent after it passed over (but making sound , when it was Over , RAT already deployed?? ) and made offhand comment it was gonna crash . Found that out later .
2025-06-20T17:56:00
permalink Post: 11907161
PPL IR engineer in air trafic flow management, so not quite the right kind of engineer to talk about physics and total energy.
Nonetheless, I ran some calculation with the Eurocontrol BADA total energy model equation using a rough estimation of the height/time profile from the CCTV footage (wingspan yardstick method) and an approximation of the drag using a constant drag hypothesis of 200 kN. That amount of drag is about what a fully configured B788 on final approach generates, I felt it was an acceptable approximation for a less draggy configuration flown at higher than normal AoA but I could be very wrong. I used the total energy equation to calculate speed, inputs are the the height gains and the drag hypothesis.
I find a B788 with a mass of 207 t and an initial speed of 185 kts with no thrust from the rotation onward could theoretically have traded 50 kts for 190 ft before descending back down at speeds between 135 kts and 130 kts, coming back to take off elevation 1950 m after the last ADSB position. The flown distance is sort of the checksum in this, it is not an input but rather the output of the model.
I also did the same exercise with the EK521 incident, assuming a 285 t B77W rotating at 155 kts for a go around on idle thrust. I guestimated 350 kN of drag (normal approach in full configuration is supposedly 380 kN) and got a speed decay to about 130 kts at 80 ft which is how high that particular aircraft got before coming back down. I am not sure this is enough to give confidence in my attempt at doing total energy calculation, but at least it doesn't seem too far off.
Nonetheless, I ran some calculation with the Eurocontrol BADA total energy model equation using a rough estimation of the height/time profile from the CCTV footage (wingspan yardstick method) and an approximation of the drag using a constant drag hypothesis of 200 kN. That amount of drag is about what a fully configured B788 on final approach generates, I felt it was an acceptable approximation for a less draggy configuration flown at higher than normal AoA but I could be very wrong. I used the total energy equation to calculate speed, inputs are the the height gains and the drag hypothesis.
I find a B788 with a mass of 207 t and an initial speed of 185 kts with no thrust from the rotation onward could theoretically have traded 50 kts for 190 ft before descending back down at speeds between 135 kts and 130 kts, coming back to take off elevation 1950 m after the last ADSB position. The flown distance is sort of the checksum in this, it is not an input but rather the output of the model.
I also did the same exercise with the EK521 incident, assuming a 285 t B77W rotating at 155 kts for a go around on idle thrust. I guestimated 350 kN of drag (normal approach in full configuration is supposedly 380 kN) and got a speed decay to about 130 kts at 80 ft which is how high that particular aircraft got before coming back down. I am not sure this is enough to give confidence in my attempt at doing total energy calculation, but at least it doesn't seem too far off.
9 users liked this post.
2025-06-21T08:19:00
permalink Post: 11907566
"Boeing explained that the RAT will remain operational as the airplane decelerates through the minimum RAT design speed of 120 knots, not 130 knots. Boeing expressed that the performance of the RAT was shown to meet the Boeing Model 787 requirement that specifies 120 knots as the minimum RAT design speed. We agree that the RAT will remain operational as the airplane decelerates through the minimum RAT design speed of 120 knots, not 130 knots..."
Again I'm not sure this is of any particular utility now, but is included here in the interests of ensuring as much factual data is available as possible.
FP.
5 users liked this post.