Hi there,

there is a process, operator take data every hour(two shift a day, total 16 data a day) and put data in computer to draw SOC chart,

qustion1: as a quality engineer, how long do you check your SPC chart? I means one hour, half day or other? Because quality engineer also have other duty.

question2: when you find out of control, what's response do you do in this two case?
a) if process is still running(stop machine, or ???)
b) if porcess is finished now(check product in stock? or??

thanks

Michael

Hi there,

there is a process, operator take data every hour(two shift a day, total 16 data a day) and put data in computer to draw SOC chart,

qustion1: as a quality engineer, how long do you check your SPC chart? I means one hour, half day or other? Because quality engineer also have other duty.Michael

There is no general rule regarding how often a quality engineer should check control charts. It depends on a number of factors, including how concerned you might be that something is going to change. Although mine is a minority opinion, I don't believe in running control charts in perpetuity. I think they should be used to provide objective evidence that a process, when controlled and operated as designed, will be efficacious. Once that has been established, there is (generally) no good reason to do SPC forever.

question2: when you find out of control, what's response do you do in this two case?
a) if process is still running(stop machine, or ???)
b) if porcess is finished now(check product in stock? or??

In general, signs of process instability are not, in and of themselves, reason to either stop the process or start containment activities. Remember--a point beyond the +/- 3 sigma limits might be normal--only 99.73% of the population is included within the limits, and given enough time, that other .27% will appear.

The nature of the out-of-control condition should also be considered; a point outside the control limits probably represents a less alarming condition than say, seven or eight consectutive points on one side of the mean. The point is that your reaction should be rational and based on both an understanding of your process and an understanding of probability and variation.

qustion1: as a quality engineer, how long do you check your SPC chart? I means one hour, half day or other? Because quality engineer also have other duty.

If the workers have been trained and are competent in updating the chart, and know how to determine if a trend (out of control condition) has occurred, and know to contact management and the quality engineer when such a condition occurs, then there really is no reason for the quality engineer to check it. Personally, I would make sure to tour the work spaces on a periodic basis anyway, and as part of the tour be sure to check the charts. Hopefully they should be posted for all to see.

question2: when you find out of control, what's response do you do in this two case?
a) if process is still running(stop machine, or ???)
b) if porcess is finished now(check product in stock? or??

It depends. The control chart has sent you a signal that something has changed. What new risk are you encurring? If the SPC chart also includes the specifications (USL and LSL) how close are you to them? Or have you already crossed them? If the change makes the product unusable or hazardous, STOP the machine. It the change is a trend towards a problem (hopefully the case) you have time to do some investigation and perhaps finish the current batch.

By the way, I disagree with the response that once a process is in control you should stop doing SPC. Once the process is in control, a major use of SPC is checking (very inexpensively I may add) that it still is in control.

The control chart has sent you a signal that something has changed.

Not necessarily--points outside the control limits--or other signs of statistical instability--might be "normal." As I asserted in my earlier post, given enough time, the improbable will happen.

By the way, I disagree with the response that once a process is in control you should stop doing SPC. Once the process is in control, a major use of SPC is checking (very inexpensively I may add) that it still is in control.
Here's the way I see it, and keep in mind that I'm not saying that ongoing SPC is unnecessary in every instance: We want people to concentrate on controlling the process. When a process is designed, parameters for its operation are established. The working hypothesis is, "If I establish parameters a, b, c... and those parameters are monitored and controlled, the process output will be in conformance." Using statistics and the scientific method, we should try to falsify the hypothesis. In other words, we should run the process, monitoring both the process controls and the output until we have credible objective evidence that the process controls don't work. If enough data has been collected and analyzed and the hypothesis has not been disproven, there is no longer a need to continue statistical analysis.
SPC, in general, should be considered a tool in the objective establishment of process controls, not an ongoing method of verifying process output.

Not necessarily--points outside the control limits--or other signs of statistical instability--might be "normal." As I asserted in my earlier post, given enough time, the improbable will happen.

True. There can and will be false alarms. But I like to use the analogy of a fire alarm going off in a building. Smart people evacuate the building as an immediate action, then check to see if the alarm was valid or false.


SPC, in general, should be considered a tool in the objective establishment of process controls, not an ongoing method of verifying process output.

I guess we'll just have to disagree on that point.

HI JSW05 and Steve,

Thank you so much for your kind opinions. I do appreciate that. :thanx:

You asked this question also on the ASQ discussion board, where I replied yesterday. Please see that thread also. Basically, I agree with Steve and JSW (that is, where they agree with each other :rolleyes: )

Another thing you need to consider is just where you are in your SPC journey. The others alluded to it, but I want to make certain that you understand it: The frequency that someone needs to "really" evaluate the chart depends on if you are just beginning your SPC program or if your operators very familiar with SPC.

You also need to use those charts for something, so don't just think, OK I have a chart for this shift. Use them for capability studies. Make use of the data. The charts by themselves while, helpful, can be misleading. You can get yourself down to a very small control limit area, yet really be setting yourself up for defects if the data is skewed to one or the other spec limits.

If the workers have been trained and are competent in updating the chart, and know how to determine if a trend (out of control condition) has occurred, and know to contact management and the quality engineer when such a condition occurs, then there really is no reason for the quality engineer to check it. Personally, I would make sure to tour the work spaces on a periodic basis anyway, and as part of the tour be sure to check the charts. Hopefully they should be posted for all to see.

It depends. The control chart has sent you a signal that something has changed. What new risk are you encurring? If the SPC chart also includes the specifications (USL and LSL) how close are you to them? Or have you already crossed them? If the change makes the product unusable or hazardous, STOP the machine. It the change is a trend towards a problem (hopefully the case) you have time to do some investigation and perhaps finish the current batch.

By the way, I disagree with the response that once a process is in control you should stop doing SPC. Once the process is in control, a major use of SPC is checking (very inexpensively I may add) that it still is in control.

:agree1: Thanks Steve, I believe this is the only way charts make sense. If we wait until the QAE gets around to evaluating the charts the damage may already be done. The first line of defense is the line operator who record on the chart. Training, training and more training.

Once that has been established, there is (generally) no good reason to do SPC forever....

In general, signs of process instability are not, in and of themselves, reason to either stop the process or start containment activities. Remember--a point beyond the +/- 3 sigma limits might be normal--only 99.73% of the population is included within the limits, and given enough time, that other .27% will appear.

The nature of the out-of-control condition should also be considered; a point outside the control limits probably represents a less alarming condition than say, seven or eight consectutive points on one side of the mean. The point is that your reaction should be rational and based on both an understanding of your process and an understanding of probability and variation.

The decision to control chart a process should be dependent on how critical it is to quality and not whether it is in control at any given instance or for a long period of time. Using your example, the out of control condition may occur as a result of a change in the process. Without the control chart, it is not as likely to identify the seven or eight consecutive points or even the points outside the control limits. If your "rational" reaction is based on "an understanding of your process and an understanding of probability and variation", SPC is a valid statistical tool for continuing to monitor the process.

Bill Pflanz

there is a process, operator take data every hour(two shift a day, total 16 data a day) and put data in computer to draw SOC chart,

qustion1: as a quality engineer, how long do you check your SPC chart? I means one hour, half day or other? Because quality engineer also have other duty.

question2: when you find out of control, what's response do you do in this two case?
a) if process is still running(stop machine, or ???)
b) if porcess is finished now(check product in stock? or??

Hi Michael

Do your operators collect the data points and enter them into computer SPC software for your Quality Engineer to review at a later date?

It seems from your second question that you are finding out of control (OOC) conditions much after the fact.

I hope this is not the case. If so you are not getting all the value possible out of SPC.

This is a downside of some software, in that you need an extra step to generate the chart (which people may not take the time to do).

The old school method of plotting charts on paper by hand was very powerful, the operator knew right away that there was an OOC condition.

As Steve has said, the operator must react to the OOC as it happens.

As a Quality Engineer here, I have NO responsibility for SPC charts....they belong to manufacturing and the operator. My role is to train, help them improve the process, do studies, look for rare occurences, zone and pattern anaysis, etc. But hour to hour, the charts are theirs.

Can you provide more background?