3 variable process controls - I have no idea

[rgray107918]

Brand spanking new to the forum. Registered just to ask this question...

We have a machine at work that has the following parameters:

tempurature
pressure
time

This machine machine processes a prodcut that has two properties:

quantity
stiffnes
shade (Not really sure if the shade actually changes the process properties but thought I'd add it)

My question is whether it's possible to create an adustment matrix encompassing the three machine variables and 3 product properties, to improve first-time-right process adjustments by machine operators?

I'm sure there is a way, but my skills in this area are exceedinly low. I'm more than willing to research possible methods (registered for this forum for starters) I just need a jumping off point, or a nudge in the right direction. Oh, and probably about 3 more years of calculus sprinkled with a little advanced statistics. But a nudge will get me started :)

Thanks, to anyone who chimes in on this.

 

[etudiant]

The easy way to approach this is to monitor the settings and the outputs these settings produce. That was the method used by the paper industry to get a grip on paper making machines.
Hopefully you find a bunch of good settings which you can then use to help new operators avoid costly mistakes.
If you find that the outputs vary even if the settings are the same, you will have learned something about the limits of your control and perhaps you need more parameters to be effective.
Either way, data is needed before you start with the construction of adjustment matrices.

 

[rgray107918]

Thanks for the reply!

We started gathering some settings data, but our problem is we don't have procedures for changing settings. One operator may choose to increase pressure while a different operator may think changing temperature would be the first choice. Additionally, I'm not even sure the same operator is consistent over time. I've heard the same operator say heat is the first adjustment to make one week, and then go on about pressure being the go-to next week.

I just walked the floor and talked to some operators (I'm 2nd shift) and got their opinions on the creation of such a matrix. They all feel they adequately make adjustments on their own, and do not need a matrix to tell them how to run their machines. So, now there's the change-management design along with the systems/controls design. But, it wouldn't be fun it it wasn't a challenge!

Etudiant,

In your opinion, would it be more beneficial to begin with recording operators' settings (even if they're changed in different orders and magnitudes), or to begin with estabishing uniformity across operators' changes?

 

[etudiant]

My $0.02 is that you want to get the operators to work with you, not against you. They can make a program succeed or screw it up more ways than you can dream of. Can you set up a trophy or something for the best output, however defined? Do the operators even know what you want to achieve, is it greater quantity or higher stiffness? What is their pay incentive set on?
As you describe the situation, there is no recording of the settings at all, so no one really knows what works, not even the operators. In the paper making example, the operators were given the ability to record the machine settings when things were working especially well. That provided a basis for the eventual control algorithms. You need something similar, but it does not seem that your machinery has any monitoring built in and settings are modified fairly routinely. Ideally, you want the operators to record their best settings, but unless they can monitor the output and judge it, they cannot do that.
Also, you do not say whether the input material is subject to variation. In the paper case, the pulp is very variable, so there are different settings for differing inputs. Is that true in your case?
As described, you really have more of a case of labor management than of process control logic, at least imho.

 

[rgray107918]

Thanks, etudiant. I'm lucky in the fact that there are several skilled and continual-improvement-minded operators I can work with. It wouldn't be hard to get them on board. The others (majority) are fairly set in their ways and feel management already asks for too much paperwork (we're FDA controlled).

There is variation in the product. It is not a great deal of variation, but it's there and measured. Product comes to us from the lab with a stiffness rating. My overall goal from the beginning of this was to minimize process impact, caused by the variation in stiffness i.e., different settings for differing inputs, and increased throughput.

But, like you said, I think the immediate goal really should be labor management. Unless I can get ALL the operators on board, and engage them to collect quality data, I won't have accomplished anything except making a faulty matrix, based on faulty data, producing faulty product. Drat.

Thanks for the replies, etudiant.

 

[etudiant]

Thanks, etudiant. I'm lucky in the fact that there are several skilled and continual-improvement-minded operators I can work with. It wouldn't be hard to get them on board. The others (majority) are fairly set in their ways and feel management already asks for too much paperwork (we're FDA controlled).

There is variation in the product. It is not a great deal of variation, but it's there and measured. Product comes to us from the lab with a stiffness rating. My overall goal from the beginning of this was to minimize process impact, caused by the variation in stiffness i.e., different settings for differing inputs, and increased throughput.

Do the operators have some way to monitor the effects of their actions? Stiffness is not a parameter that is self evident, so presumably there is some sampling pre and post processing. Is that available in real time to the operators? I assume the process is pretty quick, that the part does not require several shifts to get completed.
Is it possible for you to get useful data just from the more involved operators?
It is not easy to get people to fill in data sheets accurately, usually one gets lots of dittos, even if there has been change. So if you can just work with the willing, you avoid the burden of dragging along the unwilling.
Good luck on your efforts!

 

[Baluncore]

I would use instrumentation to log the actual temperature, pressure and process time parameters that can be controlled. Where possible I would also automatically record stock and product parameters.

When the operators know you are recording the processing parameters they may be less likely to make arbitrary adjustments, but when they do they will be more willing to record the parameters that influenced their decision to change.

 

[dlgoff]

I agree with Baluncore that you should "use instrumentation to log the actual temperature, pressure and process time parameters...".

Having worked in the instrumentation area for a pharmaceutical company, FDA compliance to Good Manufacturing Practice was a must. Records had to be maintained for numerous process and environmental variables, so I used these types of Circular Chart Recorders:

Not only would you have parameter data and records for FDA compliance, it could be used by the operators to "get a better handle" on the process's control. Have the operators know that you are installing these recorders for them. This would be the first step to a more refined SOP, for which you need the recorded data to come up with after some time, and will also give the operators some time to get comfortable with what they're seeing.