21
$\begingroup$

I'm aware that there is a lot that goes into an aircraft accident investigation, so I understand that the investigations take a long time for a reason.

But what I don't know is what those reasons are. I'm hoping someone with some familiarity can give a high level overview as to why aircraft accident investigations take so much time?

$\endgroup$
  • 10
    $\begingroup$ have you ever fully read a report? collecting the information, validating it, performing analysis, setting up simulations to validate/disprove theories, interviewing experts, rinse and repeat until a valid explanation is found, drafting the reports and checking them for mistakes. Not exactly the kind of job you can take care of during breakfast. $\endgroup$ – Federico May 20 '16 at 17:11
  • 2
    $\begingroup$ @Federico Several, and I'm aware it takes a lot of time (I said so in the question), but I'm curious if someone who has done investigations can go over the process for us. $\endgroup$ – Jay Carr May 20 '16 at 17:12
  • $\begingroup$ Also, realize both these organizations are small in the government scheme... I believe the NTSB is less than 300 people last I read. $\endgroup$ – slookabill May 20 '16 at 18:38
  • 1
    $\begingroup$ That are small organizations are working on multiple cases with limited resources $\endgroup$ – jean May 20 '16 at 19:15
  • 1
    $\begingroup$ It's may be worth pointing out that some investigations result in follow-up activities that continue long after the report is issued. The longest one that I'm aware of is still in progress after about 25 years. The evidence they are looking for (improving the detection of defects in the material used to manufacture certain components) flags up approximately one new data point per year - and of course the capabilities of the technology used in the process doesn't stand still for a few decades, while you are collecting a reasonable sized set of data points to analyze statistically. $\endgroup$ – alephzero May 20 '16 at 22:25
35
$\begingroup$

When it comes down to it, accident investigations involve lots of coordination between many different groups, companies, and agencies, often around the world. These groups may range from helpful and responsive to completely uncooperative.


Visiting the Site - First there will be a team that heads to the incident site to collect information. If the site is easily accessible, this may be a fairly quick process. If the site is tough to access, or takes a while to find, this will of course take much longer. This is generally the part most visible to the public, but is only the tip of the iceberg in generating a useful report.

Research - There is also a lot of research that must go on behind the scenes. Especially if a possible cause is not clear, investigators must look into many factors to rule out the options. This can be a lot of information and may involved many different groups around the world:

  • Previous maintenance done on the plane
  • ATC and weather information about the flight
  • Information about the crew, the passengers, and the cargo on the flight
  • Many individuals may also be interviewed

If any of this initial information provides clues as to possible causes of an accident, they will also be gathering information about related aircraft design and functions. This can be a lot of information and may involve working with many different groups. The process of narrowing the focus of the investigation may involve tracking down different leads which end up being irrelevant. It may take a lot of effort just for the report to end up saying "X was not a factor."

Flight Recorders - Gathering the flight recorder data is not always as simple as plugging it in and reading the data. If there is damage, extracting and repairing the memory may take time. The FDR data must be analyzed and it may take additional work to investigate and verify the parameters. The CVR must be read out and then transcribed. Transcription is also a long process, involving different parties to the investigation, and must be synchronized to ATC and FDR information.

Analysis - Once the information has been collected, it must all be analyzed. Investigators may need to contact and work with many different groups and experts to develop an understanding of parts and processes that took groups of engineers years to analyze and design. Components may need to be analyzed down to a microscopic level. This is generally the longest part of the process. Investigators may realize that they must go gather more information. This may involve running tests on equipment to recreate issues. It may take a lot of work to not only explain what failed, but also the reasons behind the failure and how it could be avoided in the future. Most reports will find issues in many different areas, including human factors, regulations, procedures, and aircraft design.

The Report - The investigators will eventually need to come to some conclusions among themselves. This must all be written up in a report. Some reports can be quite lengthy. If you've ever had to write something like this, you might know that writing the report can easily take much longer than the work that went into it. There will also be an opportunity for parties to the investigation to make comments and provide feedback before the report is released. This is where the investigators get to share the results of all their hard work, and it will be published to get picked apart by the whole world, so it's worth taking the time to get it right.

Law - While most agencies like the NTSB focus on producing a factual report and not assigning blame, lawyers may also need to get involved. Investigations may involve dealing with proprietary or sensitive information, or groups that are reluctant to divulge information (see "completely uncooperative" above). Some accidents may involve criminal investigations as well, in which case organizations like the FBI may also be involved.

Followup - It's not even over after the report has been released. The agency will still need to follow up with regulatory agencies (like the FAA) about recommendations made in the reports and determine whether the responses are adequate.

Additional Duties - On top of all this, many investigators will not be working on one incident full time. Agencies like the BEA and NTSB investigate many smaller incidents as well. So imagine that at any given time, an agency may have many active investigations at different points in the above process. In the case of the NTSB, all of this is accomplished by about 400 people with a budget of around $100 million per year. While aviation safety is the largest part of this with about 130 employees, the NTSB also investigates highway, marine, and railroad/pipeline/hazardous material incidents. Employees also spend time on other things such as educating the public about safety issues, looking for ways to improve safety, and producing additional types of reports.

$\endgroup$
  • 2
    $\begingroup$ And don't forget politics - that can delay things esp. if there are international "sensitivities" (e.g., see EgyptAir Flight 990 where the Egyptians really disagreed with the NTSB investigation - the article doesn't say it delayed this particular investigation but the NTSB had to do followups for quite a long time, and you can see that the same problems might occur for other investigations which might delay them.) $\endgroup$ – davidbak May 20 '16 at 22:32
  • 1
    $\begingroup$ Then there's good ol' beurocracy. Once they listened to the CVR on the Germanwings crash there was no arguing what took place. Yet it took almost a year to publish the report. $\endgroup$ – TomMcW May 21 '16 at 2:19
  • $\begingroup$ Every once in a while they also take all the wreckage and reconstruct the aircraft, like TWA 800 which took considerable time to just recover off the sea floor. $\endgroup$ – Ron Beyer May 22 '16 at 1:45
  • 1
    $\begingroup$ I would argue with TomMcW about how simple the Germanwings 9525 investigation was. Even when the root cause was obvious, the investigating authorities needed to determine a detailed description of the incident, evaluate ulterior motives, find out if the medical clearance procedures had been properly followed, if psychiatric practices took this outcome seriously enough, whether requiring two people in the cockpit would have helped, and detailed evaluation of the cockpit door operating procedures. As fooot and kevin explain, the authorities leave no stone unturned when investigating. $\endgroup$ – Cody P May 23 '16 at 17:06
17
$\begingroup$

I have not done any aviation investigation (I am far from even qualifying for the job), but I will try to relate it to something I have experience with: research paper writing.

A couple of similarities between both come to mind:

  1. You don't know what you're looking for. You take a quick look at the big picture, pick your best guess (sometimes a wild guess), then dig deeper into that direction. If you're lucky, it works; usually it takes a few trials.
  2. You need to call experts. Even within the field of aviation, there are many experts who specialize only in subfields, e.g. materials, human factors, aerodynamics etc. If you need to analyze the inner workings of a component or system, often times you have to contact the manufacturer. They have to then contact their employees, who may delegate that work to someone in their department, then that person has to pack up and travel. And it's likely you need many experts like these during the investigation.

    • If you suspect a bomb, you need an explosives expert.
    • If you suspect the weather was abnormal, you need a meteorologist.
    • If you suspect a structural failure, you need a materials expert.
    • If you suspect a defect in a component, you need an engineer from the manufacturer.
    • If you suspect pilot fatigue, you need a doctor who specializes in aviation.
  3. You need to collect data. You need to conduct experiments or examinations to collect data. You need to design an experiment that fits your purpose. You need to set them up: apparatus, environment, users, everything. You need to do it scientifically. You need accurate measurements. Typical data I've seen in major accident reports include:

    • Flight data recorder / transcript of cockpit voice recorder. If the blackbox was damaged, technicians will need more time to extract data.
    • Flight history. When did it take off? Which airway it was flying? Do radar tracks deviate from the FDR? What was the weather?
    • Pilots' background. When did the captain get their license? When did he get training? Are there any failed checks in the past? Was the license actually valid and not fake?
    • Interviews. Did the previous crew have problems with this plane? What did the mechanics do? How did the ground crew fuel the plane? What do the pilots who landed just before this plane say? What do friends of the pilot involved think about his flying ability? What about the controllers?
    • Examinations. What was the position of the switches in the cockpit before impact? Did this debris come off because of metal fatigue? Are there signs of fuel leak? Did this break before impact, during impact, or during the post-crash fire?
    • Experiments. Would putting a rubber tire before the engine inlet lead to an explosion? Would the wings oscillate at 30mph crosswind? The manufacturer says the strength of the valve should withstand 100psi. Is that true?
  4. You need to analyze. There is a huge amount of data. Do you find any anomalies? Any trends? Any correlation? Was that an experimental error, or did you find something?

    • Is the scenario reproducible in the simulator? This is common for control surface issues / corrupted flight data in the blackbox / abnormal cockpit warnings recorded.
    • Was it possible for the crew to save the plane? Quite often, for controlled-flight-into-terrain accidents, the report would state something like "if the pilot applied corrective action at or before the 13.7s mark, he would have cleared the trees by at least 200 feet".
  5. You need to coordinate. Only the core investigators (those who get to decide the direction or conclusion of the investigation) have access to every piece of information. The others only get information related to their tasks. When necessary, the team would split up, e.g. two or three people investigate engine fire, another two or three study the charts and the flight path. You need meetings for the teams to present their findings and coordinate. Sometimes you would also need to coordinate with external parties:

    • If you need to run simulator trials, you need find someone with a simulator and volunteer pilots.
    • If you need aerial trials, you (likely) need special permission from the aviation authority and ATC, along with a plane and its pilots.
    • If you suspect the airline's training program is a factor, you need to talk to the chief instructor.
  6. You need to prove. One of the reasons why NTSB reports are so reputable, is that they carry out extensive analysis to prove that every other possible scenario was not the case. They disprove every theory you can imagine, until you are left with nothing but the proposed cause of the incident.

  7. You need to write. A decent research paper takes around a month just to write. You need to organize the content. You need to proofread for mistakes. A full investigation report is several times longer than a research paper.
$\endgroup$
  • $\begingroup$ care to explain the downvote? $\endgroup$ – kevin May 20 '16 at 18:38
  • $\begingroup$ This might be a slightly higher level than I was actually looking for... Steps 1, 5, 6 are kind of inherent to the situation. Steps 2, I'd really like detail as to who (and why). Step 3, I'd like more about what is collected and when it's considered "good enough" (since it will never be perfect). Step 4, I wonder what methods are used, how do they use analysis to suss out more evidence and data? How is that data shared? How many people are involved? Etc etc. Just hoping for an answer that is probably the level of depth right below this one. $\endgroup$ – Jay Carr May 20 '16 at 18:41
  • $\begingroup$ @JayCarr I've extended Steps 2, 3 & 4 based on what I've read in reports. I don't think there are specific methods for step 4, since you don't know what you're looking for. You really are just scanning all the data and compare that with expected values. $\endgroup$ – kevin May 20 '16 at 19:41
  • 1
    $\begingroup$ Not an answer, but you might want to have a look at the investigation required as a result of British Airways Flight 38 where for a long time, there were indications as to the cause but actually replicating the issue to prove it took many attempts over many months before is was successful. $\endgroup$ – Gwyn Evans May 20 '16 at 19:51
  • 1
    $\begingroup$ RE the BA flight 38 investigation, a key quote from the Wikipedia page is "Certification requirements, with which the aircraft and engine fuel systems had to comply, did not take account of this phenomenon as the risk was unrecognised at that time." - so there would have been no "standard" testing procedures or facilities available that could reproduce the proposed cause of problem. Bear in mind that some of the tests with are performed "routinely" to certify engines may take a year to set up, run, and analyse the results, and you may get some feeling for the timescales involved. $\endgroup$ – alephzero May 20 '16 at 22:35
10
$\begingroup$

If you would like a little "light-reading" to answer the question. Take a look at the NTSB investigation manual as well as the 315 page appendix that goes with it. Together they will give you some sense of the investigation process. Just the time to read through that manual will take a while, let alone to perform the processes that it discusses.

$\endgroup$
  • 2
    $\begingroup$ Love how the appendix is over 5 times longer than the actual document... I'll give this a look though, maybe I can write my own answer ;). $\endgroup$ – Jay Carr May 20 '16 at 18:02
  • 9
    $\begingroup$ This answer is entirely correct, but it would be great if you could list or summarize some of the main activities from the manual in your answer, e.g. from the table of contents. As it is, this answer basically says "go read this document", and we usually try to avoid link-only answers here. $\endgroup$ – Pondlife May 20 '16 at 18:42
1
$\begingroup$

This is not a canonical answer to this Q as asked.

But I will add:

  • a slow investigation is not a bad thing, thorough is good
  • a completed investigation is just one part of the value in the holistic process of aviation safety
  • when investigators find something obviously wrong, corrective actions are sent out immediately to the aviation community in various directives. This often happens before an investigation is complete (I'm not an aviation professional, so someone may have good examples of this)
  • the most valuable part of this process is the accuracy of the investigation---and how good the corrective actions address the root cause of the accident
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.