Overweight and tail heavy airplane was ‘legal’ for takeoff
The fatal Jan. 8 crash of Air Midwest [AMI] Flight 5481 on takeoff at Charlotte, N.C., illustrates the danger of operating an aircraft outside of its approved performance envelope and how maintenance error can combine with an overloaded condition with deadly consequences.
The airplane was a ticking death machine when it left maintenance on the morning of Jan. 7, and in its overloaded and tail heavy condition the latent hazard lurking in the pitch control system meant the pilots a day later were unable to prevent the airplane from lurching into a stall.
The case already has spawned a major review by the Federal Aviation Administration (FAA) of the average passenger and baggage weights used to calculate takeoff weight. The fact that the airplane crashed two days after it underwent maintenance has caused the National Transportation Safety Board (NTSB) to focus its investigation on the maintenance practices that left the pilots on the accident flight with too little pitch authority to save the situation.
The morning flight of the Beech 1900D twin-turboprop airliner was carrying its maximum capacity of 19 passengers, and the pilots – Capt. Katie Leslie and First Officer Jonathan Gibbs – clearly were concerned about being over the airplane’s maximum takeoff weight and outside of the aft center of gravity (CG) limit. The transcript of the cockpit voice recording is full of pre-takeoff discussions about the airplane’s weight, and the pilots jokingly suggested moving their seats forward so as to distribute more of the load toward the nose of the airplane.
Black humor turned to horror as the airplane’s nose-up pitch increased constantly from 5� after liftoff to some 55� over a period of 15 seconds. The engines were powerful enough to carry the airplane to an altitude of some 1,200 feet, but with the steep pitch, the wings stalled and the airplane crashed near a hangar at Charlotte-Douglas International Airport.
Indeed, the nose wheel had been coming up during takeoff even as the crew pushed forward on the yoke.
This was a case where adequate information from the digital flight data recorder (DFDR) enabled investigators to quickly home in on the likely cause. The recorder captured numerous flights both before and after the aircraft’s empennage and engines underwent maintenance on the night of Jan. 6, 2003, at the Raytheon Aerospace facility at Huntington, West Virginia. This facility was performing contract maintenance for Air Midwest under its Federal Aviation Regulations (FAR) Part 121 operating certificate.
The recorder revealed that the elevator, which controls pitch, was in a markedly different position in flights after the control cables were tightened during that maintenance visit. On flights before that work, the elevator during cruise was in about a 4� nose-down position. After the hangar visit, the elevator during cruise was in about a 13� nose- down position.
According to NTSB investigator-in-charge Lorenda Ward, who was among the first to testify in two days of hearings last week, “the airplane lost about two-thirds of its down elevator capability.” That was the latent hazard. It did not rear its deadly head until two days later, when the airplane was most heavily loaded. Based on the average passenger and baggage weights used to calculate takeoff weight, First Officer Gibbs estimated the airplane weighed 17,018 pounds, or just 102 pounds under the maximum authorized takeoff weight of 17,120 pounds. As it turned out, Gibbs erred by 10 pounds, and the weight manifest should have read 17,028 – still below the maximum but closer to the limit. Nonetheless, the figures used by the crew placed the airplane within its weight and balance limits.
Investigators believe the accident airplane weighted closer to 17,400 pounds, based on the physical evidence (actual weights of persons aboard, more likely weight figures for baggage and carry-on). As such, the airplane was some 280 lbs. over its maximum takeoff weight. Because of the baggage stowed in the rear, the airplane was tail heavy, to the extent that its CG was slightly more than two inches outside of the aft limit. Other overweight Beech 1900s have been flown as much as 11 inches out of the aft CG limit, usually with unfortunate consequences. Out of balance by just two inches, the Air Midwest airplane might have been controllable had its elevator been functioning properly.
The accident was a combination of underestimated weight and improper maintenance. The nose pitched up because the airplane was tail heavy, and the pilots could not get it back down because elevator movement was restricted.
The outdated average
At the time of the accident, Air Midwest was basing its weight calculations on average passenger and baggage weights. The FAA allowed operators to use either actual weights or approved average weights; this guidance was contained in a 1996 advisory circular. Air Midwest had opted for the latter course. As such, it was among the 15 of 22 carriers operating scheduled flights in the 10-19 passenger seat category using average weights to calculate takeoff weight. The other seven carriers were using actual weights.
Of interest, the FAA requires actual weights to be used on aircraft with fewer than 10 passenger seats. The FAA’s Dennis Pratte said that actual weights are necessary for small aircraft because “they have a greater chance for error.” On larger aircraft any weight deviation can be spread over a larger area, and therefore average weights are sufficient, he explained.
His testimony highlighted a paradox. Pratte said that smaller airplanes are more sensitive to weight and balance exceedances, but the FAA allows average weights that inject a greater uncertainty into the estimate of an airplane’s takeoff weight.
As an example, in the averaging method, the passengers are presumed to be a mix of 60 percent males and 40 percent females. On Air Midwest Flight 5481, the mix was 84 percent male and 16 percent female.
An NTSB official asked, “Why wouldn’t it be prudent to use actual weights all the time on these smaller carriers?”
Pratte replied, “If the carriers’ average weight programs are adequate, there is no need to use actual weights.”
If actual weights had been used for Flight 5481, the airplane would not have been allowed to take off. In fact, Air Midwest has recently updated its average weight figures, based on a survey ordered by the FAA. The carrier’s average weight for a passenger with carry-on and one checked bag is about 40 pounds heavier than previously used for takeoff weight calculations. Had those weights been in effect Jan. 8, the airplane would not have been legal for takeoff.
Of the 15 carriers using average weights before the FAA ordered a weight sampling, one has since opted to use actual weights.
By permitting the use of average weights, and not demanding more frequent updates of the average weight estimates, the FAA may have abetted a situation where the accident airplane was at a legal weight for takeoff but nonetheless was too heavy for safe flight. In this seeming contradiction lies one of the root causes of the accident.
Maintenance morass
Two days before the crash, the airplane entered the hangar at Huntington for a scheduled Detail 6 (or D6) check, which was part of Air Midwest’s phased maintenance program. The check was begun shortly after 9:30 p.m. that evening and ended early on the morning of Jan. 7, when the airplane was returned to service.
Part of the D6 procedure required checking of the pitch control cable tension every 1,200 flight hours.
The men who did this work were not Raytheon employees. Rather, Raytheon Aerospace had retained the services of mechanics from a Florida-based company called Structural Modification and Repair Technicians (SMART). Known in the trade as a “body shop,” the company was basically an office. It did not have a facility to perform any maintenance, nor was it an approved repair station. SMART officials, who were not licensed airframe and powerplant (A&P) mechanics, interviewed prospective maintenance technicians sight unseen, on the telephone, and provided them to clients, such as Raytheon Aerospace. The successful candidates underwent a brief on-site orientation and were put to work in late December 2002.
On the night of Jan. 6, the maintenance workforce in the Huntington hangar consisted of seven people: five mechanics, an inspector and a foreman. Six of them were SMART employees; the seventh was a Raytheon Aerospace employee.
NTSB member John Goglia pointed out that the five mechanics doing the D6 check that night had “virtually zero experience on the Beech 1900.”
For the critical cable tension check, recently-hired mechanic Brian Zias was detailed to receive on-the-job-training (OJT) from George States, a more experienced mechanic who was the quality assurance (QA) inspector for the night’s work. States, a recent Raytheon Aerospace hire, was a former SMART employee.
States, visibly nervous during his testimony, said in his professional experience he had fully rigged the Beech 1900 elevator control cables one time before and had tested and adjusted the cable tension “several times.”
He basically guided and showed Zias how to do the cable tension check. A few steps were skipped during the course of this work. For example, the autopilot was not disconnected, as called for in the maintenance manual, because the airplane was not equipped with an autopilot. Other, more relevant steps in the prescribed procedure also were skipped. For example, safety clips were to be removed from the turnbuckles, devices used to adjust the tension on the two control cables, and cable tension was to be released completely before adjustment.
An elevator control system friction test also was not conducted.kdfsdkfsdfjsd
Zias found slack in the two cables. States basically performed the work, showing Zias how to use the turnbuckles to increase tension in the cables. He did it in such a way that the turnbuckles were not set to nearly the same length. Rather, there was a 1.76-inch difference in length.
There was no guidance in the aircraft maintenance manual or operators manual regarding limits for turnbuckle adjustments. Nor was there anything in an associated 1988 advisory circular (AC 43.13A), which only cautioned that a sufficient number of threads must be screwed into the turnbuckle barrel to hold the load. When the FAA issued an emergency airworthiness directive Jan. 27 to all operators to check cable adjustments, it did not mention that the turnbuckles needed to be within a certain range.
Joseph Machalek, vice president of maintenance for Air Midwest, said, “If you rig the elevator for full travel up and down it’s irrelevant how long the turnbuckles are.” His statement is correct on the understanding that full range of elevator motion is validated per the rigging procedure.
States said that after tightening the cable tension, he moved the horizontal elevator to check its motion. He did not perform this check against a graduated scale to assure full range of movement, as evidenced by the digital flight data tracings of elevator position before and after the maintenance was performed.
In effect, States performed what he believed was a subset of the full rigging procedure, skipping steps he deemed unnecessary. Even though States did the work, Zias signed the job sheet off as having completed the task per section BMM 27-30-02 of the Beech 1900 maintenance manual. This section outlines a rigging procedure, not a cable tensioning check. States applied his stamp to the paperwork as the QA inspector.
States and other witnesses, including Gary Sologub, the on-site manager at Huntington for Raytheon Aerospace, said it was common practice for inspectors to perform OJT. The practice was contrary to the company official duties, in which training was assigned to foremen and not inspectors.
States’ testimony generated considerable discussion over the semantics of tension adjustments and rigging. “Rigging involves tensioning but tensioning doesn’t necessarily involve rigging,” States said.
John Oxley, the Air Midwest site manager at Huntington, who worked during the day, said, “Adjusting the cable tension is rigging, so [skipping steps] is unacceptable.”
Russell Unangst, an official from the maintenance branch at FAA headquarters, said, “If you’re adjusting tension, you have to check the rigging to make sure that you haven’t affected anything.” The horizontal stabilizer is controlled by a system of cables, pushrods, bellcranks and the like.
In any event, the job sheet declared that the job was done per the rigging procedure. States applied his inspector’s stamp, essentially certifying his own work.
There are a number of problems with this sign off. First, trainees should not be attesting to work they have performed for the very first time under instruction. States’ aviation mechanic stamp should have been on the mechanic’s line of the job sheet. Second, the cable adjustment is a required inspection item (RII) for Air Midwest. RII items require that the work be inspected by a qualified individual holding RII inspector authority – and this individual must be other than the one who performs the work. With his inspector’s stamp affixed to the QA block, States’s self-inspection was in clear violation of federal aviation regulations.
There is an “explanation” for all of this activity beyond the bounds of policy and regulation. The Huntington center was understaffed. If States had not stamped the RII inspection (assuming no one else on that shift had RII authority), the plane would have been delayed. Oxley would have had to have been called in to relieve States of the OJT work, so that States could independently certify the work in his role as an RII-qualified inspector.
The center was understaffed with under-qualified people.
Even before the accident, Air Midwest had expressed its concern about the short staffing at Huntington. Before the accident, the FAA had given Raytheon Aerospace at other locations “Diamond Award” for training in 2002. This award, it was revealed in testimony, was based on the amount of training, not the quality of the training. Yet both RII inspections and fully trained personnel seemed sadly deficient in the detailed reconstruction of activities at the Huntington facility on the night of Jan. 6.
Not surprisingly, a number of changes have occurred since the accident. To cite a few:
- Mechanic Zias no longer works for SMART.
- QA inspector States was stripped of his stamp and now works as a mechanic.
- On-site manager Sologub is now a mechanic at Raytheon Aerospace’s facility at Panama City, Fla.
- Regional site manager Oxley now comes in for work at about 3 a.m. to more closely supervise the activity, and he remains on duty through the morning conference calls in which maintenance issues are discussed between himself, Raytheon Aerospace and Air Midwest officials.
- Two more people have been added, increasing staffing at Huntington from eight to 10.
- The maintenance manual has been revised with a new and more specific procedure for rigging and tensioning adjustment.
- Air Midwest has reiterated its policy that mechanics are not to sign off on work for which they are being trained for the first time.
- RII inspectors have been prohibited from performing OJT.
Checks and balances
The Flight 5481 crash post mortem has revealed major disconnects, deficiencies and disparities in oversight. As such, the crash is the visible tip of the proverbial iceberg, in which the systemic deficiencies are below the water line.
For example, witnesses testified that the FAA principal maintenance inspector (PMI) approves the carrier’s weight and balance program, as outlined in the airline’s operations specifications, which the PMI signs. But while the PMI may be familiar with the operating empty weight (OEW) of the airplane, which must be determined by actual weighing on a regular basis (i.e., typically every three years), he may have little knowledge of the actual weight and balance of airplanes in service, carrying fuel and payload (which do not figure into the OEW). On the other hand, the FAA’s principal operations inspector (POI), who periodically examines departure weight and balance figures for in-service aircraft, does not approve the carrier’s program. The PMI and POI exist in parallel universes, and there is no mechanism to connect the approved weight and balance program to the actual weights and balances of aircraft in service.
Jay Hiles, representing the International Association of Machinists (IAM), a party to the investigation, asked why the POI rather than the PMI isn’t tasked with ultimate responsibility for regulatory oversight of the weight and balance issue.
The FAA’s Pratte replied that the carrier has ultimate responsibly for its weight and balance program.
Hiles pressed his concern, asking, “Has the FAA increased surveillance in this area?”
Pratte replied, “I can’t answer that. I don’t have the data.”
NTSB member Goglia expressed concern not only about the qualifications of contract maintenance personnel, in this case the SMART employees on contract to Raytheon Aerospace who had to be trained on the Beech 1900, but about the QA arrangements. In this case, Air Midwest had relegated the QA function to its contract maintenance organization, Raytheon Aerospace, and QA supervision resided at its Panama City facility.
With more carriers contracting out an ever-greater proportion of their maintenance, the Air Midwest case has significance beyond the situation at one small carrier. Like the ripples spreading outward on the calm surface of a pond, this crash embodies object lessons for the entire industry. >> Bart Crotty, an aviation maintenance expert, contributed to this story. Crotty, e-mail [email protected] <<