How the concept of infant mortality came into general knowledge to the military

[Frank1940]

Back in 1963, I was hired out of college by Western Electric as an Electrical Engineer was sent off to a company school for additional training in the ways of WE and AT&T.  One of these courses was Statistical Quality Control (Of which Bell Labs was one of the leaders in providing the Mathematical basis that was to lead to its adoption).  Acceptation of Statistical Sampling for quality control was not universal in the early days.  (Many people could not believe that looking at 20 to 50 parts out of 5,000 to 10,000 parts was enough to determine if that lot of parts was acceptable or not!)  This is the story of one of the events which transpired back in WWII convinced the Military that there was some real validity behind all of this mathematical hocus-pocus.

 

In WWII, the US got  involved in the European theater against Germany and Italy with the invasion of North Africa in November of 1942.  The campaign lasted for about seven months ending the following May when the Axis force of 275,00 men surrendered at Tunisia.  The Allies quickly realized that they had a numerical superiority over the Axis forces in the south of Italy if they acted quickly. So they began planning the invasion of Sicily for early July of 1943. 

 

This required refitting and updating much of the equipment that had been used in the North Africa campaign.  When it came to overhauling the tanks, it was decided by someone in the chain of command that all of the vacuum tubes* in the radios of all of the tanks used in the North Africa Theater would be replaced with new tubes.  They also ordered that the oil and all filters be changes.  All was done.  Those radios were vital in the control of tanks from both a strategical and tactical standpoint and they figured this would prevent the failure of those radios in critical situations.

 

Suddenly, the radio repair facilities were overloaded with failed tank radios.  The radios were failing at rates that had never occurred during the entire campaign and the army was basically stationary in North Africa waiting for the invasion to start.  It was so bad that a study was ordered into the cause of the problem.  I believe Bell Labs was assigned the task.  What they discovered was that failure rate of vacuum tubes had a extremely high rate in the first few hours of operation.  The rate would then drop and remain almost constant for few thousand hours and then began to rise again.  This mode of failure was dubbed the Bathtub curve. 

 

Why had the Army seen this phenomenon earlier?  Well, they had!  But at the time, tanks were being delivered to the Army in small qualities on virtually a weekly basis.  The tubes were failing in same manner but the failures were being spread out over time so it wasn't really noticed.  When all of the tubes in a 1000 tanks were changed out in a very short period of time, the failures overloaded the repair facilities thus creating the problem.  As far as I know, most of the failures did occur before the actual invasion so it probably didn't cause major operational issues in actual invasion.  (If it did, it was hushed up so that the Germans would not hear about it.) 

 

By the way, the bathtub failure curve is still a valid one for every electrical device that I have heard of.  With many components today, the early failures are not quite as pronounced as with vacuum tubes.  But as recently as the mid 1990's, many mail order computer manufacturers advertised a 72 hour burn-in period to reduce those early failures...  They must had a good reason to do so! 

 

* I cannot think of another electrical device ever manufactured in large volume that is as unreliable as the vacuum tube.  I worked in a Radio-TV repair shop during high school and the failure rate of vacuum tubes guaranteed that I had a job as 80% of both radios and TV could be repaired by replacing a tube that had failed.  If it hadn't been for that failure rate, the shop would have had to close!

[RobJ]

Thank you Frank!  I too was hoping you would tell us the story.

[c3]

This failure pattern is not limited to electronics. Happens for humans too

 

 

Image from https://www.ssa.gov/policy/docs/workingpapers/wp111.html

 

And airplane engines, and even ships (shakedown cruise anyone?).

 

Also noted in WWII, something called Waddington effect. Basically, repair/touch anything and it will increase failures.

http://blog.aopa.org/opinionleaders/2014/01/14/the-waddington-effect/

 

 

[eroz]

When I first read the title I thought it was a spam bot. Then I saw it was not a new member, do i I was confused. Love the story.  :D

[CHBMB]

Great story, thanks for taking the trouble to post this Frank.

 

Sent from my LG-H815 using Tapatalk

 

 

[Frank1940]

 

Also noted in WWII, something called Waddington effect. Basically, repair/touch anything and it will increase failures.

http://blog.aopa.org/opinionleaders/2014/01/14/the-waddington-effect/

 

Another story to illustrate this:

 

Back in the early 1970's, I was working as a Test Set design Engineer for electro-mechanical devices--- telephone relays and Crossbar switches.  Solid state logic boards (NAND gates, binary counters, Analog IC Amplifiers, etc. ) were becoming available and reasonably cost effective that I decided to design a new family of tests to replace the twenty year old designs which were using relays and vacuum tubes amplifiers.  The project actually went fairly smoothly (mainly because I decided to over-designed the sets to prevent any problems from  both the physical and electrical environments of the factory floor).  But I was TOLD that the calibration schedule had to be identical to test sets designed with the older technology. 

 

Since we were running two shifts a day  (We were making 5M relays per year at that time!), the calibration on working in-spec test sets was done on the third shift by personnel who were totally unfamiliar with the design.  The day after calibration, 95% of the test sets would be down and the first shift test maintenance would have to re-calibrate them again.  In reality, the new designs would hold calibration for three to four times longer than the older design...

 

You might say why not just change the calibration schedule?  Well, we had a permanent in-house group that reported directly to a Bell Labs organization in New Jersey who were doing quality control inspections for the Telephone Company (Western Electric's customer) and they were the ones who controlled the calibration intervals.  (I quickly learned that in a big organization that it best to pick carefully the battles to fight that you had a chance of winning...)