Good Day!
I need help:
1) Can I use a P chart to monitor our operator performance?
Means the NC will be whenever the operator failed to follow OP/SOP/etc...
2) Let's say I have a process that always fails to meet customer spec. Can I use SPC to see the trend of my process? Or is there any other method for me to check on our stability of process?

Thanks for your help.

Sitiz, I really am not sure if I got your point in item (1). Can you elaborate further regarding this type of NC mode and why does this occur. I am sure the people here at cove are eager to help you design a monitoring system for this mode.
Regarding (2), I think what you should do for that process is to study its capability, and why does it always fail with customer specs (what process has a high variation?you start on reducing variation on this process, or study the tolerance limit, etc.). Again, can you elaborate more on this process so people here can guide you once step at a time....
wel be waiting...
cheers!

Sitiz, I really am not sure if I got your point in item (1). Can you elaborate further regarding this type of NC mode and why does this occur. I am sure the people here at cove are eager to help you design a monitoring system for this mode.
Regarding (2), I think what you should do for that process is to study its capability, and why does it always fail with customer specs (what process has a high variation?you start on reducing variation on this process, or study the tolerance limit, etc.). Again, can you elaborate more on this process so people here can guide you once step at a time....
wel be waiting...
cheers!

Hi Sativ,
1) We want to monitor the operator performance;eg;they did not adhere to Work Instructions/Visual Aid at the assembly station.Due to that;we have faulty/defective parts during our final inspection.
So my Q is wether I am able to use P-chart to monitor the operator performance.
2) For item 2;my understanding is we cant perform Cp/Cpk studies when our process is not in control.This is the situation that I am having. Our process is far beyond the spec limits given by the cust.Thus,how do I need put up a control in this type of process?
* Q throw out by customer; how do we ensure that our process are in control?:confused:

If you keep track of how many times an operator does a certain operation, and you keep track of any errors made during the operation, you could certainly do a p-chart for percentage of time an error was made. With p-chart control charts you could monitor it versus time to see if it is stable and predictable. You can even compare across operators over a fixed period of time to see if all operators are statistically the same or if there are outliers (good or bad) that you ought to learn what is going on with.

You can also take the failure type data and do Pareto charts of the most common failures.

Now, if you do not know how many times the operator does the operation, you could shift to ucharts of numbers of errors per 100 units produced, which is different from a p-chart since there may be more than one error per unit.

Good Day!
I need help:
1) Can I use a P chart to monitor our operator performance?
Means the NC will be whenever the operator failed to follow OP/SOP/etc...
2) Let's say I have a process that always fails to meet customer spec. Can I use SPC to see the trend of my process? Or is there any other method for me to check on our stability of process?

Thanks for your help.

If you have a process that always fails to meet customer specification you have no need for SPC to check on the stability of the Process.

I recommend that you perform a tool capability study along with a process capability study and reduce the variables that are causing the non conformance to customer specifications. Once you have a capable process, you can then use SPC to monitor the process.

Good luck,

Hi Sativ,
1) We want to monitor the operator performance;eg;they did not adhere to Work Instructions/Visual Aid at the assembly station.Due to that;we have faulty/defective parts during our final inspection.
So my Q is wether I am able to use P-chart to monitor the operator performance.
2) For item 2;my understanding is we cant perform Cp/Cpk studies when our process is not in control.This is the situation that I am having. Our process is far beyond the spec limits given by the cust.Thus,how do I need put up a control in this type of process?
* Q throw out by customer; how do we ensure that our process are in control?:confused:

As Steve said, of course you can use a p chart or u chart to monitor operator error rates. HOWEVER, be srue that you are monitoring operator error rates and not failure rates unless you know for absolute certain that the inspection failures are actually caused by operator error. Too often we blame operators for something that is completely beyond their control, if this is the case your monitor will only makes things worse.

As for the second I too agree that to simply monitor the stability of a non capable process is a waste of time. I would focus on improving the process first. In fact I would go back to the customer and clarify their statement: many people mistakenly confuse the terms "out of control" and incapable. Those who are not steeped in SPC will often refer to an incapable process as "out of control"...What the Customer typically wants is a stable (in statistical control) and capable process...

As Steve said, of course you can use a p chart or u chart to monitor operator error rates. HOWEVER, be srue that you are monitoring operator error rates and not failure rates unless you know for absolute certain that the inspection failures are actually caused by operator error. Too often we blame operators for something that is completely beyond their control, if this is the case your monitor will only makes things worse.



Good post by Bev. I would say that go ahead and monitor the failure / error rates. BUT as pointed out by Bev - do not necessarily assume the error was due to something the operator did. It is worth doing a causal analysis on the most common errors and deteremine what was the source of the error, and more importantly, what can be done to eliminate the error (assuming the error does cost the company money).

Sitiz,
This is what I have done in my previous "process improvement" days:
1. list the variables that directly affect the condition of the product after processing
(example: to reduce the variation in the final height of a certain electronic product with solder, the following variables are taken into consideration: solder amount, height of product (from previous processes) before soldering, etc.)
2. conduct capability study on these variables and determine which of these variables has a high variation rate. (in the example above, it was found out that solder amount placed on product has high variation)
3. conduct improvement regarding that variable (example, look for a way to stabilize the amount of solder placed in the product).
4. if variation is reduced (capable process), THEN conduct SPC for this variable (by controlling the amount of solder, the solder height variation was controlled!)

Hi! Thank you everyone for that advice.

I will start to monitor the performance with the P chart. Yea; I've told them the same thing about having to use SPC in an incapable process. But you know when customer questions this, some smart guy just throws out this idea to the customer: We will have SPC! :nope: Just say things for the sake of answering questions. Thanks again everybody!

Hi! Thank you everyone for that advice.

I will start to monitor the performance with the P chart. Yea; I've told them the same thing about having to use SPC in an incapable process. But you know when customer questions this, some smart guy just throws out this idea to the customer: We will have SPC! :nope: Just say things for the sake of answering questions. Thanks again everybody!

I do believe that you may be able to identify the sources of why your process is incapable if you do look at the data through the P chart and Pareto charts of your sorts of errors.

From God's brain thru Steve's fingers!

It's amazing the things you can see when you know how to look. If I were your customer I probably would have insisted on at least a run chart of the error rate with operators, lots, days etc. listed AND the aforementioned pareto chart of failures and immediate cause findings. This is exactly where to start to improve your process usign logical, science and reason.

I agree with Steve, and I will go further.

Using a chart for assembly errors is useless for assigning nonconformances if the errors are unacceptable, unless the chart is used to identify who, when, why and how often the errors are occuring so the causes can be investigated.

In such a case the nonconformance goes to the process, not the person, unless it can be shown the person is performing poorly intentionally. Even though a customer likes to see statistical techniques (this type of charting is not process control charting) being applied, we should avoid making charts that are not clearly helping our process somehow.
:2cents:

2) Let's say I have a process that always fails to meet customer spec. Can I use SPC to see the trend of my process? Or is there any other method for me to check on our stability of process?

SPC might be the tool you use to get the process in control and capable. Lack of capability does not always preclude SPC - although it can goof up control limits in some cases.

Hi!

I have another question with regards to this P-chart. Can we use P-chart to monitor our First Pass Yield (FPY)? I know that P-chart is used to collect the defective units but what about the pass units? Can I use P chart in this case?

Thanks

Hi!

I have another question with regards to this P-chart. Can we use P-chart to monitor our First Pass Yield (FPY)? I know that P-chart is used to collect the defective units but what about the pass units? Can I use P chart in this case?

Thanks

Yes. If you work through the math of the control limits, the chart of "p" versus "1-p" will be exactly the same, but flipped upside down. So if you want to plot the percent acceptable, it will give exactly the same signals as if you plot the percent defective.

Hi to all,

I have been using P-chart to monitor defects and to determine if the fluctuations of the defects in in control or not. So far no problem. Now, I wanted to monitor the ratio of Number of pieces produced per kilowat hour (kpcs/kW) to control the usage of electricity.

My data looks something like this (on the average)
Product produced - 15,000 kpcs
Electricity - 270,000 kW

I have collected 2 years data and inputted on my spreadsheet. The resulting UCL and LCL was too narrow. I tried to adjust my units to Mpcs/MWatts and the UCL and LCL were fine.

I would like to ask if there is a particular rule in using a p Chart for this kind of data types. I am using the right chart or is there more appropriate control chart for the abovementioned type of data.

Thank in advance for your advice/s.

Regards,

Romel:thanx:

Now, I wanted to monitor the ratio of Number of pieces produced per kilowat hour (kpcs/kW) to control the usage of electricity.

My data looks something like this (on the average)
Product produced - 15,000 kpcs
Electricity - 270,000 kW

I have collected 2 years data and inputted on my spreadsheet. The resulting UCL and LCL was too narrow. I tried to adjust my units to Mpcs/MWatts and the UCL and LCL were fine.

I would like to ask if there is a particular rule in using a p Chart for this kind of data types. I am using the right chart or is there more appropriate control chart for the above mentioned type of data.


Since this is variable data, I would suspect that using a variable chart (x-bar -R or so) would be more appropriate. If you can post the data, we can evaluate the distribution, etc. - and provide some more comments and suggestions.

Interesting!

The number of defects per some area of opportunity would be a u-chart.

Thanks bobdoering and also to Steve.

Here is my raw data.
KwkPcskPcs/kW 366,800 18,992 0.05178 337,400 21,048 0.06238 347,900 19,492 0.05603 301,700 21,694 0.07191 345,100 18,234 0.05284 308,700 16,284 0.05275 275,100 14,687 0.05339 269,500 15,433 0.05726 284,200 16,595 0.05839 279,300 15,220 0.05449 262,500 16,791 0.06396 259,000 15,550 0.06004 198,100 13,071 0.06598 193,200 12,228 0.06329 203,700 9,276 0.04554 100,800 4,848 0.04810 142,800 4,726 0.03310 155,400 10,894 0.07010

I want to monitor the kPcs/kW index so I can check if the fluctuation is within control or not and also serve as a good benchmark for efficiency improvement. With the ongoing worldwide financial crisis, orders are becoming unpredictable that is why it is necessary to strictly monitor the usage of resources using this type of index.

Your advise is highly appreciated.

The above data is difficult to read. I just copy and paste it from excel. I encoded it below for your reference. I wanted to monitor the kPcs/kW index.

Kw kPcs
336,800 18,992
337,400 21,048
347,900 19,492
301,700 21,694
345,100 18,234
308,700 16,284
275,100 14,687
269,500 15,433
284,200 16,959
279,300 15,220
262,500 16,791
259,000 15,550
198,100 13,071
193,200 12,228
203,700 9,276
100,800 4,848
142,800 4,726
155,400 10,894

thanks

I took the data and generated a run chart of power per piece (or kW/kpcs which equals W/pc). Also from that data I generated a distribution and outlier chart. If the data is sequential, there appears to be a recent event that should be investigated. If this event is a special cause, then we can look at the earlier data to develop a distribution of typical power usage and also develop the control limits. Making that assumption, I took the first 14 data points and generated an XMR chart and attached it for review.

To continue (with more attachments):

If you look at the data with the outliers removes, you get a beta distribution (or, very close to a uniform distribution). When reviewing the run chart, you can see a downward trend, with an abrupt jump upwards, followed by another downward trend. This is not normal distribution behavior, and you may want to further investigate the cause of that condition.

Downward is good (less power consumption per part)!

Thanks bobdoering for the information. Honestly this is the first time that I will be using xmr chart. I will be googling for a quick tutorial so I can understand your analysis. I am certain that this type charting is the one I am looking for because I only have one data/observation per month or any time period.

I would like to ask if xmr chart can be used to control defective products per day as well.

Thank you very much for the knowledge you imparted.

Thanks bobdoering for the information. Honestly this is the first time that I will be using xmr chart. I will be googling for a quick tutorial so I can understand your analysis. I am certain that this type charting is the one I am looking for because I only have one data/observation per month or any time period.

I would like to ask if xmr chart can be used to control defective products per day as well.

Thank you very much for the knowledge you imparted.

Don't be too quick to apply the X, MR (or more commonly known as the I, MR chart; I = Individual) to everything. There are many chart types becasue there are many situations. It is important to seelct the correct chart for the application. In general, one would start with the p chart or c chart for defectives/defects as this is categorical or counting data.

Don't be too quick to apply ANY control chart.

I looked at the data and plotted the kWh vs kPcs. The relationship is fairly linear, but the "y-intercept" is around 48,500 kWh. This would imply that you are using 48,500 kWh each period for "overhead" - for keeping the lights on and the office computers running, etc. So at some level, it would be worth subtracting this background to get a more accurate estimate of the true "parts per kWh" for actual production. This leaves an average of right around 14.0 kWh/kPcs, rather than the original 17.33 kWh/kPcs

Look once -- the months with the highest energy per part are the months with the smallest production. This is not because the process itself was less efficient, but because you were making fewer parts but you still had to pay the basic costs to keep the lights on. Or put another way -- if you want to be efficient, make lots of parts!

In fact, if you make a run chart of the original data, there is an upward trend in energy use per part. If you make a run chart of the "corrected" data, there is a downward trend!

Tim F

P.S. I just noticed the axis on the run chart is incorrectly labeled. It should be "Months" along the bottom of the run chart, not "kPcs".

Excellent analysis Tim as always! you beat me to it - but I was going to ask if the energy usage included other uses besides simply making parts...

it is always helpful to get the whole story from the poster before attempting to answer their questions in specific...

Don't be too quick to apply ANY control chart.

I looked at the data and plotted the kWh vs kPcs. The relationship is fairly linear, but the "y-intercept" is around 48,500 kWh. This would imply that you are using 48,500 kWh each period for "overhead" - for keeping the lights on and the office computers running, etc. So at some level, it would be worth subtracting this background to get a more accurate estimate of the true "parts per kWh" for actual production. This leaves an average of right around 14.0 kWh/kPcs, rather than the original 17.33 kWh/kPcs


That is a good analysis, but it assumes that this is overall usage and not measured usage on the process itself. It would be a good point to be clarified by the OP as to which it really is. In the end, the control chart will initally provide questions - not answers, but that is OK. Where is energy used (so that if a significant shift is seen it can be corrected)? There are conditions that can eat up energy - such as bad bearings, bad heaters, etc. Such a chart can indicate when these conditions exist.

Good points, Bob

I was simply "following the numbers" and trying to hypothesize about the possible meaning. There might me an "overhead" as I suggested. Or the equipment might be running worse - a low production month might be due to equipment going bad and using more energy, as you hypothesized. Or it might be something else altogether.

I really liked your statement "Control charts will initially provide questions - not answers."

Numbers have stories to tell if we will just listen to them. The numbers help us know where to start looking.

In the end, the control chart will initally provide questions - not answers, but that is OK.

That is exsctly why I recommend plotting the usually in multi-vari type format (time sequence, showing the mulitple components of variation) and THEN asking the second, third, etc. level questions. The resulting drill down will eventually enable us to pick the correct control chart scheme and to know which data is best for the baseline to calculate the stable control limits....

Good points, Bob

I was simply "following the numbers" and trying to hypothesize about the possible meaning. There might me an "overhead" as I suggested. Or the equipment might be running worse - a low production month might be due to equipment going bad and using more energy, as you hypothesized. Or it might be something else altogether.

Good thoughts. Many machines have an ongoing "overhead" power usage, such as hydraulic pumps, motors and heaters. That would always indicate that more parts are better to cover that 'cost'. But, too many parts is bad (inventory costs), too - and there are startup costs for running fewer parts more frequently. So, by itself power consumption may not be the only decision factor - unless the cost of power is far and away greater than these other costs.

Always something making a simple job complex.

Thank you giving me insights. We are on the process of establishing a baseline and your analysis on the given data is very much appreciated. Tim pointed out that the months with the highest energy per part are the months with the smallest production. This is because we still use airconditioning and lighting whether the production is High or Low. It is a good point that I should break down the energy consumption into indirect and direct or something similar.

Another factor is the size of the product. If I produce big magnets, the kPcs/kW index will go down. Maybe I should look into total surface area or total product weight instead of the kPcs.

Maybe I should look into total surface area or total product weight instead of the kPcs.
Another approach when product mix issues complicate the picture is to try to plot deviation from expected power usage. It might be more normalized for an apples to apples comparison over time.

Can you actually monitor the process itself, either with power or amperage gages?