Using Control charts on shopfloor filled by operators

[RGohil]

Hello Group,

I have a question regarding use of control charts for monitoring a process on the shop floor. Please redirect me to a previous thread if this has already been discussed.

Does anyone use control charts on there shopfloor that are supposed to be filled by the operators. We currently have a X-R charts used at one machine. The intention was to use it for informing the operator (by number and plots) if the process is showing any trends. Unfortunately none of the operators are filling this because they think that if the outcome is within spec they dont have to worry. I understand that the spec are not the limits on a control chart but in your workplace does anyone take the reading recorded by the operator and actually calulate the Rbar and Xbar with control limits? For an operator to calulate overall X bar and Rbar for about 10 reading of each subgroup of 5 its a very tedious and time consuming process. OR do you have taken the UCL and LCL for X-R chart to be a constant value from historical charts and use those on the charts everytime that paricular product is run. That is another problem the specifications change for different product so do you maintain UCL and LCL for each product spec.

Operators on the floor were suggesting that simply recording the number is good for them to undertand the process(this is because I feel they have the product spec in mind as a control rather than variation, which is what we need to see). But with our original intention in mind of checking trends I feel a simple median chart should be good. Just one graph with all reading and median value plotted. I dont even need Rbar or Xbar. I as an engineer can do that at a periodic interval to check for variation. Currently charts without any calculations gives a very bad impression and doesnt earn credibility with the operators. Also during my ISO audit the auditor may write me for this.

I would like to request the group to suggest me if my thinking is correct and suggest me a practice they follow in there plants.

I appreciate your help in advance,

Thanks
Robin

 

[Wes Bucey]

In my high tech machining business (1990 -2000), I eliminated Quality department inspectors and all operators maintained their own charts. We were able to do this because the machines, once set up, were computer controlled and did not need the constant input from an operator.

Each operator was in charge of a "cell" of machines which he programmed and ran. For first article inspection, each operator calibrated his own instruments and did a complete first article inspection according to a Control Plan jointly created by operator, Quality department, and customer. He entered the data from his readings. then he took the first article to another operator who would make a redundant first article inspection with HIS set of instruments (or the joint use of shadow graphs, CMM, etc. in the Quality lab.)

When we began, we used paper charts and operator did his own calculations for cpk, trained by our Quality department, and sometimes spot checked to assure training was effective. When we were able to afford them, we added SPC software to do calculations and ultimately to instruments wired directly into computers for real time SPC on critical dimensions.

We sometimes ran in-process SPC charts on "interim" dimensions (which would later be changed to a "final" dimension.) The interim dimensions SPC told us whether our process was operating effectively BEFORE we went on with more expensive machining. (For example, if we were cutting off lengths of a bar to feed into a grinder, we would chart the length even though the cut-off piece was being fed automatically to the grinder in the same work cell and ordinarily would not be touched by human hands between machines.)

What did the Quality department do?
I am often asked, "If the operators did the inspections and charting, what did the Quality Department do?"

Quality department acted as training center, customer and supplier liaison, and "court of last resort" when questions would arise. Quality folk conducted root cause and DOE with cooperation and assistance of operators and/or suppliers and even [sometimes] customers who were damaging parts in their handling procedures.

 

[Al Rosen]

Hello Group,

I have a question regarding use of control charts for monitoring a process on the shop floor. Please redirect me to a previous thread if this has already been discussed.

Does anyone use control charts on there shopfloor that are supposed to be filled by the operators. We currently have a X-R charts used at one machine. The intention was to use it for informing the operator (by number and plots) if the process is showing any trends. Unfortunately none of the operators are filling this because they think that if the outcome is within spec they dont have to worry. I understand that the spec are not the limits on a control chart but in your workplace does anyone take the reading recorded by the operator and actually calulate the Rbar and Xbar with control limits? For an operator to calulate overall X bar and Rbar for about 10 reading of each subgroup of 5 its a very tedious and time consuming process. OR do you have taken the UCL and LCL for X-R chart to be a constant value from historical charts and use those on the charts everytime that paricular product is run. That is another problem the specifications change for different product so do you maintain UCL and LCL for each product spec.

Operators on the floor were suggesting that simply recording the number is good for them to undertand the process(this is because I feel they have the product spec in mind as a control rather than variation, which is what we need to see). But with our original intention in mind of checking trends I feel a simple median chart should be good. Just one graph with all reading and median value plotted. I dont even need Rbar or Xbar. I as an engineer can do that at a periodic interval to check for variation. Currently charts without any calculations gives a very bad impression and doesnt earn credibility with the operators. Also during my ISO audit the auditor may write me for this.

I would like to request the group to suggest me if my thinking is correct and suggest me a practice they follow in there plants.

I appreciate your help in advance,

Thanks
Robin

My thoughts:

1. Plot variation, the delta from the target, so that when you run different parts you can use the same chart.
2. Train the operators in SPC. Not just how to calculate the mean and control limits. It will be easier to introduce #1.
3. Find a way to simplify or automate the calculations so that the operators just record the data and then enter the results on the plot. There are macros for excel that are inexpensive or free.

BTW, welcome to the Cove, Robin

 

[Steve Prevette]

The model I am used to is a statistician supports the workers on the floor by doing the initial establishment of the average and control limits. The workers plot the dots as they come in, and evaluate the data against the average and control limits using whatever set of detection rules are in effect. If there is a signal, the workers then contact their manager and the statistician.

Through exposure and periodic training, the workers do end up learning quite a bit about setting the average and control limits, but you still need the degreed statistician in support. A good company statistician will walk through the process he/she goes through with the workers when they do respond to a signal.

Another type of control chart to consider is the "Zone" chart. In those charts, the data are plotted simply within the standard deviation band that they fall in, and a cumulative score is kept and once that cumulative score exceeds a certain threshold, that is taken as a signal.

 

[Mike S.]

It sounds like it might be a buy-in and/or discipline problem you are facing, and those are tough nuts to crack. But, FWIW, something that might help if they are super-focused on the spec is pre-control. It is pretty powerful statistically, simple to use, and uses the spec. for its "limits".

 

[D.Scott]

If you are using subgroups of 5, there is a very easy way to calculate the average. It requires a bit of thought on the part of the person setting up the chart but see if you can follow this.

First, if you are measuring "big numbers" where all the beginning numbers are the same (example - the measurements are between 105.1 and 105.9) then you only chart the numbers after the decimal.

Once the number you are measuring is "small", simply add the 5 numbers together, double the total and stick the decimal back in. for example - you have 5 numbers of 2, 3, 5, 3 and 6 total 19 - double it to 38, put the decimal back .38 - 105.38 is your average for the subset which is what you record. It works just as well with bigger numbers too - try it.

Our operators loved amazing their friends without having to use a calculator and it kept our charts neater.

Dave

 

[Ron Rompen]

Another thing you might want to try is precontrol (gasp...he said the P-word!) charts.

We have had this (literally) rammed down our throats from the OEM (we are Tier 5 to them, BTW), and after initial resistance, I have found that it works remarkably well, and is easy for the operators to use. They don't have to calculate ANYTHING, I can do analysis when/if I want, and most importantly, everyone is happy.

If you are, however, using computerized SPC, then you should be able to 'teach' your software to calculate the limits for you at whatever frequency you desire, using whatever sample size you feel is appropriate. This is easily done as a macro in Excel, and most statistical software packages offer it.

 

[Caster]

IMR Charts

We used Individual and Moving Range charts with the control limits pre determined and printed on the form. It was a lot of forms.

The operator just printed a preformatted chart for the part number (set up in excel) and plotted the dots by hand.

We did some basic zone and pattern anlaysis training for operators.

We now run a computerized system, and I feel we have lost "the feel" that our operators had when they had to put pencil to paper and draw that little line.

Some people really have the knack of recalling how there process was doing 2,4,8 hours and even 3 days ago. But they are rare indeed. With a chart, everyone can see how the process is doing.

Our problem was with production people, who did not want to react if it was only out of control. We had nothing but support from operators.

Good luck

 

[Dave Dunn]

The best argument you could give to your production people who only want to react if the parts are out of specification is this:

"Okay, they're out of spec. Now it's too late and we have bad product. If we had controlled the process, we could have kept the parts good."

The other benefits of control charting and monitoring capability include:
- ability to track potential problems in the process caused by abnormal variation (assignable causes)
- ability to assure good product even when not running at nominal dimensionally.