Sunday, December 4, 2016

LMI2933 LAMIA AVRO RJ85 Medellín Deadstick

LMI2933 approach to SKRG.  The airplane position is plotted as a function of time (where it effectively flows to the left). The airplane initially arrived with conceivably enough fuel to complete the approach and landing safely.  The airplane entered a holding pattern and on the back side of the second circuit departed the holding pattern with apparently total engine failure.


KEY ASSUMPTIONS

From Start of Descent: 21,000 press alt. 
26 nm to Rwy 01 (7025 MSL feet)
16 nm to VOR (crash site ~8683 MSL feet)
8 nm to last report (15,500 press alt.)

From Last Report: 15,500 press alt.
18 nm to Rwy 01
8 nm to VOR-crash site


The lengthy discussion reflected in this blog is a tool to understand the relevance and benefit of ADS-B data as s tool for safety investigation. This is not a safety investigation of LMI2933.  The official safety investigation will conclusively resolve any uncertainties.  

Please take note that other explanations or factors not known or misunderstood herein may correctly determine different conclusions - in other words, the conclusions herein may be wrong.

As an aid to organizing the data, I have come to some conclusions that rely on assumptions.  The conclusions include:
  1. ADS-B reported position is accurate to about 4km.
  2. Aircraft transitioned to landing configuration (flaps/gear down) while still under power, in the beginning of the descent.
  3. Aircraft transitioned to un-powered flight between 15,000 and 13,000 feet.
  4. Flight continued under deadstick control, gliding at expected ratio.
ATC clearance instruction to hold is not known. Position suggests holding at GEMLI, while ILS approach would have been expected RNG VOR.  The concern that the moving map (LNAV) function was offset or a source of loss of situational awareness awaits the official investigation to resolve.  

The voice timeline gives an estimate of distance traveled - by assuming the same 140 KTAS to the point of impact - that arrives at a point southerly matching the ADS-B data.

ATC transcription overhears "Gear Down" in background, but this may not reflect any action taken. The data does show that the airplane slowed by nosing up a bit and then a stable descent was resumed at a slower speed. The airplane needed to lose 11,000 feet in about 16 nm to the final approach point (FAP) for the ILS.  They had stated a plan to join the localizer.

Each engine runs on its own feed tank, and the inner tanks may drain sooner than the outer tanks. Without scavenging pumps, more fuel may be unusable.  As well, sloshing or unusual attitude could move fuel off of the engine drains.  Engine failure may have been progressive rather than collective, and there is no information to judge their occurrence other than the pilot call.

LMI2933 ATC reported total loss of electric and fuel.  This information came at a time beyond the last ADS-B report and cannot be discerned otherwise (there is no ADS-B data beyond 15,500 feet).  If CP2933 was equipped with standard fuel tanks, the endurance and predicted payloads appear to match with arrival on fumes.  Fuel starvation is another possibility.  Fuel exhaustion is the most obvious explanation, but it cannot be determined by the ADS-B data itself.  

The last ADS-B reported position to the point of rest is connected by a straight line.  A constant 140 KTAS is assumed.  From this data, the voice timeline is overlaid. The reports are all made consistent with predicted altitude.  The voice calls stated a consistent northerly heading.  There was no sense of control loss, even to the call at 9000 feet (seconds from impact), rather situation awareness possibly by loss of localizer signal.  The achieved glide ratio (5.7) is about what BAE Systems predicts for flaps/gear down (6:1).  All with flight at night, with some cloud banks, under emergency lighting, using sole reference to standby attitude, airspeed, and a direction/bearing indicator, and a flagging localizer signal.

Contents
ATC Transcription
Last Moments
Accuracy of Recorded Data
FMS Composite Position
Holding Pattern offers Position Reference
Flight Plan
RJ85 Type Certificate
RJ85 Performance Data
RJ85 Fuel System
FEED LO LEVEL Warning
RJ85 Hydraulics
RJ85 Electrical System
RJ85 Flight Controls
RJ85 Communication Systems
RJ85 Lighting
RJ85 Navigation and Instruments
RJ85 All-Engine Out Procedures
Final Maneuvering
Conclusion
Deja Vu
Plots
Graphs
Transponder Data

SUMMARY NOTE

These observations stem from sparse ADS-B reports, an unauthorized copy of a related RJ85 Flight Crew Operations Manual, Vol 1. plus other references not known to be accurate in comparison to CP2933 airplane combined with best-intentioned judgements and helpful advice from online commentators.
  1. The flight plan from SLVR to SKRG with 77 aboard may be beyond the legal range of this RJ85, even with auxiliary pannier tanks.  The same plane had flown the city pair in reverse on Nov 4, with even a longer stage length of over 4:33 (reported ferry flight, presumed light payload).  The investigation will need to reconcile these actions.
  2. FEED LO LEVEL warning should have been raised at least 150 km from SKRG.  Each feed tank has a dedicated gage for constant monitoring.  There can should be no doubt the flight crew knew they had a little more than 20 minutes of fuel left at that point.  The FMS offers endurance calculations as well, giving considerable foresight into arrival conditions.
  3. There appears to have been sufficient fuel for a straight-in approach, and after one circuit around the holding pattern (by virtue of the estimated flameout position).  The decision to turn outbound on the second circuit would therefore have left fuel exhaustion inevitable.
  4. ADS-B position data appears to be driven from an IRS source, not GNSS.  The departure position was offset by 2,000m.  Overlay of holding pattern as a means to estimate arrival IRS position drift revealed an easterly offset of about 4 km.  Using the voice transcript, the estimated position of the last ADS-B report was 15 km south.  This matches the ADS-B position.
  5. It is assumed that LMI2933 was holding at GEMLI at 21,000 feet.  The ATC clearance and intention for an RNAV approach is unknown.  Once LMI2933 began to descend, they stated they would follow the localizer.  The last ADS-B report from LMI2933 was over 3,000 feet above the approach path.
  6. The necessary glide ratio (to make the runway threshold) while holding at 21,000 feet peaked at about 11:1 on the backside of the pattern (to SKRG runway 01).
  7. The necessary glide ratio (to make the runway threshold) from the last reported position had increased to 13:1 (due to their relatively steep descent).
  8. The RJ85 operates with no fuel (windmilling engines, no APU, no standby generator) by using a battery to power one VHF, ILS, VOR, ADF set, IRS, standby attitude and direction bearing, and hydraulic power for landing gear extension and braking.  The pitch and roll controls are mechanical.  The rudder control is possible only if the windmilling on Eng 3 is sufficient and only if the crew isolates the standby generator.  Flap, air brake: are not powered. Spoilers may be available.  Passenger Address should be available, at least from the flight crew.
  9. RJ85 predicted glide ratio is 15:1 flaps up at best glide speed.  
  10. Recommended procedure in all-engine failure is to only extend gear only when landing is assured (1,000 feet of altitude per nm), as glide ratio drops to 6:1 with gear extended (whether flaps up or flaps down).
  11. LMI2933 descended from 21,000 feet presumably under waning power.  The last engine flameout is estimated at approximately between 13,000 and 15,000 feet. 
  12. In the descent the airplane was slowed to what appears to be flaps/gear down approach configuration.  
  13. The estimated glide ratio from the last reported position to the crash site is about 5.7:1 (matching predicted gear down glide ratio of 6:1).
  14. LMI2933 may have been capable to glide to SKRG runway 01 until the point that the flight crew extended flaps and gear.  It is believed that the crew extended flaps and gear prior to flameout and may have been confident in their continued flight worthiness at that moment.   There is no provision in the RJ85 to raise flaps or gear with all engines failed.
  15. The Cockpit Voice Recorder and the Flight Data Recorder are not powered by EMERG AC/DC and would have stopped functioning when the engines flamed-out.
Detailed discussion follows below.

Introduction

On 2016-11-29 at about 02:58Z (or 28 Nov @ 9:58pm local time), flight LMI2933, an AVRO RJ85 crashed into a hillside on approach to SKRG (Medillin).  The flight appears to have exhausted its fuel supply and the pilot completed the deadstick approach and off-airport landing in total darkness.  With only a handful of survivors, the catastrophe appears to be the result of poor planning and tragically unfortunate circumstances.

The purpose of this article (the third in this series) is to examine the ADS-B data as telemetry.  The prior two reports are:

EA3452 "Pence" plane overrun at LGA
ADS-B Telemetry from EK521

ADS-B information is minimal and lacking in comparison to the onboard data and voice recorders, and the other information available to air-safety investigators.  With the tragic loss of life, a thorough and comprehensive investigation from air safety investigators will offer conclusions of probable cause and recommendations for safety improvements.  The purpose of this report is to evaluate ADS-B as a tool for air-safety investigation.  There may have significant errors and this report should not be considered authoritative in any sense.

The flight crew were faced with a dire situation when they started their deadstick approach having left the back side of the holding pattern.  The figure below portrays the night-time perspective from the trail of flight LMI2933 to the last reported ADS-B report, with the crash site and the threshold of SKRG runway 01 highlighted ahead north-northwest.


Colombia;  U.S. Defense Mapping Agency Aerospace Center, compiled 1966 
http://www.lib.utexas.edu/maps/onc/txu-pclmaps-oclc-8322829_l_26.jpg
https://skyvector.com

https://skyvector.com

Weather

Winds were light at the surface.  Temperature was one degree above dew point, making fog a possibility.  Visibility was good, in a light drizzle.

Ceilings were dropping.  At the time of the accident (290358Z) a broken layer had formed at 1500 AGL (8,500 MSL, the elevation of the crash site).  A scattered layer had formed recently at about 15,000 MSL.
ARCHIVED METAR OF: 20161129 // FROM: 0 TO: 3 UTC 
SKRG 290000Z 06003KT 9999 SCT017 SCT200 17/16 A3019
SKRG 290100Z 01003KT 9999 SCT017 SCT200 17/16 A3020
SKRG 290200Z 00000KT 9999 BKN015 SCT200 17/16 A3023
SKRG 290300Z VRB02KT 9999 -DZ BKN015 SCT080 17/16 A3025
Sunset was 5:44pm local (282244Z)
Moonset was 5:21pm local (new moon) (282221Z)

Table of Contents

ATC Transcript

The ATC recordings have been transcribed.  My complements to Alejandro Franz for providing the annotated Spanish version, and VAS Aviation for their translations to complement Google translate (I do not speak Spanish).  I have added the Spanish report at the end of this blog entry.  Please take note that this readout may not be fully accurate.

From avherald.com:
Incident: VivaColombia A320 near Medellin on Nov 28th 2016, suspected fuel leak
By Simon Hradecky, created Wednesday, Nov 30th 2016 14:05Z, last updated Sunday, Dec 4th 2016 08:59Z 
A VivaColombia Airbus A320-200, registration HK-5051 performing flight FC-8170 from Bogota to San Andres Island (Colombia), was enroute at FL360 about 150nm northwest of Medellin at 02:11Z when the crew initiated a diversion to Medellin reporting a suspected fuel leak (engine CFM56), turned around and descended to FL160 on the way to Medellin. The aircraft reached the Final Approach Fix for the approach to runway 01 at 02:45Z and touched down safely at 02:51Z.

FC8170 diverted to SKRG worried over a fuel leak, landing about the time LMI2933 started their emergency descent
FC8170 had transferred to tower control by the start of the transcript (my assumption), and does not show up in the Rio Negro approach control transcript.

FC8170 and LMI2933 arrive about the same time, but FL8170 is offered a direct approach, where LMI2933 holds.
FC8170 passes by LMI2933 about 10,000 feet lower.
FC8170 passes by LMI2933 about 10,000 feet lower.
FC8170 direct intercept for the approach to SKRG

In addition to LMI2933, there were four other airplanes communication with Rio Negro approach control.  The order for landing (planned) was:

  1. FC8170 (landed 290251Z)
  2. LA3020
  3. AV9356
  4. AV9771
  5. LMI2933

LA3020 - arrived from SKBO to SKRG ahead of LMI2933 (#1 for approach sequence).  It is vectored to the west at 13,000.


LA3019 - departed SKRG at 290246Z for SKBO, not a factor.  Tail number not known.  Normally an A320 or A319.



AV9771 - arrive from Cartagena ahead of LMI2933. Holding below LMI2933 at 18,000 feet. #3 in approach sequence.  It is vectored to the west at 18,000 feet.



AV9356- arrive from SKBO for MDE ahead of LMI2933.  Flightradar24.com had only one data point, about 30 minutes prior to the accident.  #2 in approach sequence.  It is vectored to the west and ends up holding at RNG VOR at 19,000 feet.



The time stamps are relative to the timeline on the audio recording, where 0:00.000 is about 290251Z.

Avionca 9771 start the approach, turn inbound


FC8170 lands about the time LMI2933 notifies ATC of fuel concerns.
LAN Colombia 3020 heading 010 - then a bit later 090 - 


Avionca 9771 turning left 270

LA3019 departed Medellin heading for Bogota well ahead of LMI2933 emergency



LMI2933 descending.  Avionca still on 270, will turn left for direct RNG VOR


LAN Colombia 3020 dodges weather at 13,000 feet



Avionca flew over RNG VOR (the crash site) about 10,000 AGL about three minutes afterwards

There are various versions of the ATC audio recording on-line.

VAS Aviation provided a bilingual version, but it has been edited and does not reflect real-time

bluradio.com provided an unedited version which I have used to create a time-line.

The altimeter was 30.27", so I have adjusted the ADS-B reported pressure altitude up by 350 feet to reflect MSL.  Through Google Earth, I have estimated the point of impact at 8683 feet.  In the following two charts, LMI2933 radio transmissions are matched in time and estimated altitude.



The ADS-B reports ended at 15,500 pressure altitude.  The point of impact is known.  The distance traveled past the last report to the crash site can be crudely estimated by assuming that the aircraft continued at about 140 KTAS throughout.  From the call at 16,000 to the call at 9,000 feet is 192 seconds.  That equates to an average of -2188 fpm vertical speed and about 8.1 nm traveled (14.9 km).  That puts the last ADS-B reported position about 8 nm (15 km) south of the crash site.

With this in-hand, the remainder of the discussion will use the raw ADS-B calculations, which reflect a 15 km distance to match the voice transcript.

The Google Earth images reflect an offset of the ADS-B reports to better match the expected drift to match the hold at GEMLI.  This offset is westerly 2.1 nm (4 km).  The plots are also offset northerly to match GEMLI, but may be 1 nm too far north.

Table of Contents

The Last Moments: Sorrow for Those Lost or Suffering

It is easy to overlook the human tragedy and suffering brought forth by this accident.  The official investigation will offer an explanation for the events and any recommendations for improving safety. This report is operating from only a handful of data points and does not profess to be accurate.  I am sad for the outcome and loss of life, and have taken personal initiative to try and be sure every technical aspect is considered and is sensible.

The captain of LMI2933 should have been alerted more than 10 minutes prior to arrival that fuel exhaustion was imminent.  Growing apprehension from the FMS predictions while watching the main tanks drain, to the warning that the feed tanks were draining and watching them go down, to the LO PRESS warnings and eventual flameout.  The descent began with the hope they would make it, but it quickly degraded into a cascade of associated annunciations suddenly silenced as the airplane is left to operate on battery powering one channel of minimal instruments and controls.

The passengers would have been suddenly and so unexpectedly thrust into a dark cabin and silent runnings for two minutes or less.  With the failure of all engines and no APU is the loss of pneumatics with means loss of pressurization - the oxygen masks might have been been deployed.

The cabin crew were the most surprised to find themselves in a dire situation, hastily thinking through evacuation procedures and most likely placating passenger fears.  The cabin interphone was inoperative, the Passenger Address was the only cabin comms available.

The lack of a post-crash fire, while aiding the survival of a few persons, was also a telltale of the tragedy.

Table of Contents

ADS-B Accuracy of Reported Position

The takeoff from SLVR suggests that the position reporting over ADS-B reports may have been offset by about 2 km (1 nm).

LMI2933 route is shown offset from Google Earth presentation of runway 34
While a bit disappointing, this error cannot be confirmed in magnitude in the approach to SKBR over four hours later.

Transponder terminology is confusing.

Mode A/C/S transponders respond to 1030 MHz interrogation from Secondary Surveillance Radar (SSR) with a 1090 MHz squawk, with progressively more and more information based on the type.

Mode A ident
Mode C altitude

Mode S has the capacity to deliver a wide range of parameters using three Mode-S Specific Protocols (MSP)
  1. ELS Elementary Surveillance; 
  2. EHS Enhanced Surveillance; and 
  3. ADS-B Automatic Dependent Surveillance Broadcast ADS-B 
    1. implemented in Mode S as 1090 MHz “Extended Squitter” (1090ES)
In response a 1030 MHz interrogation,  a Mode-S transponder would respond with Elemetary or Enhanced Surveillance squawk on 1090 MHz.  Neither form of a squawk broadcasts aircraft position.  Aircraft position may be determined by the SSR using bearing and range directly or through multi-lateration.


ADS-B 1090 Extended Squitter (1090ES) is broadcast without interrogation from the ground (1030 MHz).   1090ES offers a wide range of aircraft parameters, notably position.

Flightradar24.com has shared 1090ES position, altitude, ground speed, track heading only.

LMI2933 appears to have been equipped with a form of 1090ES that uses an inertial data source for position.

NOTE: the filed flight plan (see below) makes no mention of ADS-B features.  The ADS-B data appears to be driven from an inertial source of position, which would not qualify as a proper ADS-B system (reliant on GNSS).  1090ES driven by an inertial source of position would not be declared as ADS-B.

Flightradar24 makes mention that the data they provided was from ADS-B (1090ES), and that the source of position was subject to initialization and drift.

LINK TO FLIGHTRADAR24.COM REPORT
The last ADS-B message received by Flightradar24 was at 02:55 UTC from 5.848078,-75.388847, at an altitude of 15,500 feet AMSL, approximately 33 kilometers south of Medellín Airport. 
CP-2933 was equipped with an older ADS-B transponder which may be subject to positional accuracy issues. The position of the aircraft is calibrated prior to take off and the on board computer calculates positions based on speed and direction of the aircraft. With certain types of flying (including holding patterns), the calculation can become inaccurate.


I have examined prior flights of CP2933 looking at both departure and arrival accuracy.

Here are three examples:

Nov 11, CNF-EZE, time enroute 4:04 - accumulated ~ 2,200 meter error

Depart CNF - aligned along runway, departure seems reasonable with no offset

EZE arrival - 2,200 meter offset

Nov 5 VVI-EZE time enroute 3:07 ~ 1,500 meter error


Depart VVI - first ADS-B report was @100kts, threshold estimated based on later flight data
EZE arrival ~ 1,500 meter offset

Nov 4,  MDE -VVI, time enroute 4:33 ~ 3,100 meter offset

CP2933 completed the same route (in reverse) as the accident on Nov 4.  The actual flight time (which does not account for the takeoff and initial departure) was 4:33.  I have seen reports that this was a ferry flight, or without any payload (which may be very reasonable).

The existence of this flight record puts into question 
  1. the actual range of this airplane 
  2. the possibility it was operated beyond legal flight planning on a recurring basis 

Nov 4, CP2933 departed SKRG/MDE for SLVR/VVI.  Flightradar24.com did not assign the departing runway as the flight reporting started already airborne

Arrival to SLVR/VVI with about 3,100 offset

An inertial system should be initialized and aligned so that errors would be minimal upon departure.  Of the four departures, two show errors of 1500-2000 meters.  Normal procedures should emphasize proper alignment prior to departure.

The Inertial Reference System (IRS) provides position, body heading, track heading, and velocities to the GNS-X.  The GNS-X provides the IRS initialization parameters used during pre-flight alignment.


The IRS has four selectable modes.


The IRS internal sensors are unable to resolve longitude.  An accurate position is needed to align the IRS.


The IRS retains a last position (the stored position) from the last time it was in navigating to serve as a blunder test against the entered position.


 The IRS will not accept a bad latitude.


Align downmode, or fast align, is a way to use reset the IRS by "brute force".  Position can be updated, if entered manually.


Inertial position errors follow the Schuler cycle of about 84 minutes; a few measurements are not adequate to guarantee peak values are noted.

An understandable east offset of about 1 nm is evident in the arrival vector of LMI2933 towards the RNG VOR.  The N/S offset cannot be confirmed conclusively.

Arrival errors are the combination of alignment and drift.  It would seem a drift of 2,000 meters is reasonable over 4 hours of flight time.

For LMI2933, an initial error of 2,000 meters coupled with a drift contribution of another 2,000 meters puts the position uncertainty at least at 4,000 meters or about 2 nm.

In any case, the magnitude of the error does not materially change any conclusion or observation.

The FAA has no example of a non-GNSS (inertial in this case) position source that meets the criteria of 14 CFR 91.227 for ADS-B Out.

https://www.faa.gov/documentLibrary/media/Advisory_Circular/AC_20-165B.pdf

Technically, a non-GNSS ADS-B out position source could offer correct reporting by setting the performance parameters to reflect the lack of position confidence.  Multi-lateration techniques have the potential to determine position from a non-GNSS 1090ES by virtue of the transmission itself.

http://www.ecfr.gov/cgi-bin/text-idx?SID=8137158693744ba666e318c1f474d81b&node=se14.2.91_1227&rgn=div8

Table of Contents

GNS-XLS/Collins FMS and Composite Position

By virtue of the equipment listed for LMI2933, there is a clear statement for relatively modern GNSS-based navigation source and VNAV (Flight Management System - FMS).

For this time, only GNS-XLS will be represented.  There are some minor differences with the Collins. FMS.

There is some reason they did not connect GNSS to allow fully compliant ADS-B 1090ES.  By virtue of the flightradar24.com data, the was 1090ES data was driven by inertial data source.

When a moving map is provided, a navigation data base presents the fixed features, and the sensors provides the position and heading source.

GNSS provides track data (position, altitude, track heading, track velocity); a point-mass perspective used for navigation.

IRS provides heading, from which wind drift can be measured along with air data sensors.

Position confidence may be assessed by comparing radio navigation aids to reveal any offset.  Inherent in this calculation is the navigation data base that stores the location of the reference navaid.

The favored update is to take intersecting DME measurements from two VOR/DME.  In the event only one navaid is available, a VOR/DME measurement may suffice.

The RJ85 has four options for lateral navigation (LNAV).  The information LMI2933 provided identifies FMS functionality, but it is not determined the specific configuration.


The FMS stores a Navigation Data Bank (Base) (NDB).  Any association of VOR/DME measurements rely on an accurate NDB.


Waypoints associated to navaids may not be revised.  The flight crew can enter personalized waypoints.  An offset waypoint is used in flight planning as a convenience to the pilot in planning a flight.


The flight crew reference to the Direction Bearing Indicator (DBI) and their dedicated two VOR/DME receivers is entirely separate from the FMS and based on dedicated tuning panels.

The FMS (GNS) relies on an internal VOR/DME receiver to measure radial and distance from up to five different stations simultaneously.  

The FMS tunes the VPU as a function of the NDB and current aircraft position.  

The FMS VOR/DME receivers are used only within the FMS.
Each GNS has a VPU: VPU 1 for GNS 1 and VPU 2 for GNS 2. The VPUs take inputs from DME 1, DME 2, VOR 1 and VOR 2.

Each DME has five channels: channel 1 to channel 5.
  • VHF NAV 1 controller tunes channel 1 of DME 1; 
  • VPU 1 tunes channels 2 to 5 of DME 1. 
  • VHF NAV 2 controller tunes channel 1 of DME 2; 
  • VPU 2 tunes channels 2 to 5 of DME 2. 
VPU 1 uses all five channels from DME 1 and channel 1 from DME 2.
VPU 2 uses all five channels from DME 2 and channel 1 from DME 1.
Thus each VPU has up to six DME ranges to calculate position.

The VOR bearings associated with the frequencies set on number 1 and 2 VHF NAV controllers are also sent to both VPUs.  
However the VOR bearings are only used if the VOR has a co located DME and this DME is being received.

A VPU can operate in two modes: DMEIDME or VOR/DME.
  1. In the DMEIDME mode, fixing is primarily made using the DME ranges. 
  2. In the VOR/DME mode, heavy reliance is made on VOR bearing.
VPU accuracy is normally very good when enough DME stations are in range to allow the VPUs to be in the DME/DME mode and there are good geometric cuts on the circular lines of position.  
In these cases, the accuracy of the VPUs and the accuracy of the GPS are similar.
When DME coverage is sparse, the VPUs may drop into the VORIDME mode.  
In this mode the accuracy is degraded; however it is no worse than pilot navigation using VOR bearing and DME distance.

On most routes, the VPUs will perform well for the bulk of the flight.
However at low altitude, fewer DME stations will be in range and the accuracy may suffer.

The GNS assigns a quality factor to VPU position.  
The quality factor is between 2 and 99 inclusive; the best quality factor is 2; a figure of 99 indicates that the VPU is in its DR mode.

A quality factor reference value can be entered through the CDU.  
If the GNS assigned value exceeds the reference value, a VPU QUALITY sensor message is given.

Composite Position

GPS is the primary source of position for flight management.
If RAIM is available, the composite position is the GPS position.

If RAIM is not available, the position will be a blend of the GPS position and the VPU position.

IRS position is not used in the determination of the composite position.

On the ground after initialization until take off:

The VPU positions are not used because they can be significantly in error on the ground due to multi path reception of the signals.

If the GPS is not available, the composite position is the initialised position modified by integrating the average of the two IRS velocities.

It is important that the ramp position is entered accurately, as it is the starting point for the IRS position.
If both VPUs and the GPS are lost, theGNS composite position becomes:
If the VHF NAV reception is adequate, the VPUs remain in the navigation mode on landing and thus may still contribute to the composite position.
Multi-path reception can corrupt the position; this can be avoided by switching the GNSs off and then on during a turn-round.
If the position needs to be updated before take off, do not use the HOLD facility but perform a downmode align. If the HOLD facility is used, the VPUs will be put in the navigation mode and thus may contribute to the composite position; of course, if VHF NAV signals are not being received, the VPUs will not be used.
The composite position when the last sensor was lost. UPDATED BY the change in position found by integrating the average of the two IRS velocities.

IRS Position Initialization

The FMS provides a helpful service to the IRS by telling the IRS what level of offset (for drift) has been recently received.  This information is consumed by the IRS only during pre-flight alignment.


There are some scenarios that can lead to errors in initialization.


Another sequence that can yield the GNS-X loading a bad position for initialization.


The FMS offers endurance calculations.
PLAN 1 is the FUEL STATUS  page. It displays:
  • The units in use: LB or KG. 
  • The fuel remaining.
  • The pilot entered reserve fuel. 
  • The current total engine fuel flow. 
  • The time left before the reserve fuel quantity is reached. 
  • The distance left before the reserve fuel quantity is reached. 
  • The ground nautical miles per unit mass of fuel. 

Additional items on PLAN 3 are:
  • The total engine fuel flow. 
  • The fuel required to the TO waypoint. 
  • The fuel required from current position to destination following the flight plan route. 
  • The fuel remaining at destination following the flight plan route 

The FMS offers a simple means to insert a holding pattern.
A HOLD key is just to the top right of the LCD; the key has two functions:
  • To access a holding pattern page. 
  • To access a position fix page. 
The holding pattern page is accessed if the cursor is placed over a waypoint and the HOG key is pressed.  
A holding pattern at this waypoint can be set up on the holding pattern page.

The position fix page is accessed if the HOLD key is pressed when the cursor is not over a waypoint.  
The position fix page provides the means of checking and updating the system position.

Table of Contents

Holding Pattern offers Position Reference

The ADS-B reports from LMI2933 during the time it was in the holding pattern are correlated to  GEMLI, used in the RNAV approach.


http://www.aerocivil.gov.co/servicios-a-la-navegacion/servicio-de-informacion-aeronautica-ais/Documents/43%20SKRG.pdf



AVI9771 was in the pattern with LMI2933, possibly holding at RNG VOR. 

In the Google Earth projection below, both flight patterns are shown (red LMI2933, blue AVI9771).  

LMI2933 (red) and AVI9771 (blue) flight paths - LMI2933 position shifted 2.7 nm to offset drift
AVI9771 appears to have taken at least a couple of turns apparently over the RNG VOR. 

LMI2933 orbits 8 nm south at GEMLI.

Table of Contents

Loss of the Moving Map

While a sufficient number of GPS satellites are available for RAIM, the FMS will use GPS position solely as a position reference.  When the number of satellites available for RAIM is inadequate, the FMS will blend the position with radio navigation readings.

The FMS offers RNAV capability.  The approach would be managed in the FMS and displayed on EFIS.

The flight crew had power when they started their emergency descent and were probably thinking they will make it.  They had a reference to the feed tank levels for any tactical reference.

The flight crew were diving for 12,000 feet at GEMLI.  They were set to cross at over 3,000 feet above the approach.

LMI2933 powered rate of descent and configuration would be without regard to the issue of flameout.  The flight crew would most likely have operated with the confidence that the engines will continue to run.

The last ADS-B report showed their rate of descent at around 2,500 fpm (prior to flameout). They were much too high for the approach, and may have even been using spoilers.

Note that FMS, EFIS and DME are lost on EMERG AC/DC power.  The crew would have had to switch to a raw-data ILS approach by reference to the standby attitude indicator as part of their response to the all-engine flameout.

Table of Contents

LMI2933 Flight Plan: SLVR - SKRG alt SKBO

LMI2933 originated in Santa Cruz, Bolivia (SLVR).

The flight to Medellin was estimated at 4:22 minutes.

The flight plan endurance was listed the same 4:22.


Google Earth projection of LMI2933 route based on ADS-B reports


LMI2933 International Flight Plan- original source not determined 
Filing Time:                                28 20:30 
Box 3 Message Type:                 FPL                      Flight Plan
Box 7 Aircraft Identification:     LMI2933
Box 8 Flight Rules:                    I                            IFR
Type of Flight:                           N                           non-scheduled
Type of Aircraft                          RJ85
Wake Turbulence Category        Medium
Box 10 Equipment                     SDFGRWY/S       S=VHF, VOR, ILS
                                                                                 D=DME
                                                                                 F=ADF
                                                                                 G=GNSS
                                                                                 R=PBN
                                                                                 W=RVSM
                                                                                 Y=8.33 kHz VHF Comm
                                                                                 /S=Mode S transponder
                                                                                 <note: ADS-B not declared>
                                                                                 <no 1090ES>
                                                                                 <note: INS not declared>
Box 11 Departure Aerodrome  SLVR                     VVI
Time                                          22:00
Box 15 Cruising Speed             N0380                     380 KTAS
Level                                         F280                        FL280
Route:          UN420 RCO UL417 PABON U0111 8OA DCT MOU DCT RNG DCT
Box 16 Destination                   SKRG
Total Est. Time Enroute            04:22
Alternate                                    SKBO
Box 18  Other Info:                    PBN/B1C1D1S1S2      
                                                              B1=RNAV5 (GNSS, DME/DME, VOR/DME, INS)
                                                              C1=RNAV2 (GNSS, DME/DME, DME/DME/INS)
                                                              D1=RNAV1 (GNSS, DME/DME, DME/DME/INS)
                                                              S1=RNP APCH
                                                              S2=RNP APCH w/Baro-VNAV
                                                    REG/CP2933                 Registration number
                                                    EET/SBAZ0126 SKED0236   Est. Enroute Time
                                                    SEL/GPDK                       Selcal
                                                    PER/C                               ICAO Perf. Category
                                                    OPR/LAMIA                    Operator
Box 19 Endurance                      E/04:22
Persons on board                        P/TBN
Emergency Radio                       VHF, ELT
Survival Equipment                    S/ Dinghies, M Maritime, J Jungle
Jackets                                         J/F
Aircraft Color and Markings      A/WH                              White
Pilot in Command                       C/Miguel Ouiroga Lic 198943
Filed By                                      /s/ 3992715


The great circle route from runway to runway is 2,947 km.

By integrating the ADS-B data, the airplane traveled 3,039 km (1,641 nm) in 4:42 from takeoff to flameout (noting there is a 2 hour gap in the reports that I have straight-lined; there may have been some additional maneuvering).

At it's closest initial approach (as it entered holding) it was just 34 km from the runway, and from there it traveled 80 km to flameout.

Without the holding, the stage length would have been about 2,995 km.

There is little doubt that LMI2933 would have been able to make a successful approach to runway 01 if it had proceeded directly without holding.

The FEED LO LEVEL warning had already been illuminated for at least ten minutes at the point LMI2933 entered holding, signaling a countdown to flameout.

Google Earth projection of LMI2933 route - noting that FUEL LO LEVEL warning would have been set well before arriving at SKRG.  Note the holding pattern, and the locations of the crash site and the threshold to SKRG runway 01
Table of Contents

AVRO RJ85 Type Certificate

EASA published the RJ85 data sheet with some useful references.


https://www.easa.europa.eu/system/files/dfu/EASA-TCDS-A.182_Bae_146---AVRO_146_RJ-02-20102010.pdf
LMI2933, tail CP2933, is s/n E2348, according to flightradar24.com.

https://www.easa.europa.eu/system/files/dfu/EASA-TCDS-A.182_Bae_146---AVRO_146_RJ-02-20102010.pdf
CP2933 would be a series 2 RJ85 build reference 348.

Table of Contents

AVRO RJ85 Performance Data

The following are excerpts from the BAE Systems Technical Data brochure:

BAE Systems Marketing Brochure
http://www.angelwingsva.com/lib/DOWNLOAD/pdf/RJ_Tech_data_at_a_glance__April_2004.pdf

BAE Systems markets the AVRO RJ85 for routes up to 3,000 km.


Instrument flight rules for turbojet aircraft generally mandate enough fuel to fly to the destination, divert to an alternate, and hold for 30 minutes at the alternate.

The flight plan listed SKBO (Bogota), 215 km from SKRG.  215 km at cruise speed is about 20 minutes.  Add 30 minutes of holding, for a reserve of at least 50 minutes beyond the SKRG routing that should have been provided.  In retrospect, LMI2933 would have had less than 10 minutes of fuel if it had arrived straight-away, or it was short at least 40 minutes of flying.

The fuel consumed presumably would have been an entire tank of 11,728 liters over the 4:42 minutes of flying, or about 42 liters per minute.  40 minutes would have required about 1680 liters (the airplane was about 15% short on fuel for a legal flight plan).  Take these numbers as approximations - just in the ball park.

Note, the auxiliary fuel tanks provide an additional 968 liters of usable fuel.  This amount would still not have been enough for a legal flight plan (about 700 liters short).  The data would suggest LMI2933 did not operate with auxiliary fuel, in any case.

Avherald.com offered a modeled RJ85 flight plan to examine the fuel requirements for this flight.  Their conclusion was a more severe 35% shortfall in fuel at release.
LINK TO FLIGHT PLAN

The information is provided at face value, it may not be relevant or accurate.
Great circle route 1606 nm Air distance 1637 nm
Alternate 131 nm
Dry Operating Weight 25,000 kg
Pax 68
Total 6,052 kg payload
4 knot headwind.
Zero Fuel Weight 31,052 kg
Takeoff Weight 43,513 kg
Landing Weight 34,133 kg
Fuel tank capacity 9,362 kg
Fuel to destination 9,380 kg
Fuel allowance for contingency 533 kg
Fuel to fly to SKBO 1277 kg
Fuel for 30 min hold 1271 kg
Taxi fuel 210 kg
Total Fuel for Release 12,671 kg --- note this is includes 3,309 kg over fuel capacity
http://avherald.com/files/lamia_flightplans.pdf

http://avherald.com/files/lamia_flightplans.pdf

http://avherald.com/files/lamia_flightplans.pdf

http://avherald.com/files/lamia_flightplans.pdf


http://www.angelwingsva.com/lib/DOWNLOAD/pdf/RJ_Tech_data_at_a_glance__April_2004.pdf

The RJ85 is offered in standard or optional take-off weight limits.  The optional take-off weight limit adds an additional 4,000 lbs of capacity.  LMI2933 did not require the benefit of optional take-off weight limit, as the payload and full fuel were within normal limits.

http://www.angelwingsva.com/lib/DOWNLOAD/pdf/RJ_Tech_data_at_a_glance__April_2004.pdf

The basic fuel capacity is 3,099 US gallons.  The brochure makes mention of pannier fuel tanks for extended range.  By virtue of LMI2933 apparently encountering fuel exhaustion upon arriving after about a 1600 nm flight, it would appear they were NOT installed.




BAE offers some examples of range capability.  LMI2933 carried 68 passengers and 9 crew for a payload of about 7,400 kg.
http://www.angelwingsva.com/lib/DOWNLOAD/pdf/RJ_Tech_data_at_a_glance__April_2004.pdf

A payload of 7,400 kg would project a range of about 1,375 nm with the above BAE basic assumptions of an additional 30 minutes of holding and 150 nm alternate.  

If auxiliary fuel tanks were available, the range would be about 1,530 nm.  

A flight of 1,600 nm would have restricted payload up to about 2,500 kg.


http://www.angelwingsva.com/lib/DOWNLOAD/pdf/RJ_Tech_data_at_a_glance__April_2004.pdf

In the LMI2933 case, the alternate was only 116 nm away.  

Winds were from 050 True at 20 knots (at 21,000 feet over the holding pattern).

Winds aloft estimate from level holding pattern speed and track measurements
Over 4 hours, a 20 KTAS headwind would add 80 nm to the route - the winds enroute were not determined for this work. With that canceled out, the maximum range attempted should have been as the brochure projects, or less than 1,375 nm - nowhere close to the 1,600 nm along the planned route.

Table of Contents

AVRO RJ85 Fuel Systems

The AVRO RJ85 is a four-engine airplane.  Each engine is fed from a dedicated feed tank.  Normally, each feed tank is kept full (capacity 272 kg each).

The operator's Avro RJ aircraft have three separate fuel tanks: one main tank on each wing and one centreline tank.

The wing tanks comprise three compartments: the main wing tank as well as inner and outer feed tanks.

Each engine has its own feed tank, with a capacity of 272 kg of fuel.

In a normal situation the feed tanks are constantly full of fuel.

RJ85 Fuel Transfer.  Ref Vol. 1 FCOM
The fuel from the centre tank is transferred to the wing tanks from where it is fed to the engines via dedicated feed tanks. 
The wing tank structure is shown in Figure 4.2 as a plan view of the left wing and a vertical section through the left wing looking forward. 
The wing tank is an integral part of its wing. The tank is formed by sealing the volume enclosed by:
  • The front and rear spar. 
  • The top and bottom skin. 
  • The innermost and outermost ribs. 
There are 19 ribs in each wing: numbered 2 to 20 from inboard to outboard. Fuel can flow freely through most of the ribs. The remaining ribs block or restrict the flow of fuel; some of these ribs divide the wing tank into:
  • The main compartment. 
  • The two feed tanks. 
  • The surge tank. 
Each feed tank contains a pump compartment; it contains the electrically driven pump. 
Rib 13 forms the boundary between the main compartment and the inner feed tank. The rib is solid and sealed apart from some small vent holes at the top and some flap valves at the bottom. The flap valves allow gravity transfer into the feed tank but prevent flow back to the main compartment.

Rib 15 is the boundary between the inner and outer feed tanks. It is solid and sealed everywhere apart from at the very top. It acts as a high level weir; fluid can move between the two feed tanks across the weir. Generally, the flow is from inner to outer feed tank.

Rib 18 is solid and completely sealed. It is the boundary between the surge tank and the outer feed tank.

Ribs 6 and 1oare baffle ribs. They restrict span wise rate of fuel flow to reduce fuel slosh.
Each pump compartment is formed by the lower part of rib 15 and a rib that is open apart from its lower portion: rib 14 for the inner pump compartment and rib 16 for the outer pump compartment. Each of these ribs has flap valves at the bottom which allow flow into the pump compartment but not out of it.

In the vertical section of Figure 4.2, dashed line 1 is at the height of the top of rib 13; dashed line 2 is at the height of the top of the weir. Line 1 is equivalent to a wing fuel quantity of about 2 000 kg (4 400 lb); line 2 is equivalent to about 1 300 kg (2 860 lb).

The feed tanks can be kept full by gravity transfer alone provided the level is above dashed line 1. Below line 1, the jet pumps are required to keep the feed tanks full.

If the level falls below line 2, no more fuel can enter the outer feed tank by gravity transfer. Without jet pump transfer, the outer feed tank is isolated when the wing tank level falls below line 2.

RJ85 Fuel Transfer.  Ref Vol. 1 FCOM
Feed Tanks,  Ref Vol. 1 FCOM


RJ85 Fuel Panel  Ref Vol. 1 FCOM

RJ85 Fuel Panel Ref Vol 1 FCOM
Table of Contents

AVRO RJ85 Fuel FEED LO LEVEL Warning

In the event of fuel exhaustion,  a warning is raised for each wing once any feed tank begins to drain.
If the amount of fuel in a feed tank drops below 272 kg, the aircraft's fuel panel annunciates a FEED LO LEVEL warning.

Feed Tank Quantity Gauges and L/R FEED LO LEVEL Warning.  Ref Vol 1 FCOM

The warning (FEED LO LEVEL) indicates that flameout is imminent.  Each feed tank can power each engine for at least 23 minutes in cruise -or- enough fuel to make an approach, miss the approach, and land in the second approach.  

23 minutes back from flameout puts LMI2933 150-250 km from SKRG, while the flight was in cruise.

The flight crew would have been alerted to their dire situation at least ten minutes before they started their descent as each engine FEED LO LEVEL warning was triggered.  The fuel situation must have been glaring during their turns in the holding pattern.  First they would have noted the fuel tanks draining, then the FEED LO LEVEL warnings.  Finally the Lo Press warnings.

The AVRO RJ85 suffered from FEED LO LEVEL warnings in other circumstances.  A investigation report from an event in 2009 provides a reference for the fuel system.

An AVRO RJ85 very similar to flight LMI2933 suffered FUEL LO LEVEL warnings  due to fuel contamination.  While the circumstances are not the same, the report documented the BAE jet fuel system features.

Incident aircraft E2389, a series 2 RJ85 build sequence 389 (delivered 2001).

If the warning does not clear as a result of checklist action the flight crew must land as soon as possible.

A full feed tank will supply fuel for one engine for a minimum of 23 minutes of continuous operation at cruise power, or for a descent from high level, an approach, a go-around and a further approach to landing.

The checklist notes that in an abnormal situation the pilots can only rely upon the fuel in the feed tanks on the affected side.

http://www.skybrary.aero/bookshelf/books/2197.pdf

RJ85 All Engine Out Considerations


This section relies on an unofficial copy of Volume 1 of the flight crew operations manual (FCOM) to the RJ85.  I cannot attest to its accuracy or relevance, but I do profess that it is probably pretty accurate.


Hydraulics

The RJ85 has two hydraulic systems: Yellow and Green.

The Yellow system normally is driven by Engine 2, with an AC pump backup in the case of Eng 2 failure.

The Green system is driven by Engine 3.  The backup to the green system is a pump driven by the Yellow system.

The Green system runs a backup generator (AC/DC).



The Yellow system has a DC MOTOR that can operate on battery power.   The DC Motor offers hydraulic power to assist lowering the gear and for braking only.  The crew must select battery power to activate the pump.


No other hydraulic system is powered with all engines off. 

In the event all engines are lost, windmilling will deliver limited hydraulics.

However, the Yellow system is apparently bogged down and will not drive the rudder.

The Green system can only drive the rudder if (1) the windmilling is sufficient and (2) the crew select the standby generator off.



Guidance for loss of engines and APU.
RJ85 FCOM Vol 1

If the crew does not take this action, they may lose control of the rudder.  The engine flameouts may have been asymmetric, causing yawing moments. 


Electrical

The RJ85 has two engine-driven generators, an APU, and a hydraulic-powered standby generator plus one or two batteries.

The APU may be started inflight as high as 20,000 feet altitude and can operate as high as 25,000 feet.


With no fuel, and with no engines or APU running, the only remaining source of power is one battery and one optional battery.  One battery should be sufficient for about 30 minutes.

The EMERG DC bus is powered by a battery.  A standby inverter uses the EMERG DC bus power to generate EMERG AC bus power.   

The ESS DC bus can be tied to the EMERG DC bus if the DC BUS-TIE is selected to OPEN.



It is not apparent whether a choice was made to utilize ESS DC when all engines failed.  There appears to be no operational benefit, and it would tax the batteries.






Flight Controls

Elevator trim tab

The pitch elevator tabs operate mechanically (and without regard to loss of hydraulic power).

Roll Aileron trim tab and Spoilers

The roll aileron tabs operate mechanically (and without regard to loss of hydraulic power).

Roll spoilers are powered by the yellow hydraulics and may be serviceable from the number 2 windmilling engine (not confirmed).
Control in roll is by two servo tab operated ailerons and two hydraulically powered roll spoilers. The handwheels are mechanically connected to the roll spoiler actuators and to the servo tabs not to the ailerons. The aerodynamic feel of the servo tabs is enhanced by a spring.

Rudder

Rudder relies on either Yellow or Green Hydraulic power.  
Control in yaw is by a hydraulically actuated rudder. 
There are two hydraulic rudder actuators. One is powered by the green system and one by the yellow system. Either actuator provides adequate control in yaw. 
The actuators are mechanically signalled by a mechanical summing unit. The summing unit takes inputs from:
  • The rudder pedals. The maximum rudder pedal input is reduced as speed increases by a rudder limiter. The rudder limiter is positioned by a rudder 0-pot.
  • A manually operated rudder trim wheel on the centre console. There is no electric rudder trim.
  • A yaw damper (YD)

 

With loss of fuel, the number 3 windmilling engine can drive the rudder if the crew selects the standby generator off.
Loss of All Engines and APU not Running
If all engines are lost and the APU is not running, the electrical system will descend to the emergency power level.

If an inboard engine is windmilling with its engine driven pump on, a limited amount of hydraulic power is available.

If the standby generator switch is at ARM, the standby generator will automatically be signalled to run.
If engine 3 is windmilling, the standby generator motor will absorb all of the green hydraulic power but will not achieve the minimum speed to allow the generator to come on line; the isolation valve will be closed, thus the other green services will be isolated.

However, EMERG DC is available, so the standby generator valve and the isolation valve can be controlled. The standby generator should be selected OFF so that any green hydraulic power is available for rudder operation.

Hydraulic power is not available to the rudder from engine 2 because the power is absorbed by the standby fuel pumps. When main DC is lost, the standby fuel pumps:
• Automatically run
• Cannot be selected off.

Flaps 

Flaps rely on both Yellow and Green Hydraulic power to operate at full speed.  Half-speed operation is possible on one hydraulic source.  There is no provision to move flaps under windmilling hydraulic power.

Yellow hydraulics flap actuators control is powered by EMERG AC and EMERG DC.

Landing Gear and Brakes

The landing gear can be deployed using a manual handle release and EMERG Yellow hydraulic power.  Braking is available from EMERG Yellow hydraulic power.  There is no provision to raise the landing gear while under EMERG Yellow hydraulic power.


Lift Spoilers

Lift Spoilers are locked down.

The yellow lift spoiler actuator system control is powered by EMERG AC and EMERG DC


Airbrake

The airbrake may trail slightly open.

The airbrake actuator control is on DC 2, and would be lost in emergency power.
When hydraulic power is lost, air loads will move the airbrake towards the fully in position. The airbrake may trail slightly open. In this case the AIRBRK annunciators will remain illuminated, but the drag will be small.

When the command signal to the servo valve is lost, the airbrake moves slowly to fully in. When electrical power is lost, the airbrake moves slowly to fully in.

Stall Warning

The vane heaters are powered at the essential power level as are both channels of the stall protection system. The stall protection system is completely serviceable at the essential power level. 
The stall protection system is not available at the emergency power level.



Communication

The audio "A" channel operates on the EMERG DC battery bus.  Passenger Address, VHF 1 and RMP 1 are functional.

The right flight deck speaker and mic interfaces are lost.

ACARS, HF and SELCAL do not operate on EMERG DC or EMERG AC power.



Lighting

Flight Deck Lighting

Flight Deck Emergency Lights are automatically activated if ESS DC is lost by connection to the EMERG BATT bus.

Flight Deck Emergency Lighting

None of the remaining flight deck lighting would operate on emergency power.  Instrument panel and center console flood lights could be powered if ESS DC was manually activated.

Instrument panel, console lighting, storm, flight deck entry lights are all on AC Bus 1 or AC Bus 2.

Instrument panel and Center Console flood lights are on ESS DC.

Standby Compass has an integral light on DC BUS 1.

Sill and Lap lights, Flight Kit lights are on DC Bus 1 and DC Bus 2.

Cabin Lighting

Cabin Emergency Lights (aisle, exit) operate on internal batteries. It is possible to manually power them from BATT 1 bus as long as BATT 1 voltage is above 18V.

None of the remaining cabin lighting would operate on emergency power.  The lavatory standby light could operate if ESS DC was manually activated.

Lavatory normal lighting is on AC Bus 2.  Lavatory standby incandescent light is on ESS DC.

All cabin signs are on DC Bus 2

Half of the Center Cabin lights are on ESS AC.  The remainder of Center Cabin lights, Side, Reading, and Vestibule lights are on AC Bus 2 (inflight).

Exterior Lighting

None of the exterior lights would operate on emergency power unless ESS DC was manual activated.

Wingtip and Tail Nav lights have two independent lights, one on ESS DC the other via AC Bus 2.

Strobe lights are on AC Bus 2.

Anti-Collision beacons have two configs: either powered ESS AC or powered ESS DC.

Left landing and taxi lights are on ESS AC, the right side on AC Bus 2.

Left wing inspection light is on DC Bus 1, the right wing on DC Bus 2.

The Logo lights are on AC Bus 2.

Both Air Data Computers are not functional on EMERG DC/AC power.  

A standby indicator provides altitude and airspeed.


IRS 1 can operates on BATT 1 power.


ILS/VOR/MB/LOC/ADF 1 are operable with EMERG AC/DC.

EFIS, DME, Radio Altimeter, Temperature, Transponders, TCAS, EGPWS, WXR Radar are lost on EMERG AC/DC power.

IRS 1 supplies heading under EMERG AC/DC power to the left VOR/DME/NDB Direction/Bearing Indicator (DBI).











It may be possible the navigation unit operated for one minute after power loss (it has an internal battery).  Otherwise it is lost.







Standby attitude, LOC/GS deviation indicators.



Magnetic compass


Master warning and status annunciators operate partially on EMERG DC power.

The clock runs, but the backlighting is lost.

The flight data and voice recorders run on ESS AC power, flight data acquisition unit on ESS DC, and all would likely stop operating when the engines flamed-out.  Note that the electrical sources for the flight data acquisition units may not be operable on EMERG AC/DC power.

There are four indicating and recording system circuit breakers on the COMMUNICATION panel.
  • VOICE REC (827) supplies power to the CVR
The only control in the flight deck that would stop the CVR from operating (in-flight) is to pull the VOICE REC circuit breaker.

The FAA has mandated a 10 minute Remote Independent Power Supply (RIPS) for newly manufactured aircraft.  They did not mandate its retrofit.  I am doubtful LMI2933 was modified to introduce a RIPS voluntarily; I also am unfamiliar with their certification basis.

The following is from the FAA NPRM proceeding.  This is where they decided to not force the retrofit.



FAR 25.1457 covers cockpit voice recorders.  (d)(5) brings forth the RIPS requirement.


FAR 121.359 (US flag carriers) covers cockpit voice recorders.  (i) applies to new rules for retrofit application do not express a mandate to comply with 25.1457 (d)(5).


FAR 121.359 (US flag carriers) covers cockpit voice recorders.  (j) applies to new rules for newly manufactured airplanes that does express a mandate to comply with 25.1457 (d)(5).



Engine-Out Procedures

Thanks to AerocatS2A and pprune.org, I have two matching references for the glide ratio related to the RJ85.

Glide ratio is 15:1 for flaps/gear up
Glide ratio with gear down (flaps up) is 6:1
Glide ratio with flaps/gear down is not stated.

http://www.pprune.org/rumours-news/587574-jet-goes-down-its-way-medellin-colombia-41.html





Loss of engines means loss of pneumatics, which means loss of pressurization.  Drop the oxygen masks and descend at VFTO+30.

Initial procedure is to drop oxygen masks (loss of pressurization with engine loss) and establish max glide speed.
https://quizlet.com/136145662/rj85-memory-items-flash-cards/
With terrain a factor, a landing site is the priority.

Maintain VFTO+30 until landing area is assured.

Then deploy flaps and gear - noting that in the loss of fuel case, no APU, it may not be possible to lower flaps.

Configure for EMERG DC power, EMERG Yellow hydraulics, standby gen OFF.

For loss at low altitude, glide ratio becomes a factor, hence flaps/gear only as required.  

With loss of fuel includes loss of APU as a backup.  This puts the aircraft to rely solely on battery power.

https://quizlet.com/136145662/rj85-memory-items-flash-cards/


https://quizlet.com/136145662/rj85-memory-items-flash-cards/

Table of Contents

Final Maneuvering

LMI2933 entered a holding pattern while waiting to land, presumably at GEMLI.

LMI2933 departed the holding pattern by descending once turning inbound on their second circuit.

The ATC transcript indicates the intention to fly direct to RNG VOR and intercept the localizer inbound.  This was the obvious intention.

The airplane glide left no altitude by the time they arrived over the VOR and crashed in close proximity.



Minimum terrain avoidance altitudes in the vicinity of RNG VOR.  Notably, 12,000 feet from the point LMI2933 left the holding pattern and 11,000 feet in proximity to RNG VOR.

http://www.aerocivil.gov.co/AIS/AIP/AIP%20Generalidades/Aerodromos/43%20SKRG.pdf

From the holding pattern, it appears LMI2933 may have initially planned an RNAV approach.  With the loss of engines, RNAV capability is lost.  With total loss of power, the crew would have had the option to fly VOR, LOC, and glide slope using the standby attitude indicator and direction bearing indicator.  ATC clearance to hold is not known.

Once LMI2933 started to descend, they stated that they planned to fly inbound on the localizer.


SKRG ILS 01 approach
http://www.aerocivil.gov.co/servicios-a-la-navegacion/servicio-de-informacion-aeronautica-ais/Documents/43%20SKRG.pdf

LMI2933 hold and final approach (red) RNAV approach in blue.

The RNAV waypoint KUGNU sits 1.5 nm north of the VOR at 9,700 feet.

http://www.aerocivil.gov.co/servicios-a-la-navegacion/servicio-de-informacion-aeronautica-ais/Documents/43%20SKRG.pdf

The ILS Final Approach Point is about 1/2 nm north of the VOR at 10,000 feet.

http://www.aerocivil.gov.co/servicios-a-la-navegacion/servicio-de-informacion-aeronautica-ais/Documents/43%20SKRG.pdf

Both the RNAV and ILS approaches follow nearly the same flight path.  The RNAV approach is shown for convenience.

LMI2933 was over 3,000 feet above the glide path when crossing GEMLI, presumably at flameout.  They crossed through the RNAV approach at about 10,500 feet (estimated) and still about 3 nm short of the FAP (KUNGU).
LMI2933 (red) and RNAV approach (blue) shown to the runway


A 3 degree glide path is a glide ratio of 19:1, beyond the power-off capability of pretty much any airplane.

From GEMLI, they were about 16 nm to the runway.  They had descended already to about 15,000 feet, leaving about 8,000 to the runway.

It may be possible to deadstick in a clean configuration (2.5 nm per 1,000 feet) from GEMLI at 15,000 feet.

With gear down, at 1 nm per 1,000 feet, their glide range was only half what was required from GEMLI.  They persevered in the last three minutes of their deadstick descent along the approach path.

The RJ85 has no provision to raise the flaps or the gear with engines off.  With flaps and gear down at flameout, the crew were stuck.

From the last ADS-B reported position to the point of impact shows about a 6:1 glide ratio, as predicted with gear down.

The impact point is very near the VOR, giving credit that the crew were faithfully steering straight for the localizer.  The point of impact was regrettably the last ridge they needed to cross, having cleared it, there would have been about another 1,500 feet of elevation to glide, and possibly a more favorable impact zone.

Table of Contents

Conclusion

LMI2933 planned a route that was probably more than 300 km beyond what should have been allowed (assuming a standard RJ85 or even with pannier tanks).

The flight crew would have been alerted at least 23 minutes prior to flameout (and more likely even earlier) to impending fuel exhaustion.  

LMI2933 entered the holding pattern having come as close as 34 km to runway 01.

The first circuit around the holding pattern consumed their last margins.




Taking a close look at the final descent reveals the airplane being reconfigured into landing configuration.  The transcript makes mention of "gear down" in a radio call.

Initially in the descent they held speed, then the airplane pitched up slightly as the airplane was decelerated to approach speed (142 KTAS ground speed).

It was noted that the airplane would have been into about a 20 knot headwind, making the airspeed more like 160 KTAS - or about 125 KCAS (airspeed) at 16,000 feet.


The flight crew were still operating with a fully operable airplane and were following a natural progression to configure for a stable capture of the approach path at 9,700 feet at the FAP.

In hindsight, with engine failure imminent, a more prudent course of action may be to fly a route consistent with engine failure.

In any case, up until the engines flamed out, the flight crew appear to be operating under the presumption of continued flight-worthiness.  The crew have dedicated fuel feed tank gages to assess their endurance.

The flight crew crippled their glide ratio by configuring for landing.

The achieved glide ratio from flameout to crash is less than 6:1.  This turns out to be about as expected with gear down.

The following plots glide ratio and altitude during the final approach.  These flight segments were flown under power, with the assumption engine failure was just after the last reported position.  The last segment to the crash site was straight-lined given no other information, and based in part on the ATC transcript.


The flight crew worsened their glide options slightly from the time they left the holding pattern to the point of flameout.

They would have needed about 11:1 to make it from the point they started to descend from the holding pattern, but by the time they flew to their last reported position, they would have need 13:1 to make the airport.

If they had proceeded to runway 01 at the end of the prior inbound leg, the needed glide ratio would have been less than 8:1.

If they had stayed clean and maintained 21,000 feet until flameout, the glide ratio may have gotten to as little as 9:1.

The RJ85 flaps/gear up glide ratio is 15:1.  In every case, it appears that LMI2933 had the capacity to glide to the airport by remaining in clean configuration and appropriate speed.

The flight crew had no option to raise flaps and gear once the engines failed.

S-turns, circling, tailwind landing, or even a slip if needed can correct an overshoot.  Overshoots are much easier to overcome.


Point of Impact

LMI2933 came to rest after hitting a ridge at about 8,640 feet, in close proximity to the RNG VOR.  

from 2.75 nm south of the point of impact

Terrain in the vicinity south of SKRG has a mix of flat land around 7,000 feet with mountains over 8,000 feet in all directions.

http://www.aerocivil.gov.co/servicios-a-la-navegacion/servicio-de-informacion-aeronautica-ais/Documents/43%20SKRG.pdf



Had LMI2933 cleared this last range, they may have been able to glide to a more suitable landing zone at about 7,600 feet and 1 nm south.





Note additional information below

  1. Incident report from another RJ85 that ran low on fuel
  2. Google Earth views along the final approach
  3. Plots of ADS-B data
  4. Raw ADS-B data table

Stay tuned!

Peter Lemme
peter @ satcom.guru

Follow me on twitter: @Satcom_Guru

Copyright 2016 satcom.guru     All Rights Reserved

Peter Lemme has been a leader in avionics engineering for 35 years. He offers independent consulting services largely focused on avionics and L, Ku, and Ka band satellite communications to aircraft. Peter chairs the SAE-ITC AEEC Ku/Ka-band satcom subcommittee developing PP848, ARINC 791, and PP792 standards and characteristics. 

Peter was Boeing avionics supervisor for 767 and 747-400 data link recording, data link reporting, and satellite communications. He was an FAA designated engineering representative (DER) for ACARS, satellite communications, DFDAU, DFDR, ACMS and printers. Peter was lead engineer for Thrust Management System (757, 767, 747-400), also supervisor for satellite communications for 777, and was manager of terminal-area projects (GLS, MLS, enhanced vision).

An instrument-rated private pilot, single engine land and sea, Peter has enjoyed perspectives from both operating and designing airplanes.  Hundreds of hours of flight test analysis and thousands of hours in simulators have given him an appreciation for the many aspects that drive aviation; whether tandem complexity, policy, human, or technical; and the difficulties and challenges to achieving success.


Table of Contents

Deja Vu?

Aviation is loathe to repeat catastrophic history.  The whole point of air safety investigation is to be sure we learn and fix.  In 2010 a City Jet RJ85 was inbound to Zurich and had to divert to Basel due to weather.  Upon arriving at Basel, the crew ended up declaring an emergency for fear that traffic ahead would delay their ability to land above minimum required fuel.  Here are the highlights from the report.


From the report: 

At 17 h 26 min 16, while the BAe146 was 1.8 NM from the runway threshold, the crew of the A319 not having taken off, the tower controller told them : “stop immediately, hold position, repeat, stop immediately, a BAe 46 on go-around“. Then he asked the crew of the BAe to make a go-around. The latter refused because they did not have enough fuel and requested that the A319 vacate the runway. 

Note: At this time and according to the airline’s analysis, the quantity of fuel remaining was estimated at 1,400 kg. The nal reserve is de ned as 850 kg. 

At 17 h 26 min 36, the controller ordered a go-around, which the crew performed. 

At 17 h 26 min 58, the crew stated: “we are declaring a fuel emergency now we request priority vectors for landing“. 

The tower controller contacted the approach controller by telephone. They decided to have the airplane climb to 6,000 ft on the extended runway centreline and to “make it as short as possible “. The tower controller asked the crew to climb to 6,000 ft and to change frequency. 

At 17 h 28 min 23, the crew of the BAe 146 contacted approach control: “Mayday Mayday Mayday, City 108X, declaring fuel emergency, request priority landing“. 

After ensuring that they had the runway in sight, the approach controller offered the crew of the BAe 146 a visual approach, which was accepted. 

At 17 h 34, the crew landed.

On the ground, the quantity of fuel remaining was 1,220 kg. 

Regulatory requirements for air traffic control 

The definition of minimum fuel is not integrated in the French regulations (Decree of 3 March 2006 modified in relation to the rules of the air and to air traffic control services (RDA). 

This definition is however mentioned in ICAO Doc 4444 (15th edition, 2007) – Procedures for air navigation services– Air Traffic Management – chapter 1 definitions: 
Minimum fuel. The term used to describe a situation in which an aircraft’s fuel supply has reached a state where little or no delay can be accepted. 

Note: This is not an emergency situation but merely indicates that an emergency situation is possible, should any undue delay occur. 
A national DSNA instruction dated 6 July 2004 reminds air traffic control organisations of the requirements relating to a minimum fuel situation or to an emergency situation. 

This DSNA instruction was distributed locally to controllers at Basel-Mulhouse- Freiburg in the form of a service memo dated 8 September 2004, and concludes with: “this information does not thus lead to the granting of any type of priority“. 

Regulatory requirements for public transport operators 
ˆThe EU OPS regulations state in paragraph OPS 1.375 part b) 3) that: 
“The commander shall declare an emergency when calculated usable fuel on landing, at the nearest adequate aerodrome where a safe landing can be performed, is less than final reserve fuel.“ 
An emergency situation can be transmitted either by an urgency “PAN PAN“ message or by a distress “MAYDAY“ message. 

Airline’s Operations Manual 
The airline’s operations manual (Part A) states: 
“A fuel emergency exists when it is estimated to have reduced to an amount where an approach and landing should be commenced without delay. The amount of fuel remaining at this stage is 850 kg. 
In the case of the RJ this equates to the OPS minimum reserve fuel, which is sufficient fuel for holding for 30 min at 1,500 ft. 
A Mayday shall be declared if it is estimated that the aircraft will land with less than the OPS minimum reserve fuel.  
This fuel is not to be considered as a separate requirement in the fuel planning process. 
The 850 kg is useable fuel and may be consumed as part of the arrival procedure at any aerodrome.  
When it is estimated that the fuel remaining upon landing will be reduced to an amount of 1,200 kg or less then the Commander shall declare a PAN.  
This fuel is not to be considered as a separate requirement in the fuel planning process. 
The 1,200 kg is useable fuel and may be consumed as part of the arrival procedure at any aerodrome. The requirement that crews declare a PAN or Mayday ensures crews benefit from the priority ATC will place on an aircraft declaring such emergency.“ 
CONCLUSIONS 
The incident was due to the late communication by the crew to the Air Traffic controller of their low fuel situation and their emergency situation. This led to the controller being unaware of the emergency situation. 

The following element contributed to the event: 
  • The lack of an appropriate “minimum fuel“ procedure associated with the remaining flying time. 
Notion of minimum fuel 
The notion of minimum fuel defined by ICAO allows a crew to describe to the air traffic services a potentially critical situation during a diversion while avoiding the declaration of a distress or emergency situation. 

This notion of minimum fuel is not defined in the European regulation. 

In its report on the serious incident on 28 August 1999 at Paris Charles de Gaulle (95) to the Boeing 737-528 registered F-GJNF operated by Air France, the BEA had already recommended that the DGAC define the “Minimum fuel“ callout. In answer to this recommendation the DGAC considered that: “The minimum fuel callout is a source of confusion. This callout does not lead to any action by ATC, so the crew must then declare a distress situation as soon as the quantity of fuel planned for the landing is lower than the final reserve“. 

In the light of this event, the BEA recommends that: 
  • €the DGAC and EASA implement the “minimum fuel” message already defined by ICAO, with the associated procedures. [Recommendation FRAN‐2012‐026]. 


Table of Contents

Google Earth Plots 

The following projections are based on ADS-B reported position offset about 2.7 nm to correct for observed drift (and may still not be an accurate representation).  Note the last reported position is highlighted (the airplane icon) and the path to the crash site is straight-lined, having no other information to utilize.

LMI2933 last reported position shown above the Blue RNAV approach

LMI2933 was over 3,000 feet above the glide path when crossing GEMLI, presumably at flameout.  They crossed through the RNAV approach at about 10,500 feet (estimated) and still about 3 nm short of the FAP (KUNGU).

View along LMI2933 final descent, with last reported position, crash site, and SKRG pin-pointed

LMI2933 (red) and RNAV approach (blue) shown to the point of impact

LMI2933 impact (red).  RNAV approach in blue

LMI2933 impact

Table of Contents

ADS-B Data

The following are plots from the raw (not corrected for IRS drift) ADS-B data used to analyze the accident.
 












Table of Contents

ADS-B data from flightradar24.com

There are 275 points from taxi, takeoff, enroute, and approach.  The end is the top, the beginning the bottom.  This forms the basis of the analysis.  


The final data points are provided below.  


UTC Position Altitude Gnd Speed Direction
Pressure Nmi/hr km/hr Track, True
2016-11-29T02:55:48Z 5.848078,-75.388847 15550 142 263 3
2016-11-29T02:55:33Z 5.838272,-75.389717 16175 139 257 4
2016-11-29T02:55:02Z 5.818085,-75.391113 17325 144 267 2
2016-11-29T02:53:58Z 5.765259,-75.39489 19025 221 409 3
2016-11-29T02:53:52Z 5.759024,-75.395241 19250 226 419 3
2016-11-29T02:53:46Z 5.752553,-75.395569 19525 230 426 3
2016-11-29T02:53:40Z 5.746129,-75.396095 19775 234 433 5
2016-11-29T02:53:34Z 5.739612,-75.396896 20050 236 437 11
2016-11-29T02:53:28Z 5.732986,-75.39875 20425 234 433 19
2016-11-29T02:53:22Z 5.726074,-75.40168 20750 234 433 28
2016-11-29T02:53:15Z 5.720386,-75.40509 20950 235 435 35
2016-11-29T02:53:09Z 5.715729,-75.408852 21000 239 443 69
2016-11-29T02:52:48Z 5.707537,-75.43071 21000 245 454 95
2016-11-29T02:52:42Z 5.709074,-75.43734 21000 249 461 111
2016-11-29T02:52:36Z 5.712286,-75.444016 21000 252 467 124
2016-11-29T02:52:30Z 5.716736,-75.449402 21000 255 472 137
2016-11-29T02:52:24Z 5.722549,-75.453682 21000 259 480 150
2016-11-29T02:52:17Z 5.729874,-75.456612 21000 266 493 164
2016-11-29T02:51:13Z 5.811069,-75.45652 21000 276 511 181
2016-11-29T02:51:07Z 5.818983,-75.456329 21000 276 511 181
2016-11-29T02:51:01Z 5.827083,-75.455994 21000 275 509 183
2016-11-29T02:50:55Z 5.834793,-75.455315 21025 275 509 186
2016-11-29T02:50:49Z 5.842818,-75.454147 21000 277 513 192
2016-11-29T02:50:43Z 5.849894,-75.45211 21000 280 519 203
2016-11-29T02:50:36Z 5.857132,-75.448006 21000 282 522 216
2016-11-29T02:50:30Z 5.864185,-75.441742 21025 283 524 230
2016-11-29T02:50:05Z 5.872375,-75.412201 21000 274 507 280
2016-11-29T02:49:59Z 5.870361,-75.405037 21000 269 498 293
2016-11-29T02:49:53Z 5.866792,-75.39846 21000 266 493 305
2016-11-29T02:49:47Z 5.86171,-75.392792 21000 257 476 323
2016-11-29T02:49:41Z 5.855347,-75.38842 21000 253 469 333
2016-11-29T02:49:35Z 5.849304,-75.385994 21000 246 456 346
2016-11-29T02:49:01Z 5.811209,-75.384918 21000 244 452 1
2016-11-29T02:48:55Z 5.804077,-75.385254 21000 244 452 2
2016-11-29T02:48:49Z 5.79715,-75.385582 21000 243 450 4
2016-11-29T02:48:43Z 5.790121,-75.386292 21000 241 446 6
2016-11-29T02:48:37Z 5.783386,-75.387299 21000 241 446 10
2016-11-29T02:48:00Z 5.744313,-75.400635 21000 242 448 19
2016-11-29T02:47:54Z 5.737885,-75.402893 21000 241 446 18
2016-11-29T02:47:48Z 5.731558,-75.404808 21000 242 448 15
2016-11-29T02:47:42Z 5.724854,-75.4067 21000 241 446 17
2016-11-29T02:47:36Z 5.718338,-75.409225 21025 238 441 27
2016-11-29T02:47:30Z 5.713028,-75.412811 21000 237 439 41
2016-11-29T02:47:23Z 5.707817,-75.418633 21000 237 439 57
2016-11-29T02:47:17Z 5.70488,-75.424446 21000 238 441 71
2016-11-29T02:47:11Z 5.703461,-75.431061 21000 240 444 85
2016-11-29T02:46:48Z 5.71431,-75.455917 21000 263 487 159
2016-11-29T02:46:23Z 5.743475,-75.463051 21000 276 511 180
2016-11-29T02:46:08Z 5.763493,-75.462669 21000 277 513 181
2016-11-29T02:45:03Z 5.844866,-75.455521 21025 276 511 194
2016-11-29T02:42:41Z 5.753577,-75.3825 22625 292 541 359
2016-11-29T02:42:29Z 5.737237,-75.382027 22950 293 543 357
2016-11-29T02:41:01Z 5.611174,-75.37516 24000 329 609 356
2016-11-29T02:39:52Z 5.49754,-75.368813 24225 362 670 356
2016-11-29T02:39:31Z 5.465093,-75.36692 24725 362 670 356
2016-11-29T02:38:25Z 5.3535,-75.362625 25025 367 680 358
2016-11-29T02:38:19Z 5.34288,-75.362251 25000 367 680 357



RNG APP 0:00:000 RUIDOS DE TECLADO

RNG APP 0:03.351 LIMA MIKE INDIA DOS NUEVE TRES TRES NOTIFIQUE RUMBO

LMI-2933 0:10.532 UNO.. UNO SIETE NUEVE EN ALEJAMIENTO

RNG APP 0:13.883 MANTENGA PRESENTE RUMBO Y ESPERE PARA

INICIAR EL.. PARA CONTINUAR SU DESCENSO

LMI-2933 0:19.149 MANTENEMOS PRESENTE RUMBO Y ESTAMOS ATENTOS

PARA EL VOR

RNG APP 0:28.021 ( RUIDOS TECLADO)

AV-9771 0:29.936 RIO APROXIMACIÓN AVIANCA 9771 ES POSIBLE CONTINUAR

NUESTRO ALEJAMIENTO HASTA LA MILLA ..CINCO MILLAS

MÁS HASTA LA MILLA QUINCE?

RNG APP 0:41.905 NO CAPITÁN, NECESITO QUE INICIE EL ACERCAMIENTO LA

AERONAVE ADELANTE SUYO ESTÁ A UNA MILLA NECESITO

INICIAR EL DESCENSO ESA AERONAVE SE DECLARO CON

PRIORIDAD PARA EL ATERRIZAJE

AV-9771 0:52.916 ENTERADO ENTONCES AHORA INICIAMOS EL ACERCAMIENTO

AVIANCA NUEVE SIETE SIETE NU..SIETE UNO

RNG APP 0:57.704 SI GRACIAS POR FAVOR

LMI-2933 1:05.364 LIMA MIKE INDIA DOS NUEVE TRES TRES SOLICITA VECTORES

PARA ACERCAMIENTO SEÑORITA

RNG APP 1:10.630 ATENTO TENGO UNA AERONAVE POR DEBAJO DE US.... POR

DEBAJO SUYO EFECTUANDO LA APROXIMACIÓN Y ADICIONAL

ESTÁN EFECTUANDO UNA REVISION DE PISTA

RNG APP 1:18.769 QUÉ TIEMPO TIENE PARA PERMANECER EN SU APROXIMACIÓN

LIMA MIKE INDIA?

LMI-2933 1:23.556 ...TAMOS EH CON EMERGENCIA DE COMBUSTIBLE SEÑORITA

POR ESO LE PIDO DE UNA VEZ CURSO FINAL

RNG APP 1:31.973 AVIANCA NUEVE TRES CINCO SEIS INICIE EL ACERCAMIENTO

AHORA

AV-9356 1:35.324 INICIANDO ACERCAMIENTO AHORA UNO NUEVE CERO AVIANCA

NUEVE TRES CINCO SEIS

LMI-2933 1:41.548 SOLICITO DESCENSO INMEDIATO LIMA MIKE INDIA DOS NUEVE

TRES TRES

RNG APP 1:45.132 LAN COLOMBIA TRES CERO DOS CERO CANCELE LA

AUTORIZACIÓN DE APROXIMACIÓN VIRE POR SU IZQUIERDA

CON RUMBO CERO UNO CERO AHORA ( LO SACUDIÓ VOZ DE

FONDO VARONIL)

LA 3020 1:52.792 POR LA IZQUIERDA RUMBO CERO UNO CERO Y..QUÉ MIL

MANTENDRÍAMOS?

RNG APP 1:58.058 MANTENGA UNO TRES MIL PIES

LA 3020 1:59.973 UNO TRES MIL PIES LAN COLOMBIA TRES CERO DOS CERO

RNG APP 2:04.282 LIMA MIKE INDIA DOS NUEVE TRES TRES PUEDE EFECTUAR

USTED EL VIRAJE POR LA DERECHA PARA INICIAR LA.. EL

DESCENSO? TIENE USTED DOS TRÁNSITOS A UNA MILLA

POR DEBAJO DE USTED

LMI-2933 2:14.815 TRÁFICO A LA VISTA NO ES FACTOR Y SOLICITAMOS

INCORPORARNOS DE UNA VEZ AL LOCALIZADOR (GEAR DOWN)

RNG APP 2:21.038 CAPITÁN USTED TIENE DOS UNO CERO NECESITO BAJARLO

DE NIVEL TENDRÍA QUE MANTENERSE A POR SU DERECHA

PARA INICIAR SU DESCENSO

LMI-2933 2:30.962 NEGATIVO SEÑORITA ESTAMOS YA INICIANDO EL DESCENSO

Y ESTAMOS PARA EL LOCALIZADOR

RNG APP 2:37.665 AVIANCA NUEVE SIETE SIETE UNO VIRE DE INMEDIATO POR

SU IZQUIERDA CON RUMBO DOS SIETE CERO

AV-9771 2:43.410 VIRANDO POR LA IZQUIERDA CON RUMBO DOS SIETE CERO

AVIANCA NUEVE SIETE SIETE UNO

¿ME CONFIRMA LA OTRA AERONAVE YA FUE DECLARADA EN

EMERGENCIA?

RNG APP 2:52.985 LAN COLOMBIA TRES CERO DOS CERO MANTENGA AHORA

RUMBO CERO NUEVE CERO

LA 3020 2:56.634 CERO NUEVE CERO LAN COLOMBIA TRES CERO DOS CERO Y

.......EN FRECUENCIA

RNG APP 3:00.199 AVIANCA NUEVE TRES CINCO SEIS DE INMEDIATO POR LA

IZQUIERDA CON RUMBO DOS NUEVE CERO

AV-9356 3:06.250 DOS NUEVE CERO POR LA IZQUIERDA AVIANCA NUEVE TRES

CINCO SEIS

RNG APP 3:10.107 LIMA MIKE INDIA DOS NUEVE TRES TRES TIENE UN TRÁNSITO

ADELANTE SUYO DIEZ Y OCHO MIL PIES ALFA 320 (AIRBUS 320)

LMI-2933 3:20.245 ESTÁ IDENTIFICADO EN EL TCAS Y LO TENEMOS ARRIBA DE

NOSOTROS SEÑORITA Y ESTAMOS EN CURSO FINAL

RNG APP 3:25.549 LA AERONAVE ESTÁ CON DIEZ Y OCHO MIL PIES CAPITÁN ESTÁ

EL TRÁNSITO AHORA ABANDONANDO POR LA IZQUIERDA .. ADICIONAL

TIENE UN TRÁNSITO ..O KEY YA DEJÓ DIEZ Y OCHO QUINIENTOS

LMI-2933 3:36.728 A LA VISTA Y ESTAMOS CON UNO OCHO MIL NOSOTROS

LA 3019 3:42.620 MEDELLÍN LAN COLOMBIA TRES CERO UNO OCHO LISTO A

CAMBIAR

RNG APP 3:46.034 LAN COLOMBIA TRES CERO UNO NUEVE VEINTIUNO TRES

BOGOTA NORTE

LA 3020 3:50.523 LAN COLOMBIA TRES CERO DOS CERO ¿POR CUÁNTO TIEMPO

MÁS MANTENEMOS EL RUMBO?

RNG APP 3:54.091 ATENTO, LIMA MIKE INDIA DOS NUEVE TRES TRES DIEZ Y SIETE

SETECIENTOS CONTINÚE LA APROXIMACIÓN PISTA HÚMEDA

LLAME VOR UNO CERO MIL Y EN LO PRACTICABLE SI REQUIERE

ALGÚN SERVICIO EN TIERRA

LMI-2933 4:05.051 LE CONFIRMAREMOS PARA EL SERVICIO EN TIERRA Y ESTAMOS

EHH..A TRAVES DE UNO SEIS MIL PARA EL LOCALIZADOR

RNG APP 4:13.156 DE ACUERDO QNH TREINTA VEINTISIETE

LMI-2933 4:15.897 TREINTA VEINTISIETE

LA 3020 4:19.813 NECESITAMOS EHH.. RUMBO POR LA DERECHA LAN COLOMBIA

TRES CERO DOS CERO POR MAL TIEMPO

RNG APP 4:23.456 AUTORIZADO LAN COLOMBIA TRES CERO DOS CERO POR LA

DERECHA ¿CON RUMBO DOS CERO CERO LE ES PRACTICABLE?

LA 3020 4:30.396 DOS CERO CERO EH.. CON INICIAREMOS EL VIRAJE Y YA LE

INFORMAMOS

RNG APP 4:34.493 O KEY LAN COLOMBIA TRES CERO DOS CERO RECIBIDO

AV-9771 4:41.048 AVIANCA NUEVE SIETE SIETE UNO SIGUE CON RUMBO DOS

SIETE CERO

RNG APP 4:44.548 AVIANCA NUEVE SIETE SIETE UNO PROCEDA NUEVAMENTE AL

VOR DE RIO NEGRO MANTENGA UNO OCHO MIL PIES

AV-9771 4:50.823 CONFIRMAR POR LA IZQUIERDA O POR LA DERECHA AL VOR DE

RIO NEGRO AVIANCA NUEVE SIETE SIETE UNO

RNG APP 4:54.206 A DISCRESIÓN EL VIRAJE AVIANCA NUEVE SIETE SIETE UNO

AV-9771 4:57.233 VIRAJE A IZQUIERDA AL VOR DE RIO NEGRO AVIANCA NUEVE

SIETE SIETE UNO

RNG APP 5:00.826 AVIANCA NUEVE TRES CINCO SEIS A DISCRECIÓN SU VIRAJE

PROCEDA NUEVAMENTE EN ESPERA AL VOR DE RIO NEGRO

Y MANTENGA UNO NUEVE CERO

AV-9356 5:08.370 A LA IZQUIERDA EL VIRAJE A LA ESPERA EL VOR DE RIO NEGRO

UNO NUEVE CERO AVIANCA NUEVE TRES CINCO SEIS

RNG APP 5:16.933 LAN COLOMBIA TRES CERO DOS CERO ¿ME CONFIRMA SI SE

ENCUENTRA LIBRE DE MAL TIEMPO?

LA 3020 5:22.945 CON RUMBO DOS CERO CERO ESTARÍAMOS LIBRES DE MAL

TIEMPO Y NIVEL DE REGRESO AL PLAN DE VUELO PODRÍAMOS

VERLOS

RNG APP 5:28.715 PROCEDA CON RUMBO DOS TRES CERO AHORA LAN COLOMBIA

TRES CERO DOS CERO LA AERONAVE SE ENCUENTRA EN FINAL

MANTENGA UNO TRES MIL PIES

LC 3020 5:37.638 DOS TRES CERO EH.. LAN COLOMBIA TRES CERO DOS CERO Y

ENTERADO DE LA AERONAVE

AV-9356 5:48.474 RIO NEGRO AVIANCA NUEVE TRES CINCO SEIS VIRANDO POR

LA IZQUIERDA DIRECTO A RIO NEGRO Y SE INCORPORARÁ EN

RIO NEGRO A UNO NUEVE CERO

LMI-2933 5:57.932 SEÑORITA LIMA MIKE INDIA DOS NUEVE TRES TRES ESTÁ EN

FALLA EH.. TOTAL ELÉCTRICA TOTAL SIN COMBUSTIBLE

RNG APP 6:12.117 PISTA LIBRE Y OPERANDO LLUVIA SOBRE LA ESTACIÓN LIMA

MIKE INDIA DOS NUEVE TRES TRES BOMBEROS ALERTADOS

LMI-2933 6:23.188 ENTERADO LIMA MIKE INDIA VECTORES VECTORES SEÑORITA

VECTORES A LA PISTA

RNG APP 6:29.389 LA SEÑAL RADAR SE PERDIÓ NO LO TENGO NOTIFIQUE RUMBO

AHORA

LMI-2933 6:39.171 ESTAMOS CON XX CON RUMBO TRES SEIS CERO TRES SEIS CERO

RNG APP 6:43.958 CON RUMBO, VIRE POR LA IZQUIERDA CERO UNO CERO

PROCEDERÍA AL LOCALIZADOR DEL VOR DE RIO NEGRO UNA

MILLA ADELANTE DEL VOR AL MOMENTO USTED SE ENCUENTRA

CORRECTO LE .. LE CONFIRMO POR LA IZQUIERDA CON RUMBO

TRES CINCO CERO

LMI-2933 7:01.596 IZQUIERDA TRES CINCO CERO SEÑORITA

RNG APP 7:04.014 SI CORRECTO USTED ESTÁ A CERO COMA UNA MILLA DEL VOR

DE RIO NEGRO

RNG APP 7:12.157 "LUGAN" (VOZ VARONIL) NO LO TENGO CON LA ALTITUD LIMA

MIKE INDIA

LMI-2933 7:21.428 NUEVE MIL PIES SEÑORITA

LMI-2933 7:25.064 VECTORES! VECTORES!

RNG APP 7:30.514 ESTÁ A OCHO PUNTO DOS MILLAS DE LA PISTA

RNG APP 7:36.246 ...TITUD (POSIBLE PALABRA ALTITUD CORTADA)

RNG APP 7:37.902 QUE ALTITUD TIENE AHORA

RNG APP 7:42.106 YA NO CONTESTA ( VOZ VARONIL )

RNG APP 7:43.645 NO

25 comments:

  1. Replies
    1. Thanks Stefano! I am still working on it as new aspects come to light, or as better ways to present the info occur to me.

      Delete
  2. They should have declared fuel emergency in the first place. Great analysis!

    ReplyDelete
    Replies
    1. Neither can I reconcile the fact that the FEED LO LEVEL warning for engines 1, 2, 3, and 4 would have been raised 150+ km out. A straight-in approach would have been assured if they had declared, and there was plenty of fuel to make it.

      Delete
    2. As a formally trained computer scientist with reliability engineering and failure mode analysis, and some limited exposure to general aviation ops, plus having been in Colombia for month after I graduated, I found myself engrossed in trying to understand what had happened here and why. Yours is one of the most impressive pieces of engineering data analysis I have ever seen. I had already visited close to twenty sites, a mix of news agencies and aviation specialist sites, and had begun to suspect something similar to what you have likely confirmed. But the one thing that you didn't reference in the above comment was the external (non-aircraft performance data) situation which probably caused them to hold back from declaring. I have heard that they would have faced impoundment of the plane and a $25,000 USD fine if their fuel condition had been discovered, as it would have been in a post-declaration review. And that coupled with the fact that they had apparently gotten away with some similar flights in the past. But of course, they hadn't anticipated the Avianca flight's requested priority landing, for one thing. Another factor may have been that they had quoted the Chapeco FC a low price, and only afterwards discovered that they would not be able to fly direct to SKRG from Brazil. This, coupled with the airline's apparent chronic financial difficulties may have left them in a situation where they might have been required to pay for fuel and airport services at the time of their arrival, and they may not have had sufficient money until they got to Medellin, where they might have made prior arrangements to be paid. This is only conjecture on my part, but given that I have also been involved in technical auditing, their financial position coupled with the additional problems caused by the need for a their flight to originate in Bolivia rather than Brazil certainly, (in my not even remotely humble opinion :-] )merits investigation as a possible, perhaps even probable, further aggravating factor. Thank you once again for this excellent report.

      It has been said that curiosity killed the cat but that satisfaction brought it back to life. Certainly your report has gone a long way towards satisfying my curiosity, pending the arrival of a final definitive investigative report.

      Delete
  3. Impressive analysis with somewhat limited data available

    ReplyDelete
    Replies
    1. Hi Seinar,

      The phrase, "connect the dots" comes to mind. I am also impressed with how much insight can be revealed by ground speed, altitude, track heading, position.

      If only we had calibrated airspeed, thrust, flap/gear.

      Thanks! Peter

      Delete
  4. Very comprehensive analysis.

    You might want to add a reference to the ADS-B coordinates sent by the aircraft being offset by approximately 1 nm on departure from Viru Viru and likely by a similar amount towards the end of the flight.

    That would have the effect of shifting the "Distance to Runway" scale on your graphs, though probably by an amount that's only discernable on the last 4 charts. It doesn't materially affect the conclusions, though it strengthens the case that the aircraft had more than enough height to have enabled it to reach the destination airport in a glide.

    ReplyDelete
    Replies
    1. Hi Dave,
      Thanks for your observation. I have added another figure to showcase the offset during the departure as you suggest. I cannot be certain the extent of the offset at SKBR - although I will dig into the holding pattern instructions. As you surmise, an error of 2 km does not change any conclusion...Peter

      Delete
    2. Noted.

      The offset, by the way, is down to the fact that most, if not all, BAe RJs don't use GNSS as the source for their ADS-B horizontal position data. Instead it's derived from inertial sources, which are subject to gradual drift.

      The accident aircraft previously flew for several years in Europe where this phenomenon was frequently observed.

      The statement that "this may be the result of a survey error in Google Earth" is a bit unfair on GE !

      Delete
    3. Dave, I took your advice and had a closer look at three prior flights of this tail (the beauty of ADS-B is history!) and was able to confirm your point, that it appears to have drift that is typical of inertial sensors. I am disappointed as this is an abomination of ADS-B position confidence features. Also, why was there a 2,000 meter offset on departure - it should have been aligned. I estimate the error at SKRG arrival at about 2 nm. Note-I discovered on Nov 4 this plane flew SKRG-SLVR, 4:33 minutes plus time to takeoff and climb 3,000 feet. Did this plane have the aux tanks? Thanks for your help.

      Delete
    4. > I am disappointed as this is an abomination of ADS-B position confidence features.

      Like all aircraft with non-GNSS ADS-B, the RJ transmits coordinates with an accompanying parameter (NUCp) that effectively tells users of the data that it's useless for ATS surveillance purposes.

      > Also, why was there a 2,000 meter offset on departure - it should have been aligned.

      I guess that's just one of the many things about this flight that should have happened, but didn't. No real surprise there.

      > I estimate the error at SKRG arrival at about 2 nm

      It occurs to me that, based on the published data, the aircraft's heading during the 4 minutes or so before it entered the hold was pretty consistent at about 357°, suggesting that it was tracking a VOR radial (presumably the RNG). If that's the case, it would mean that the east-west component of the ADS-B error at that point was about 1 nm E (to which, of course, the unknown N-S error component needs to be added). I agree that the aircraft couldn't have been far enough north to be holding directly over the VOR as that would imply either a vertical descent to the crash site (not supported by the pattern of wreckage) or a third orbit of the hold, unseen by ADS-B (not possible in the time between the last received transmission and the ground impact).

      > Did this plane have the aux tanks?

      Only a few RJs have the wing stub auxiliary tanks (also referred to as "pannier" tanks) and I don't believe the accident aircraft was one of those.

      Delete
    5. Hi Dave, All great points, thanks for sharing!

      Delete
  5. Any chance you could overlay the holding pattern flown by AV-9771 over the LAMIA holding pattern? The LAMIA holding pattern appears to be someway southand/or shortened compared to the pattern indicated in the approach charts. Thanks.

    ReplyDelete
  6. I have provided some additional discussion on technical performance in light of the apparent 10 nm offset of LMI2933 holding pattern. AVI9771 appears to be holding as expected. If LMI2933 was in fact 10 nm north of its reported position, then its glide ratio needs would have been considerably softened, and yet the the airplane would have made little progress its final 8,000 feet in descent, which is not consistent with its ATC reports. I cannot resolve the quandary, but either it was near where it said it was, and if so the analysis is accurate; or if it were 10 nm north, the failure to glide to the airplane seems even more tragic.

    ReplyDelete
  7. Very good work!!! Impressive analysis with just few data and information, congratulations!!!
    Just a point: if the aircraft position follows your estimative of maximum 2 nm deviance, the decision to stay in the holding pattern in that position definitely contributed with the accident (crash)? In this case, is it only pilot's fault or involves ATC decisions too?

    ReplyDelete
    Replies
    1. Thanks! The ADS-B is a reporting service. The flight crew fly to other references including radio updates. The captain is responsible for the safe operation of the plane ultimately, and he had options to land safely that he failed to embrace. Why he chose to do that remains the subject of the investigation...here the focus is what happened. Questions to answer include the flight plan, the failure to fly straight in, the failure to fly in after one circuit, the decision to lower flaps and gear.

      Delete
    2. Couldn't find a reply button below where the subject of what was said in Spanish came up. I am close to, if not at, advanced proficiency in Spanish as a 2nd language. After having read this, I now want to go back and more carefully analyze the tower and pilot conversation in light of what I have learned here. However, it is too late, I am too busy and too tired, and it will have to wait for now. But I will try to find the better (more complete) audios I heard and if I can, I will come back to post anything that might be relevant to the types of situations alluded to here.

      Delete
  8. Peter, do you think that maybe the pilot thought he was much closer to the airport? That the VOR and the runway were next to each other? That would maybe contribute to explain some things, like no panic, landing gear down. He descended and descended, and only panicked and asked for vectors when the runway lights didn't turn up in his visual where he expected them to be?

    ReplyDelete
    Replies
    1. Hi Pablo,

      There is no question of the orientation and separation of the VOR/DME and the SKRG runway 01. Two crew members on an IFR flight plan would be expected to become familiar with the arrival procedures and contingencies.

      I am unable to understand the reason for the apparent hold displaced to the south. I get your point that if LMI2933 thought it was holding correctly then it would have been more confident in its descent.

      The issue is the airplane has dual VOR/DME and some form of moving map. A radio nav update (DME/DME for example) would normally reset the IRS offsets - a radio nav failure can propagate into a blended solution with the IRS as mistaken for drift. This can also happen with nav data base errors relating to the stored location of the navaids.

      Following the thinking that the crew thought they were holding over the VOR, the crew would have still faced 10 nm to the runway threshold at the VOR, and they descended from 21,000 to 15,000 - with about 8,000 left before they thought they had gotten there. With the memory item saying it would take 10,000 feet to make it to runway 01 at RNG VOR crossing, it still seems a bold move if you are worried about flameout. My guess is they descended under power not knowing that the flameout was going to catch them, rather as we sit here in hindsight knowing the inevitability.

      Runway 01 is 1.9 nm long. You could literally fly at 1,000 feet above the threshold and still make the runway at midfield. I wonder if there was concern over a flaps up landing, that they wanted to extend flaps before they lost hydraulics.

      When EMERG DC/AC bus became the last resort, there was still IRS, artificial horizon, altimeter, speed indicator, ILS, and VOR with direction bearing indication. There is no EFIS. I cannot profess to know whether the LOC would be received once the airplane descended below about 12,000. This would leave him with dead-reckoning (vectors) - but the VOR should still have been available.


      Delete
    2. I do speak spanish. In the ATC recording you hear them saying that they are lined up with the localizer. So, I would understand this as "established". I have no LPC in spanish. I don't know if he is using the right term for being "established". At 10'000 they would have localizer signal, but they were below this, behind the montain at 9000 feet, it is was foggy at that altitude. So, they had no more visual. Initially they said, "we have visual with the ground" That was at FL210. I believe that when they passed the 10000 and went behind the mountain, the they lost signal with the localizer and this is why they were desperately asking for the vectors. ATC did no longer have their position either. Honestly, even proper guidance by TWR was sort of impossible, not having the on the screen. I don't think they use PSR in SKRG. I think it was SSR and as such any vector could lead to sudden death. The last you hear "jesus" was the second when they saw the hilltop in front of them. I agree with you Peter, I believe too that they did think to land and lowered the gear. Then the surprise with the flameout and no way to retract the gear again.

      Delete
  9. Very comprehensive !

    28 Nov : 2:55 - 22:18 = 4:37 and 4 Nov : 1:16 - 20:43 = 4:33, right ?
    I wonder if it's possible to know approximatively the payload on 4 Nov...

    ReplyDelete
  10. I belive Lamia crew doesnt have any situational awareness and become panicked from a flameout in one of engines. They werent prepared to manage a fuel exhaustion and clearly bid to a normal landing. Until last seconds they were asking for vectors, they really dont understand how dire was the situation. I think they never know about technical information of the aircraft or total flameout emergency procedures.

    ReplyDelete
  11. Keep up the good work! Brought here from another site, but no more on that...
    I thought a (large) flow-chart of all the possible cases re the flame-out and all the related systems complexities with all the different emergency avenues could possibly help people to grasp the situation.

    ReplyDelete
  12. Well done! There is a lot of great information here. Well sourced too. It isn't that readable, however. If anyone wants opinions on 11 different things that helped cause this crash (besides just the ridiculous flight plan), see here: https://www.facebook.com/notes/jason-arthur-taylor/media-reverses-facts-to-cover-up-causes-of-lamia-airlines-flight-2933-crash/10207725094468895/

    ReplyDelete