Subscribe to DSC Newsletter

Boeing's Dreamliner turns into a nightmare due to bad analytics

The problems with this plane, which forced FAA to ground all of them today in US, is its new type of lithium-ion battery made in Japan, and never used in a plane before. This type of powerful battery is overheating easily, catches fire etc. resulting in a number of emergency landings over a short period of time.

Now I have a few questions:

  • Aren't these batteries (like pretty much any product that you can purchase, such as car or laptop battery) going through extensive quality control testing, using sound statistical techniques to make sure that faulty batteries, or risk of failure over the lifetime of the product, is below an acceptable threshold, say below 0.001%?
  • Could it be that the quality control tests were not performed according to best practices? Maybe overheating simulations were not representative of real word conditions as found in an airplane taking off, or did not "stress" the battery for long enough? From a statistical point of view, this would amount to poor design of experiment. Something the FAA should carefully look at before allowing the Dreamliner to fly again.
  • Maybe standards for quality control are lower in Japan? Or greed played a role? Maybe statistical reports about the reliability of these lithium-ion batteries made by this Japanese company were fraudulent or wrong?
  • Maybe Boeing could have used better mechanisms to cool this type of battery, which has never been used in an airplane before, but found in all cell phones, and responsible for spectacular cell phone fires in the past. Unlike in a cell phone or a laptop, it is very easy to cool (and freeze) anything in a plane since the outside temperature is well below freezing point. Could this be used to fix the problem?

What are your thoughts?

Views: 4998

Reply to This

Replies to This Discussion

It wasn't just the batteries that caused the airplane to be grounded. 

I suspect that analytics use (or lack of use) is not the issue.  I would guess it is the problem of building a plane with a host of subcontractors and trying to assemble their parts.  i'm sure Boeing is reviewing the subcontractors who were responsible for the failing parts to determine how to solve the problem and get the airplane back in the air. 

Agree w/ Alex. Anything of this sort can be spinned as an analytics problem, but I would point more towards policies and procedures used by the subcontractor for the battery part as the most likely point of failure. Those might be based on some analytics, but saying this an "analytics problem" is not worth it.

I would only be speculating here, but before anyone jumps to folks being dishonest or greedy, let us think back to the Space Shuttle and O-rings.  The reliability of the O-ring had not been tested at the unusually cold temperatures on the day of the launch of Challenger, and so the likelihood of failure under that condition was grossly underestimated.

I suspect there is something similar going on here.  On their own, the batteries don't overheat much, but under the conditions in which they are being used in this airplane the reliability stats don't mean anything because the conditions are fundamentally different.   When this happens, you can end up with finger pointing (supplier blames airplane manufacturer for using the battery outside of its specified usage, or manufacturer blames supplier for not robustly testing their product).

Analytics may not be the root cause of the problem, but perhaps analytics could have prevented a problem from happening.

 It's a classic operations research issue: quality control of a product involves both sound design of experiments (DOE in statistical terminology) and statistical models. At the very least, DOE was poor in this case.

The statistical testing used for any product (to make sure fewer than say 0.01% units will fail over their lifetime) guarantees that the product is reliable enough for its purpose, so that the manufacturer won't face recalls and lawsuits.

To put it differently, better analytics would have got the battery in question rejected until it passed the quality control tests successfully. In this case it passed these tests because these tests were (from an analytic point of view) faulty.

Or maybe non-existent.

I agree with you Titus and Keith.  Classic Ops research problem.  As an operations research analyst for the last 13 years my experience has taught me to use a variety of techniques to answer questions/solve problems.  And then I think it is a matter of the System of Systems approach and appropriate testing of the batteries in extreme conditions (the environment in which the airplane operates).  Need to look at the reliability engineering/analysis of the batteries (MTTF, MTBF, etc) and under which conditions?    

In the end certainly better and more appropriate analyses COULD have prevented this...

The cost of testing 5,000 batteries in (simulated) flight conditions (low atmospheric pressure etc.) for extended time periods as well as testing all other critical / dangerous components (fuel tank, tires, etc.) and how they interact, might be very high. But it should be measured against the cost of deadly disasters, losing out to competitor Airbus, or the cost of possible bankruptcy should Boeing be unable to sell the 5,000 Dreamliners it expects to sell in the next 10 years (should trust in the Dreamliner evaporates).

Certainly - several critical factors to look at.  However, as you stated these batteries had never been used before.  Maybe just component testing and not necessarily interoperability testing could have made the difference.  I like your idea of measuring it against the alternatives.  Could be a good problem set involving expected value and decision making.

It wasn't the battery.

Since human life could be lost from failed lithium batteries, I would imagine that greater care was taken in designing tests and in simulating extreme flying conditions. I don't think they mis-interpreted data as in the shuttle disaster, because we usually learn from our mistakes when lives are lost. It's just guessing unless we know more about the testing and the analysis of results. But I'm speculating that greed and competition may have favored shabby testing, which would have been expensive and time-consuming if it was done right. And I do agree that this case is similar to the shuttle disaster if (1) they did not test for extreme environmental conditions, and (2) these batteries are similar to batteries that failed in cars.
Pls enlighten us Jim

The FAA stated that the battery wasn't the problem. It was well within it's temperature range per the FAA.

RSS

Follow Us

On Data Science Central

On DataViz

On Hadoop

© 2017   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Terms of Service