by Palmer, Bill
On February 2, 2013, an Etihad Airways A340-600 experienced an unreliable airspeed incident en route from Abu Dhabi, UAE to Melbourne, Australia. The incident’s preliminary report13 states:
While cruising at FL350, just leaving Colombo FIR and entering Melbourne FIR, the Aircraft encountered moderate to heavy turbulence and experienced significant airspeed oscillations on the captain’s and standby indicators. The autopilot, autothrust and flight directors were disconnected automatically. The aircraft’s flight control law changed from Normal to Alternate Law, which caused the loss of some of the flight mode and flight envelope protections. The change from Normal to Alternate Law occurred twice, thereafter the Alternate Law stayed until the end of the flight. Auto-thrust and flight directors were successfully re-engaged, however, both autopilots (autopilot 1 and 2) could not be re-engaged thus the Aircraft was controlled manually until its landing.
Weather Radar
All of these affected flights had operating weather radar, but interpretation of weather radar requires skill, experience, and understanding of the principles involved. It is not unlike a doctor reading an x-ray. We can all find the bone on an x-ray, but it takes training and experience to discern meaning from the subtle shadows and gradients. Airborne weather radar is much the same. The radar control allows the pilot to adjust the tilt of the radar beam, the sensitivity of the receiver (gain), and the range and brightness of the display. The colors and shapes have to be interpreted. Unfortunately the pilot does not have the option to call in a radiologist.
Pilots rely on weather radar to navigate around storms. The objective is to avoid dangerous turbulence. Radar does not directly indicate turbulence. But operating and interpreting it correctly can indicate conditions where varying degrees of turbulence are likely to be.
Radar sends out pulses of radio waves and then listens for those waves to bound back off of water droplets about the size of a raindrop.
Heavy concentrations of water droplets are often associated with the strongest part of storms, and the correlation with turbulence is quite good. The additional use of Doppler signal processing on some radar sets allows the radar unit to measure the movement of different areas of water droplets within a storm for an additional indication of turbulence, but this is generally only available at short range (40 miles).
Radar reflects poorly when liquid water is not present. It does not reflect off water vapor, the micro-sized droplets that form most clouds, and reflects poorly off ice crystals ranging from snow flakes to hail stones. Unfortunately, the upper portions of thunderstorms have fewer water droplets and more ice crystals than the lower portions and therefore do not show up well on radar.
Radar returns on the A330 are displayed in three colors on the Navigation Display (ND), overlaid with navigation and TCAS (traffic) information.
The operation of airborne weather radar is something that is unfortunately not taught very well or formally in a classroom environment. The basic principles of operation are well known by most, if not all, commercial pilots. Pilots are taught some basic tips for interpretation, but for the most part it is on-the-job training in radar operation. Few flight simulators allow for depiction of weather on the simulator’s displays. Those that do are fairly unsophisticated, lacking the realism necessary for in depth training on interpreting the subtleties of the display. It is safe to say that valuable simulator time is not spent learning how to interpret weather radar displays, even if it had the fidelity to do so.
There are many aspects that a pilot must take into consideration when interpreting a radar display. Among them are the intensity of the return (shown in colors), the gradient and shapes of the different intensities, blocking of returns by heavy weather or the curvature of the earth, and differentiating between radar returns reflecting off the ground and actual weather.
Not all areas displayed on radar need to be avoided. It is often impossible to avoid them all. Patterns of the radar display intensity and the gradient of radar returns as well as the shape of those returns provide clues as to the nature of the weather. Additionally, the altitude of the airplane and the curvature of the earth make it impossible to see storms in close proximity yet below the airplane, or those over the horizon. When determining the height of a displayed area of weather a pilot must also consider the width of the radar beam, and that ice crystals and hail (often found in the upper portions of a storm) reflect radar signals poorly.
At cruise flight levels, when storms are suspected pilots must adjust the tilt of the radar antenna beam down to look for the more reflective parts of the storm that may cause turbulence at their altitude.
Antenna gain must also be manipulated to try to enhance weak signals. The normal position of the gain control is in the “calibrated” position (center). The gain selection can increase the receiver’s sensitivity to try to display weather that is too weak for the default setting, such as those within the ITCZ. The gain can also be decreased in order to concentrate on only the highest intensity returns.
Shortly after First Officer Robert’s return to the cockpit, he selected MAX on the gain setting which then prompted him to suggest a deviation around at least part of the weather ahead.
Below is a typical A330 radar display on the Navigation Display (ND). Green is the least intense returns, red is the most intense returns. The airplane position, course line, and other navigation data are superimposed on the radar data. This makes navigating around weather much easier and more precise than older setups where the radar is a separate display from the navigation instruments. In this image, a 10 mile offset to the right of the active course is being evaluated. Parameters in the lower right of the display indicate the gain is being manually controlled (MAN GAIN) and the antenna tilt is 2.5° down.
In the daytime it is often easy to see and avoid storms hundreds of miles away by vision alone. At night, moonlit clouds or patterns against a starry background may offer some visual clues. Lighting illuminating clouds from within provides a solid visual cue, but is less frequent in the ITCZ storms.
When Air France 447 was approaching the weather, the half moon was setting in the west leaving a dark sky out the window. Surrounding cloud layers prevented them from seeing much.
AF447 was fitted with a Rockwell-Collins WXR 700 weather radar. Adjustments to the tilt and the gain are made manually. Each pilot has the ability to select a radar display range from 10 to 320 nautical miles on their respective navigation displays.
A330’s are equipped with two radar systems for redundancy, but only one is active at a time. Both systems use the same antenna and control head.
The WXR-700 was an excellent unit for its day and captured more than 50% of the major airline market share, but it requires manual operation.
The most recent radar units, such as the Rockwell-Collins WXR-2100 Multi-scan Radar, provide higher levels of signal processing and automatic operation. During the WXR-2100’s development (about 2001 time frame) Rockwell-Collins Engineers discovered that oceanic weather is significantly less reflective than storms of similar height over land masses. A lot of time, money, and effort was put into designing algorithms to compensate for this lower reflectivity.14 The newer units will increase the gain (sensitivity) at altitudes above the freezing level, as well as in low latitude oceanic areas to show weather that may be hidden when set to the standard setting of calibrate. Other filters and internal functions limit non-critical returns, such as low altitude weather and ground returns, so that the radar only shows weather that is a threat.
Ten minutes before AF447’s autopilot disconnected, Captain Dubois left for a routine rest break leaving the two first officers in charge of the airplane. They discussed the intertropical convergence zone, the outside temperature and the resulting limit on the maximum altitude. They called the cabin crew and told them of an area they were approaching where it will ‘start moving about a bit,’ to ‘watch out,’ and that it would be a ‘good idea’ to sit down.
At 02:08 Robert suggested, “
Don’t you maybe want to go to the left a bit?”
Bonin replied “Excuse me?”
Then Robert reiterated, “You can possibly go a bit to the left, I agree that we’re not in manual, eh?” He then appeared to point out something. “Well, you see at twenty with the..” He then selected “MAX” on the radar’s gain control to better show portions of the storm not displayed. It seems that the image he obtained appeared sufficiently different as to require a change of strategy. Heading mode was selected and the airplane began a slight left turn. Satellite imagery indicates that they were navigating between two heavy areas within the storm.
When possible, weather deviations should be requested ahead of time, and a clearance issued. While traffic separation may not have been considered critical at their location, because the adjacent tracks are 100 miles to the right and 120 to the left, ATC notification was still required. Pilots are permitted to make weather deviations as required. If a clearance has not been obtained with a deviation in excess of 10 miles, a 300 foot altitude adjustment is required to help avoid traffic on the adjacent track doing the same thing. The transcript reveals no attempt or discussion of informing ATC of their deviation and acquiring the clearance.
The crew’s failure to operate the radar in such a way as to avoid the massive area of weather ahead of time, put them in a position of just avoiding the worst parts as they entered the line of weather. But even while avoiding the heaviest radar returns the airplane soon encountered conditions that overwhelmed the pitot tubes’ ability to measure airspeed.
Pitot Tubes and their Replacement History
The pitot tube was invented by the French engineer Henri Pitot in the early 18th century and was modified to its modern form in the mid-19th century. It is widely used to determine the airspeed of an aircraft and to measure air and gas velocities in industrial applications. A simplified version appears below.
On the A330, three pitot tubes are mounted on the left side, lower half of the airplane, just forward of the cockpit. The pitot tubes are automatically heated whenever an engine is running or the airplane is in flight. This fact refutes any theories that the crew forgot to turn the pitot heat on. There was no way for the pilots to turn it off using normal cockpit controls.
The pitot tubes in use were designed to handle known icing and water ingestion issues with a comfortable safety margin. However, the standards at the time did not consider the type of icing encountered by Air France 447. Those conditions were simply not known.
It is hard to blame Thales (pronounced: tal-ess), the pitot tube manufacturer, for a design that lacked the ability to handle an unknown threat.
Nevertheless, the refinement of pitot tube designs has been an ongoing effort.
In 1995 Airbus developed a set of specifications designed to improve the performance of pitot tubes in a wide range of conditions including ice crystals. While rigorous tests were performed, the ice crystal diameter was set at a hypothetical 1 mm. However, the true size and density of ice crystals, as well as the dividing line between supercooled water and ice crystals was not known with sufficient precision, and may still not be.
In the chart below15 the blue and red regions represent the certification requirements, and the purple and amber represent additional Airbus requirements. Below -40° all water is considered to be frozen, and therefore not an icing hazard for supercooled water.
In 2001 there were issues with ingestion of ice crystals and/or water with a specific model of Goodrich manufactured probes (851GR). The solution then was the mandatory replacement of the Goodrich GR probes either with Goodrich type 0851HL (HL probes) or by Thales model C16195AA (AA probes) before December 31st, 2003. Air France received its first A330s in December 2001. They came equipped with the Thales AA probes.
In September 2007, following measured speed inconsistencies observed at the time of heavy precipitation or icing conditions on A320s, and in some cases on A330/340 aircraft, Airbus published a service bulletin which recommended the replacement of the AA probes with the BA model (C16195BA). The Service Bulletin initially indicated that this model performed better in the case of water ingestion in heavy rainfall, and icing in severe conditions. Due to the absence of problems of this type affecting its long-haul fleet, Air France decided to replace the Pitot AA probes with the BA probes, but only in the event of a failure.
Between May 2008 and March 2009, nine incidents of unreliable airspeed indications were reported through pilot Air Safety Reports (ASR) for Air France’s A330/A340 fleet. All occurred in cruise between FL310 and FL380. In seven cases, the ASRs mentioned the activation of the stall warning. Air France studied the events, mostly on the maintenance side. Starting in July 2008, Air France reported these events to Airbus.
The Air France flight safety officer (OSV) for A330/A340 interviewed most of the pilots who reported these incidents. The accounts given by these pilots did not suggest an immediate risk. The heads of the flight safety division, technical information office, and professional Standards for the A330/340 division also interviewed some of these crew members.
Analysis of the incidents revealed that the pitch attitude during these incidents varied from -3° to +7°, and that the maximum angle of attack was 13 degrees. (The normal value for both is around 2.5°.) Stall warnings were momentary and no loss of control had occurred.
After the accident, six additional events, not filed as air safety reports, were found by analyzing recorded flight parameters from the fleet and maintenance reports.
Meanwhile in 2007, prompted by the earlier incidents caused by water ingestion on A320 aircraft, an exercise incorporating the “flight with unreliable airspeed” procedure conducted in an after-takeoff scenario was added to Air France’s 2008-2009 training program. The exercise was considered to be representative of the main difficulties in conducting the procedure in all flight phases. However, when conducted in a low-altitude environment, the procedure calls for pitch attitudes between five and fifteen degrees. When conducted above the minimum safe altitude, the procedure calls for maintaining level flight for troubleshooting. There were no simulator exercises added to reflect the high altitude environment events that were encountered in 2009.
In September and October 2008, Air France asked Airbus for information about the cause of these events and the solutions. They also asked if the Thales BA probe could resolve these problems. Airbus replied that the cause of the problem was probably probe obstruction by a rapid accumulation of ice crystals, and that the Thales BA probe was unlikely to improve the performance in an ice crystal environment.
From October 2008 onward, Air France alerted Thales about the increasing problem of icing at high altitude. Thales started an internal procedure to perform a technical analysis of these incidents.
During the autumn of 2008, Air France considered that flight safety was not immediately affected by this type of incident.
Four Air Safety Reports (ASRs) relating to these incidents were published during this period in several issues of the Sûrvol flight safety bulletin, which was circulated to Air France pilots. On November 6, 2008, information about the airspeed anomalies that had occurred in cruise and that affected the A330/A340 fleet was circulated as an operational memo titled Info OSV within Air France to the pilots working in the sector. The document indicated that six events of this type were reported by crews. First Officer Bonin was about to check out on the A330 when this bulletin came out. He received his type rating in December 2008.
The Info OSV document stated that the incidents are characterized by losses of airspeed indication, numerous ECAM messages, and in some cases, configuration alarms. The events occurred at high altitude in turbulence, in zones in which icing was forecast or observed, for aircraft flying at a Mach of 0.80 to 0.82, with autopilot and autothrust engaged. The chronology of the anomalies was described. It stated that, “during this phase, which lasted for approximately a few minutes, the crews did not report any feeling of over-speed (vibration, acceleration) or the approach to st
all (pitch attitude, angle of attack, reference to the horizon) despite the activation of the stall warning.”
It stated in bold red letters “be vigilant in flight conditions of high altitude, icing, and turbulence.”
Four general recommendations were included in the document. (Approximate translations)
Read the complementary technical information carefully.
Do not be taken by surprise.
Identify and confirm the situation.
Recovery in case of manual control of the aircraft. Proceed by making small corrections.
The presence of this information and the dissemination of these bulletins indicates that Air France pilots, especially A330 & A340 pilots, should have been aware of these incidents. The conditions that AF447 encountered were exactly the conditions that the bulletins and Air Safety Reports referred to.
Unfortunately, the “Flight with Unreliable Airspeed” procedure and the conditions for its application were not mentioned in the Info OSV document. A safe attitude and power setting to go to in the event of the loss of automation and/or airspeed indications was also not mentioned. Yet, that key piece of information is what pilots who have successfully flown through the loss of airspeed events report having used in the critical first seconds.
Training by bulletin rarely has lasting effectiveness, and perhaps even less so when the guidance is general in nature and includes only vague corrective actions.
On November 12, 2008 Airbus revised the earlier 2007 service bulletin. Like the earlier version, this version mentioned the improvement that could be provided by the Thales BA probe in relation to water ingestion, but no longer mentioned icing conditions.