[AT] Statistics In Tractor Manufacturing Was Bicycle Program

Sun Sep 15 09:04:19 PDT 2019

With respect to today’s tolerencing versus yesterday’s tolerencing, I spent over thirty years doing engineering on the Abrams main battle tank. When we first started, it was on a drafting table using t-squares, pencils, and plenty of erasers. Tank went into production and remains one of the world’s best till today, able of taking hits ffom Soviet designed tanks with very little or no damage in battle. (Somewhere I have a pictured a Soviet round fused to an Abrams turret front glacious, unable to penetrate the armor.)

As with most industries we switched over to a CAD system for later upgrade programs. The army wanted to convert all of our hand drawn drawings into CAD generated drawings so that they could be sent to suppliers electronically instead of through the mail. We started a program to do so. Only problem was that the CAD system changed the tolerencing to what can be manufactured today. We started to have issues with not being able to assemble tanks. Parts were too long, too short, through holes didn’t line up between mating pieces, etc. We had to go back and revise the new CAD generated tolerances to the old hand calculated tolerances. It was obvious that the older draftsmen knew more about manufacturing issues than today’s CAD operators and designed with those issues in mind. It was also obvious that all designing shouldn’t be done as if you’re building a fine Swiss watch.

During WWII the German’s built a transmission for the Tiger tank that was supposed go for 5000+ hours without needing maintenance, allowing their tanks to be on the move constantly with stopping for any maintenance. The highly designed parts were all designed and machined to very high tolerances. Worked well until something did fail. At that point it took six or seven engineers (not mechanics) to work on fixing the issue, and the tank could be out of service for weeks while obtaining and fitting all the precision made pieces back together. Needless to say, those tanks saw very little battlefield time, and they went back to the original transmissions.
----- Original Message -----
From: Jim Thomson <macowboy at comcast.net>
To: Antique Tractor Email Discussion Group <at at lists.antique-tractor.com>, James Peck <jamesgpeck at hotmail.com>
Sent: Sun, 15 Sep 2019 08:41:24 -0400 (EDT)
Subject: Re: [AT] Statistics In Tractor Manufacturing Was Bicycle Program

I have used statistics for the past 25 years or so in a manufacturing setting. I would bet the farm that if a good root cause investigation took place on this model from NH, one of the results you would probably find that every component or system was designed without the interactions with other systems taken into consideration. For example, I led a investigation for a class III medical device that was under FDA recall notification for a major global medical device manufacture. The device was a defibrillator which is considered a life sustaining device. The device was failing its IPX5 water ingress test and devices  were shorting out in the field. To sum things up, the R&D designers put the design on paper, threw it over the wall to manufacturing. As the individual processes of components varied over time, there was a tolerance stack up issue which caused the cases not to seal correctly. If, the variability of the individual components were modeled using a Monte Carlo simulation which takes this into account, the problem would have been addressed before it was released to the public. Toyota does this with their cars on all their systems. This is why they are so reliable but even they miss things once in a while. On the antique tractor side of things, they were designed with generous tolerances and overbuilt for what they had to do. That is why they are still around after many years of use. The John Deere G is a perfect example of a over built tractor designed to last.

Jim Thomson
Rehoboth, MA   

> On September 14, 2019 at 11:26 AM James Peck <jamesgpeck at hotmail.com> wrote:
> 
> 
> Some years ago I did participate in an academic Statistical Quality Control course. Much of the course involved the Weibull Distribution. If I remember correctly, the Weibull Curve predicted the lifespan of a manufactured assembly such as a tractor.  It appears to be a gift from the mathematicians.
> 
> https://en.wikipedia.org/wiki/Weibull_distribution
> 
> The same distribution probably predicts that Cecil’s New Holland money pit will continue to be so and will suffer an early demise.
> 
> [Stephen Offiler] I never took a whole semester of Statistics.  Instead, we had a course called Engineering Experimentation, which was heavy on experiment design and statistical data analysis.  That gave me a very good appreciation for the practical application of statistics.   Out in the real world, on-the-job training programs in quality control principles in manufacturing (Deming, Juran, Lean Six Sigma) continued to solidify the practical applications.
> 
> [Cecil Bearden] Steve: I nearly flunked statistics I only passed because I was a graduating senior.  However, I did flunk Rocks & Clods 2124 and had to find another 4 hours to graduate.   Then 35 years later I retire as a Geotechnical engineer designing foundations. !!!  
> 
> [Stephen Offiler] I'm not sure if that is an interesting statistic, or simply predictable statistically.  All you just said is that a bell-curve distribution for 2-year degrees overlaps a bell-curve distribution for 4-year degrees.   
> 
> [ James Peck] The interesting statistic is that some technical 2 year programs have higher starting incomes than many 4 year degree programs. People who complete such a program can later take a 2+2 program to get a four year degree if they choose.
> 
> _______________________________________________
> AT mailing list
> AT at lists.antique-tractor.com
> http://lists.antique-tractor.com/listinfo.cgi/at-antique-tractor.com
_______________________________________________
AT mailing list
AT at lists.antique-tractor.com
http://lists.antique-tractor.com/listinfo.cgi/at-antique-tractor.com