Automated Machinery: How many stops is too many? 1

[ataslaki]

I'm not quite sure if this is the appropriate forum for this question, but I'm sure I'll be guided.

I would like to open a discussion today with the following question: in an automated assembly line, where a stand-alone automatic machine assembles 30,000 parts a day, how many fault stoppages resulting in 2 to 5 minutes of down time are acceptable?

The question is really open eded, you might say. "acceptable" could be counted as dollars, or down time for maintenance, etc. However, my question is from a standpoint of the machine maker.

We made this machine, with a group of engineers. Cost us about $400,000 USD. It is almost bullet proof in its production, but unexpected occurences, such as non-conforming plastic parts that are fed into the machine, or operator errors (placing one part into the machine) cause faults and stoppages that are out of our control.

Is 50 faults/day too much? with 1-2 minute of down time for
each fault?

 

[PSE]

As you indicate, this is a bit of an open ended question. If you are looking at validating your machine design, look to see what percentage of the downtime is attributable to machine faults (non-maintenance related). Looking at these faults as well as how much maintenance is needed and energy (facility resources) is used will allow you to compare the design with other "equivalent" machines in order to judge how well it truly performs.

Downtime is the bane of any production line so if there are problems with non-conforming parts (non conforming vs the machine or actual part requirements?), the quality system needs to address the problem. If operator error, training needs to be looked at. The goal being to always drive unplanned machine downtime toward 0.

Regards,

 

[tygerdawg]

This is one of the bugaboos for machine builders and users. The normal request from management is that a machine must produce "N parts per unit time". After much pain and suffering buying machines, I finally developed a rigorous way of specifying machine performance required during trial.

I'd take the "number of parts per year" required to meet the sales forcast (or whatever else you have to work with). Subtract downtime like shift changes, holidays, etc etc to get total available work time. Divide by an efficiency factor to account for laziness, potty breaks, other machines' problems, poor supplied parts, whatever. You'll have to be creative in this efficiency, we settled on 75% - 80% depending on the situation. Then I'd also divide by a guessimate of machine efficiency. THAT would provide me a "N parts per unit time".

After that, I would specify an 8-hour trial at vendor & our site. Breakdowns and stoppages could occur, but they must be recoverable within a certain small period of time and quantity of stops allowed. This was calculated from the overall efficiency and specified "reliability percentage" that we made up. For example, 90% reliable = (uptime/total time). If the stoppage was not recovered in time, trial starts over. At the end of the trial, we had to have a minimum number of parts.

It seems like a lot of fudge and overkill, but rarely was all the incoming parts, etc., perfect enough to do a good trial without this approach. That is, "N parts per unit time" without efficiency and reliability fudge factors assumes everything else is perfect and in lockstep, and this never happens in the real world.

This approach weeded out the unsophisticated or sloppy machine builders. It also made the management spend more money, but now they had a semi-rigorous justification to do so. It seemed to work well for us. This method built machines that were overcapacitized, but when the crush was on to hurry up and produce parts, they would crank'em out faster than they needed to.

TygerDawg

 

[griffengm]

I think there are ways of addressing some of the nonconformance issues during trials. For example, we all know that there will be some number of non-conforming parts. Whenever these are identified, keep a sample to mix into the trial run. Color them differently for various defects so that when one stops the process, it is easily identified.
Operator issues can be similarly addressed by mixing trained and "semi-trained" at the trials.
Combine the two techniques and with a little design of experiments input you should be able to develop a reasonable prediction of downtime.
Additionally, the vendor should provide a maintenance schedule that can be used to factor in the PM time.
Tyger, Do you have any "feel" for the cost adder for levels of spec development on production machinery?

Griffy

 

[tygerdawg]

Griff
Hmmmm....maybe +20% ? Maybe +30% ? I really am guessing, I have no way of giving actual quantitative data on that. And I second your suggestion of inserting known bad parts in order to see if the machine caught them. I did that, too, but forgot to mention that as part of the trial requirements.

Being younger and cursed with an engineering mentality, it always made me grind my teeth that "perfection" in manufacturing couldn't be attained. It was always me yelling in frustration "Well, why AREN'T all the incoming parts meeting a six-sigma quality level??" and "How come the supplier can't deliver his parts Just In Time??" and "Doesn't Design Engineering know that this machine wasn't designed to build this new part?? " and "Joe quit, so now nobody knows how to run the machine properly."

Then I grew up and realized that Production typically ruled the roost and all that really mattered was providing them a way of making parts as fast as material could be delivered, within reason. As long as a semi-rigorous justification for extra machine costs was there, Management would be much more willing to approve the extra fat I specified into the machine design.



TygerDawg

 

[ShadowManager]

As an opener, I design and build automated assembly machinery to assemble stamped and molded component parts into a functional assembly. The company I work for mass produces Automotive Door Checkers in a tier 2 field. Ultimate frustration of company owners and managers alike point the finger at the machine designer/builder when it comes to down time. Fault conditions are mainly attributed to non conforming parts over anything else. Molded component parts as a norm retain tolerances quite well but dealing with metal stampings is a whole other ball park. Once these stampings go through the plating process, parts arrive bent, twisted, warped etc. General criteria in machine design must accommodate to constant varying conditions for smooth operation. A key factor to keep in mind during the design phase is to limit the use of rigid fixture tooling and adopt a tooling system that adapts to changing conditions of supply parts. In process checking although sometimes costly always pays off in the end. Non conforming parts should always be kicked out of the process either physically or through programming methods. This method literally removes operator error from the equation and can isolate non conformances while maintaining up-time. As a general rule of thumb, if production requirements dictate 20K parts/ 8hr shift, the machine should be designed to produce the same in 5 hours, allowing a three hour window for maintenance and supply shortages.
After all, a machine is only as good as its designer and programmer.



Terry A Hrushka Ph.D., M.S.Eng., B.Sc.

 

[ornerynorsk]

The number of stoppages does not seem to be the question here. Can you foolproof the feeding to prevent or segregate parts that are mis-oriented or non-conforming parts. Better yet, why are non-conforming parts making it into the assembly stream at all? Simplistic, but good solutions sometimes are.

 

[Drazen]

As PSE mentioned, the only legitimate target for number of fault is zero! Anything else doesn't make sense for several reasons. To avoid overdetailng, I'll mention one: how can you deliberately design machine to have say 3 small failures per day? Theoretically, there should be some relation between money invested in research and construction and number of faults that occur in operation.

In practice it cannot work because the failure is unpredictable event and you cannot plan it! The only way is to have zero target ALL the time when designing and building and then, in real life, some relation will be established between real number of failures and level of invested resources.

That really means that if you plan to ask your boss for extension of budget and need to justify it with higher level of stated goals, than do your absolutely best, plan resources for zero defects and go for it! You will never attain it but there is lot of job that can be done: from continual improvement, system of feedbeck from clients maintenance dept.to good after sales service. Life is not ideal, but you can do a lot.

Forget about investing 5% more money to achieve 2% less failures or something like that...