Tony D Posted January 30, 2010 Share Posted January 30, 2010 My most famous 'to specification' story revolves around someting W. Edwards Demming would talk about during his lectures. It involved JATCO (Japan Automatic Transmission Company) C4 Transmissions made for Ford in the 70's. Seems Ford was plauged by a 50% failure rate on these units in the fleet. Not specifically the JATCO stuff, but the C4 in some models would fail about half the time. The transmissions when investigated were all 'within spec'. This was at the beginning of Ford's SPC program, and someone there noted they had two suppliers... When investigating further, they found that the JATCO supplied transmission had a statistical ZERO failure rate. The transmissions from the other supplier had a 100% failure rate! They got some examples in from the field, and true to what was noted in all the failures above, the trannies were all 'within spec'---when the JATCO transmissions were disassembled though, they couldn't measure a difference between parts. That is to say, each transmission disassembled did not have a RANGE of tolerance, with the tools the Ford engineers were using, they couldn't MEASURE a difference from transmission to transmission. They were 'identical' as far as they could tell. So there you have a classic case of what is referred to as 'tolerance stack'---meaning if you measure to a RANGE, then 'in spec' says it's good. But if you have something at the low end of the acceptable specification range coupled to something to another component at the high end of the acceptable specification range...and then bolt it to a third component.... all those 'acceptable specifications' add up to something so loose and terribly assembled that while it may work it doesn't work very long. And this isn't bashing Ford, they figured out the issues and saw the value of repeatable narrow production tolerances. Several years later Volvo got a bad batch of ZF Trannies... 100% failure rate by 35K miles. It was just a matter of time till they reached the mark and failed. D'OH! Working for a compressor company we got a bad batch of bearings from SKF one time, and we knew every element assembled with those bearings would fail prematurely. Some stuff you can't help, it happens. But measurement and tolerance creep IS totally controllable. Keep your tolerances within a range for the intended use of the machine. Don't get crazy (I have seen guys plane a head 0.001"---bit nutzo for me when the block is flat!) but never accept nebulous answers. If they did an inspection, they should be able to give you numbers. If the guy does like most do, they grabbed a machinist's straightedge and a 0.003" feeler gauge and started poking about. If it didn't slip under anywhere, 'it's in spec'... That's where the head going one place, and the block to another can cause a problem. This is probably the only site where I can say 'Be an in-tolerant buggar!' and not get lovey-yellow submarined to death about being inclusive and all that rot... Be intolerant of tolerance! 1 Quote Link to comment Share on other sites More sharing options...
Drax240z Posted January 30, 2010 Share Posted January 30, 2010 As someone who works with a couple hundred mechanics on a daily basis... if they don't/can't give you a number, chances are they didn't measure. I can't count how many cases where I've measured something after it was passed in an inspection, and found it's actually a failure because the mechanic only eyeballed it. Quote Link to comment Share on other sites More sharing options...
jmai86 Posted January 30, 2010 Author Share Posted January 30, 2010 Be intolerant of tolerance! Awesome line Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.