10/15/2014 - Tech Talk - Engine Series: Part 1

[ross]

Wasn't aware six sigmas was not the norm in automotive manufacturing but good to get corrected. We hear it a lot in my business, though I am far from an expert at the manufacturing end of it and am in R&D myself. Yikes. Gantt charts are the bane of all existence - never thought I see that one written here, but I agree totally and your timing sounds much more reasonable to me.

As for six sigmas, it is an esoteric topic but you've piqued my curiosity. I guess if QA/QC catches parts that fail before they go into the vehicle fails in the thousands would be fine. But if QA/QC does not catch parts and if the failure of any one part in, say, a motor of 500 parts dooms the whole, then the probability of engine failure would be something like 1 in 2 for a motor with 500 parts and a 1/1000 chance of any one part failing. In other words, in stats, p(A or B) = p(A)+p(B), so 1/1000 + 1/1000 + ..... n number of parts. I get the feeling you probably know this. Just a question - not a dig - I'm genuinely curious.

As for getting run out of town, that's fine. I'm shouldered up against the Pacific so as long as I run west and not east they can only run me about 4 miles. I guess I'd just have to surf and go all ninja quiet on the site til Elio Motors got all sorted out and reduced to practice.

What language is that?

 

[ross]

I'm going to rescue myself and hijack this thread back to discussions about the Elio engine by posing a question to those far more knowledgeable than myself in these matters.

Elio Motors originally evaluated the Geo Metro motor and decided that while the basic design principles were sound ones, the motor carried excess weight and had (I'm going out on a limb here) a torque/horsepower curve that was unneeded to power a vehicle so much lighter. And, rather than fiddling about with mods Elio decided to do a melt and re-pour, designing a new engine with the same basic framework but much lighter. The question is: wouldn't there be liquid cooled motorcycle engines that would meet these same requirements as the Elio that would not have necessitated an engine re-design. I realize that motorcycles are not as fuel efficient but this probably has as much to do with their relatively high drag coefficient than basic engine design.

Keep in mind, one of the BIG requirements is it needs to be made in the good old U S of A. To my knowledge I don't think a Harley or Victory V twin is going to suffice, and I for one really really REALLY don't want a rice grinder engine in mine!

 

[Snick]

Wasn't aware six sigmas was not the norm in automotive manufacturing but good to get corrected. We hear it a lot in my business, though I am far from an expert at the manufacturing end of it and am in R&D myself. Yikes. Gantt charts are the bane of all existence - never thought I see that one written here, but I agree totally and your timing sounds much more reasonable to me.

As for six sigmas, it is an esoteric topic but you've piqued my curiosity. I guess if QA/QC catches parts that fail before they go into the vehicle fails in the thousands would be fine. But if QA/QC does not catch parts and if the failure of any one part in, say, a motor of 500 parts dooms the whole, then the probability of engine failure would be something like 1 in 2 for a motor with 500 parts and a 1/1000 chance of any one part failing. In other words, in stats, p(A or B) = p(A)+p(B), so 1/1000 + 1/1000 + ..... n number of parts. I get the feeling you probably know this. Just a question - not a dig - I'm genuinely curious.

As for getting run out of town, that's fine. I'm shouldered up against the Pacific so as long as I run west and not east they can only run me about 4 miles. I guess I'd just have to surf and go all ninja quiet on the site til Elio Motors got all sorted out and reduced to practice.

It can be boiled down very simply: each part is assessed for it's ability to either cause catastrophic failure directly, or by being the head of a chain (cascade failure) of failures. A number is assigned for risk assessment severity. Then, occurance is predicted from past historical review. Finally, Detection capability is assessed: how can either the user or the ECU detect imminent failure before it becomes catastrophic?

All the numbers are crunched through an xFMEA series (Design, Use, and Process Failure Mode Effect Analysis') and you get a risk factor such as an "RPN" for Risk Prioritization Number. You tackle the biggest RPN's with the most stringent validation tools (increase sample size, tighten tolerance bands, increase margins of safety) and run your series of tests or a Design of Experiment. <-click for background link. And a background link on the software used to set up multi-variable designs--we use JMP-SAS.

DoE's are most robust if set up well an the right variables selected with good adherence to orthagonality of design.

At the end of it, you know what manufacturing features or critical dimensions need to be tracked most closely BEFORE you even accept the part. You should also learn what the most common assembly boo-boos are and then you either train especially hard or try to change the process or part design the make it harder for assembly boo-boos to occur. Finally, if done well, you know what the user is most likely to screw up and where you should put a "lifetime" fluid or lock out the device from user tampering.

For less critical parts, you have to accept fallout rates in the thousands or tens of thousands, or you'd never stay in business. Some parts you might even accept fallout rates in the low thousands range. Examples are interior bits and window regulators. There's a reason your modern car's electric window fails after a few thousand uses. It's fully allowed!

Whereas for parts that could cause explosions or user deaths or dismemberments, dropout rates *after* final QC should be in the tens or hundreds of millions.

 

[Snick]

I'm going to rescue myself and hijack this thread back to discussions about the Elio engine by posing a question to those far more knowledgeable than myself in these matters.

Elio Motors originally evaluated the Geo Metro motor and decided that while the basic design principles were sound ones, the motor carried excess weight and had (I'm going out on a limb here) a torque/horsepower curve that was unneeded to power a vehicle so much lighter. And, rather than fiddling about with mods Elio decided to do a melt and re-pour, designing a new engine with the same basic framework but much lighter. The question is: wouldn't there be liquid cooled motorcycle engines that would meet these same requirements as the Elio that would not have necessitated an engine re-design. I realize that motorcycles are not as fuel efficient but this probably has as much to do with their relatively high drag coefficient than basic engine design.

Yes! Any of a dozen easily obtainable twin-cylinder, 500-800 cc engines could have powered this for far less development timeline while meeting most of the mpg goal at some NVH cost (but not unbearable). I think a fuel injected, 4 valve/cylinder, 650cc V-twin from the Hyosung could be had in volume numbers for ~ $900/engine, but they would have had to mated the gearbox to a differential, so some engineering time involved, still.

 

[harlan stephens]

More talk and no running engine. ..
Sos

 

[wheaters]

Most modern motorcycles engines have relatively small, integral clutches and gearboxes and are designed for rear wheel drive. As Snick said, redesigning them to fit a front wheel drive vehicle is not straightforward. Also, most motorcycle engines are high revving, relatively low torque designs, which may not be best suited to a heavier vehicle, especially where longevity of the clutch is concerned. Not to mention fuel economy.

The Geo Metro engine, designed by Suzuki, is a very sound one, it's rare for anything to fail, especially as it's a non-interference design. Using it as the basis for a modern update makes a huge amount of sense, especially bearing in mind that it has a car type dry clutch and fwd transmission.

 

[Dustoff]

It can be boiled down very simply: each part is assessed for it's ability to either cause catastrophic failure directly, or by being the head of a chain (cascade failure) of failures. A number is assigned for risk assessment severity. Then, occurance is predicted from past historical review. Finally, Detection capability is assessed: how can either the user or the ECU detect imminent failure before it becomes catastrophic?

All the numbers are crunched through an xFMEA series (Design, Use, and Process Failure Mode Effect Analysis') and you get a risk factor such as an "RPN" for Risk Prioritization Number. You tackle the biggest RPN's with the most stringent validation tools (increase sample size, tighten tolerance bands, increase margins of safety) and run your series of tests or a Design of Experiment. <-click for background link. And a background link on the software used to set up multi-variable designs--we use JMP-SAS.

DoE's are most robust if set up well an the right variables selected with good adherence to orthagonality of design.

At the end of it, you know what manufacturing features or critical dimensions need to be tracked most closely BEFORE you even accept the part. You should also learn what the most common assembly boo-boos are and then you either train especially hard or try to change the process or part design the make it harder for assembly boo-boos to occur. Finally, if done well, you know what the user is most likely to screw up and where you should put a "lifetime" fluid or lock out the device from user tampering.

For less critical parts, you have to accept fallout rates in the thousands or tens of thousands, or you'd never stay in business. Some parts you might even accept fallout rates in the low thousands range. Examples are interior bits and window regulators. There's a reason your modern car's electric window fails after a few thousand uses. It's fully allowed!

Whereas for parts that could cause explosions or user deaths or dismemberments, dropout rates *after* final QC should be in the tens or hundreds of millions.

WOW!
You had better educate the Engineers at IAV.
They may not be aware of this simple explanation.
Oh wait, they may have done this before.

 

[Jay3wheel]

How many motorcycle engines does Honda make?
How many motorcycle engines are in any Hondas other than motorcycles?

If I am not mistaken (which I often am), motorcycles and therefore their engines don't have to reach the same level of pollution control as an auto-engine.

 

[tonyspumoni]

Whereas for parts that could cause explosions or user deaths or dismemberments, dropout rates *after* final QC should be in the tens or hundreds of millions.

Snick,

Thanks. I appreciate you taking the time to walk me through this. I'm in drug development and work for a major pharma company. All we do is make pills and QA/QC is naturally a huge deal insofar as I get near the manufacturing side.

 

[goofyone]

Similar argument can be made for the Transmissions,
there are definite economies of scale at play
here. Aisen, ZF & others make private label Trans.
for the global market place & Elio has been shopping
there for some time trying to get the right 'fit', if you
will. Thereby meeting their cost/quality/quanity constraints.
The motor is a different animal altogether, your points
are well taken............

You are correct that there is a big difference between the markets for transmissions and engines.

Transmissions are often manufactured by either outside suppliers or subsidiaries which are also industry suppliers so the transmission business is structured with plenty of transmission choices and the manufacturing capacity to make supplying large numbers of transmissions possible.

Engines are not typically designed and manufactured for sale to other companies and when this does happen it is nearly universally in low volumes such as the Polaris Slingshot which uses a Ford engine however slingshot production will be measured in thousands each year not tens or hundreds of thousands. For higher engine volumes these transactions happen almost exclusively because of a partnership between two, or more companies, in which the partners buy into a joint venture and/or because partners own a portion of the other partner. A great example is the GM/Suzuki joint venture which was based around the fact that GM actually owned part of Suzuki.