Originally Posted by
CVividasku
I'm not saying the tailscrape at LHR was caused by PIO.
I'm saying that there is something that caused the PIO.
Something that should not have happened, that caused the PIO.
And the same thing that caused the PIO, also caused the tailstrike.
First, do you agree that the triangles on the elevator deflection are clear signs of PIO ?
However, there is not a PIO problem, prior to the tail strike.
There is a "flight control dynamics problem"
I zoomed in on the curves available here.
https://assets.publishing.service.go...XWBC_09-22.pdf
If you put the two curves together, you get this picture :
https://i.gyazo.com/7805953bd1bc0dfc...89e5c86214.png
You can see that the elevator follow closely the sidestick input, up until a point...
There is one nose down input, approximately 5 seconds (two and a half squares) before the red event (tail strike). Third black vertical line before the red event.
After that, there is one pitch up order. It is followed by the elevators, but late. And not to the extent that you would expect.
Then, the stick is released a bit. There is still some delay in that.
The stick is pulled again. This time there is a half second delay at the beginning of the pulling.
Then in the middle of the nose-up elevator movement, the blue line slows down. The delay is now one second between the elevators and sidestick pitch order.
As soon as the elevators reach full aft, the pilot releases his input up until the tailstrike. While the pilot releases his input, the elevators remain generally almost at full aft. You can even see, at the event, the pilot order is towards nose down. Has been in a nose down direction for 2 seconds. And the current position of the pilot order is neutral.
However, the elevators are near full aft, and are going full aft. Doing exactly the opposite of what the pilot had been ordering for two seconds.
Hence, this is clearly not the case.
Maybe in theory, not in practise.
The deed is done. The PIO occurs just after this moment, but it's already too late.
The triangles were a good indication that something weird was happening before them. And looking into it, it is the case.
There is a lot more to say about this. Notably, what would have happened if the elevators had followed the nose down tendency in the two seconds before tailstrike ? Since it's a very short duration, it's likely the pilot input would have been very similar.
It is possible to plot pitch, pitch rate, pitch rate rate (which is proportional to pitching moment), but I'm fairly sure that everything will be very consistent. It is aerodynamics, except if one aileron was broken in half, for sure everything is going to be consistent.
The discrepancy is between sidestick order and elevator response. The discrepancy during a PIO itself is due to elevator maximum speed being reached. The discrepancy before that is less obvious.
In any case, I think it wiser to pull up seriously when you hear the thrust reaching maximum. Not before that...
As for the rest, your post was really interesting. Some general comments :
I really like the APU missing incident.
We do not have the same notations so it's sometimes not easy to follow. What would V19 be, for example ?
You talking about the C* law shows in-depth knowledge of airbus FBW. It's fun to see how reality is different than the simple stuff told in the FCOM. It would be interesting to teach this in engineering school. Do you mean there is a PI filter on top of this equation, or that just having this equation plugged into the system is enough to make it work PI-like ?
For normal cases, going into all the matrix thingys and even the non linear effects is not really necessary. The simple stuff, pitch rate rate times pitch inertia equals pitch moment, without taking into account the other terms, is more than enough. In this case there is no "cross axis effect", all the angles are small..
K. I note your issue on the elevator position vs the SSC. Some background. The data in the report of the LHR case is based on the DFDR data, and that has various recording rates for various data channels. The recording in accordance with ARINC 717, however that data is converted from the backbone of the aircraft that was originally ARINC 429 for Airbus, (the standard was upgraded to ARINC 629 in 1995 but not applied to all OEMS. A far faster and greater bandwidth protocol is proposed as the ARINC 664 (AFDX protocol)). The QAR will take data from the DFDAU for easy access. The DFDR takes the output 717 sentence. (Normally, data systems generally are designed to annoy everyone involved, particularly the habit of OEMs and operators not bothering to pass on the Frames definition for the system that is installed, despite this normally being an obligation on the sale of an aircraft fitted with a FDR).
Where there is an option on the time period, it is dependent on the date of certification of the aircraft, Aug 19, 2002 being the determinative date. [1]
- Ground Air sensing; required at 1 sec resolution, or 0.25s for designs with TC issue post 19 AUG 2002.
- Radio Altimeter: 1s
- Control input, FBW system: 0.5, or 0.25s
- Pitch control surface: 0.5 or 0.25s
- Pitch attitude: 1.0 or 0.25s
- Normal acceleration, (vertical, Gz): 0.125s (freq
Hz)
etc.
What arises is that when doing an analysis based on the data, what looks like a smooth line, may not be at all, it is possible to introduce artefacts. The timing of the contact is open to timing error depending on what the investigator used as his metric. Did they use ground air sensing? if so, there is up to 0.25s timing error in that value, from the granularity of the data. If using the vertical G, then it is even more problematic, as the question arises as to what component of the fuselage bending and shock transmissions impact that value. For the SSC, and elevator, there absolutely can be differences between the SSC command and the elevator position.... the elevators in this case of the direct law are commanded electronically to alter a position by a hydraulic system. The electronic process lag time is not significant, but does exist, and normally would be nulled by the recording system for post process analysis. The hydraulic system is very dependent on airloads, and rates in order to follow the command of the pilot. The hydraulic actuator has a finite time to respond to a change in the valve porting of the hydraulic controls in response to the electrical actuation of the control valve "bobbin". Thereafter, with a high order command applied, the elevators take a finite time to respond, it is a hydromechanics system with all of the delays that involves. There is additionally a high likelihood that a control input that is being sampled for the DFDR at a modest rate, will not reflect the human input fully, and a sudden change in the control input that is then nulled out or reversed will not necessarily show the control input, or the response of the control surface correctly. In this case, the actuator response is able to be determined, and it shows the elevators are not instantaneous in changing their position, nor would they be expected to be.
The DFCS-FBW system operates at its own proprietary sampling frequency, which is certainly faster than the DFDR/DFDAU/QAR sampling. Looking at any system with a time domain analysis where sampling rates are different, or where there is a sequential data sampling within the sentence can introduce artefacts.
For the case that you bring up of the A350 @ Heathrow, (see below) The apparent "anomaly" that shows up is able to be explained as a sampling artefact. The only random momentary FCS anomaly that I am aware of was Kev Sullivans wild ride near Learmonth AUS, in a Qantas A330. In that case, the FCC had a suspected bit corruption from a possible cosmic ray badness. That resulted in a wild ride. If there has been any other momentary anomaly, I am not aware of it offhand, but they probably have occurred.
In the view below, the second red, double ended arrow is the time range where the elevator has an artefact of its position. This is within the error margins of the timing for the ground contact, and has a potential that a very short order SSC pitch command was sent, and not sampled, but responded to in due course by the FCC and the elevator hydraulic actuator, leading to what looks like an anomaly. The human body has its own resonant frequency for various parts, and in response to a load from the impact of the tail on the ground, a force will be transmitted up what is a flexible beam, the fuselage, to the pilot seat, and gives an acceleration to the protoplasm that sits proudly in the seat, holding the SSC, with/without the aid of the arm rest. momentary involuntary inputs can occur. These may be unwanted, but the aircraft response is not itself a PIO, nor is the pilots, it is an unwanted short period input, and that is highly speculative, the acceleration at the flight deck from the tail contact is not determined, and is way outside of the scope of this conversation (the CFRP barrel of the airframe of the A350 is pretty rigid, but it still has flexure under loads, just like planet earth does).
[1] Appendix M to Part 121, Airplane Flight Recorder Systems. (EASA EU OPS 1 is similar standard)