ASQ Six Sigma Green Belt – Bonus Lectures
And then in I, which is the improved phase, we improved the process and we did some fine tuning using tools such as Design of Experiment. We have achieved the objective with which we started our Six Sigma project. But then we are still left with one more phase in DMAC, which is C. And in C we want to control. So in control phase we want to control the process so that our new or the revised process stays in the revised state and doesn’t drift back to the old way of doing things.
So this is the purpose of the control phase. In control phase we will be talking about three main topics. One is SPC. Statistical Process Control. Second is Control Plan and third is Lean tools for process control. Let’s start with SPC or statistical process Control. In this lecture in SPC or statistical process control, we will understand the basics related to SPC. Then we will talk about rational subgrouping and then we will look at number of control charts. So depending on the type of variable you want to control, there are number of control charts. So we will go through each of these control charts in that section. In this lecture let’s look at some of the basics related to statistical process control.
Now, what is SPC or statistical process control? SPC helps to monitor and control a process, a process which you have improved. Now, what you want to do is you want to monitor, you want to monitor so that the process stays in the current state and doesn’t drift back to the old way of working. Or let’s say if you have reduced defects, you don’t want your process to start producing defects once again and go back to the old state. So that’s what you do in SPC. You monitor and control a process and you want to monitor and control to make sure that your process operates at its full potential. And when I say full potential in regards to defect, you might want to run your process with minimum defects and that in fact will lead to more and more conforming products and less rejectable products. So that’s what you want to do in SPC.
Now, the product which you are making will have some specification limit. So let’s say if you are making a rod, a rod has to be of 1 meter, but then you have some tolerance on that. That tolerance will be plus -1 Mm. Let’s say. So if anything is between 99 and 101 mm, then you will say that that is a conforming product. So if the item is more than 10 one or less than 99 mm, then that product will be non conforming. So whether a product is good or not good is decided by the fact whether this is within specification or outside the specification limit.
Now, when we talk of SPC or statistical process control, there are two phases in this. And as we go further into these videos we will understand that SPC is done by control charts. We have two phases in SPC. The first one is understanding the process variation. So when you produce something and let’s go back to the same example of making a rod of 100 mm in plus minus one millimeters. So the first thing what you want to do is you want to understand the current performance of your process, what are the limits within which your process is producing? So for that, what you will do is you will make some number of rods.
And based on that, you will decide that, okay, my process is capable of producing rods in this range, in these control limits. And then once you have established those control limits, once you have understood the process, and then what you will do is, going forward, you will use those limits to monitor the process. So, for example, going back to the same example of making a rod of 100 Mm, which has a specification of plus -1. Mm. So rod has to be made within 99 and 101. But then once you start Producing that once you Monitor your process, then you find out that your process is so good that instead of plus -1. Mm your process can produce those rods in plus -0. 5 mm so now you know that whatever rods you are producing, most of the rods will be from 99. 5 to 100. 5. So within 0. 5 range, rather than one millimeters, which is the specification limit.
So now what you will do is you will monitor your process. So anytime you see that now the process is producing more variation, then this will give you an alert that something has changed. And now you need to make some adjustment to the process. So, these are the two phases of SPC. First is understanding the process variation and second is monitoring and controlling. When you are monitoring and controlling your process, then you need to take action when things are going out of control. But important thing is that you don’t want to take action too quickly or too early.
And also you don’t want to take action too late. So if you take action too early, let’s say now your process is producing these rods with 100 plus -0. 5 and then you get one piece which is 100. 6 Mm this is an indication that things might have changed and then you start acting on that. And then you adjust the process, but then you don’t want to do this too early because you know that there is a natural variation, there is a normal distribution curve, there is always a possibility that you might get higher value or lower value once in a while. So if you get one reading which is slightly more or slightly less than the limit, you don’t want to immediately jump onto that and take action and adjust the process because that might have come just because of natural variation.
Because if you take too many action on your machine. You keep on adjusting your machine, then you will be messing with the process. But on the other hand, you don’t want to be too late. Also, you don’t want that your machine keeps on producing things which are out of specification. And you are not taking action thinking that this might be just by chance. So this is where SPC helps you in finding out the trend so that you don’t take action too early or too late. So when we talk of taking action too early or too late, we need to understand two types of variation. One is variation because of common cause and second is variation because of special cause. Once we understand the difference between these two things, that will help us in taking the right action at the right time.
Between the common cause and special cause will help us in understanding when and when not to take action. That will help us in avoiding taking action too early or too late. So let’s see the difference between these two and which is shown here. On the left it’s common Cause, on the right it is Special Cause. To understand the difference between common cause and special special calls, let’s take an example of me traveling from my home to my office. I need to reach my office at 06:30 A. m. . And then for that I need to optimize that at what time I should start from my home. That will depend on the journey time, how much time it takes to travel from my home to my office. Now I keep on recording these numbers. That how many minutes it takes from my home to office. There are a number of things which happen during this travel.
There are red light signals, there is traffic somewhere, there is different road conditions, sometimes it’s snowing, sometimes it’s raining. So all these things happen which affect my travel time. So if I keep on recording the time I take, this might come as, let’s say ten minute, one day. Second day it comes out to be eleven minutes. Next day it comes out to be nine minutes. Next day it comes out to be 10. 5 minutes. And one day it comes out to be 20 minutes. Now, 910, eleven, this is something a common range. And these changes are happening because of some common causes. These common causes could be instead of two red lights, let’s say I get three red lights one day. Sometimes I don’t get a red light and I straight away reach to office in nine minutes. So this is common cause. Common causes are many which have minimum impact on the outcome. So there could be a number of things which could happen, but then there are some special causes. Let’s say one day there was a major roadblock and that day I reached the office in 20 minutes. That was a special cause, that was something special. That was something above and beyond the normal variation. This is the special cause. So when it comes to common cause, you really cannot attribute that. What is the reason that one day I reach 90 minutes, another day I reach eleven minutes? Because there are so many factors, you really cannot attribute that variation to something specific.
On the other hand, when there is a major change, when there is a major roadblock or major weather impact or something, that is a special call because the variation you have on that day could be attributed to one or few important activities or things which have happened. So this is the key difference between the common cause and special cause. There are many common causes which have minimum impact. And then on the right side, if you see in special causes, there are few causes which have the significant impact. Same thing in case of manufacturing the rod which we were making in plus minus one millimeters. There are so many things which are happening which you really cannot control and because of this you get fluctuation in the dimension of the rod which you are manufacturing. This is because of common cause. Someday the machine has a problem that day you get too much of a variation. This is because of a special cause which you can attribute to something special that this particular variation is happening because of this particular reason. So if that variation is happening because of so many factors, you really cannot remove all those factors. It is not economic to remove all those factors. And when it comes to special cause which are few in numbers, it is economical to remove those special causes to bring the process back to normal.
Common causes and special causes are also known by few other names. For example, when it comes to common cause, the common cause is also called as the random cause, chance cause or nona assignable cause because you cannot assign this to a specific thing. On the other hand, special causes are called as the signal, the systematic causes or assignable causes. So as you go further into the topic of control charts you will understand that that control charts will ignore the common causes.
They will not ask you to take action for common causes, they will identify special causes. So if there’s any point which goes above and below three sigma and we will talk about these things, then that would mean that there is something special here. So there is a special cause which is making this particular point go above and beyond three sigma limits. So that is special cause. And once you see a special cause on the control chart then you take action on that particular process. And when things are within plus minus three sigma then you don’t take action because those are because of the common cause. So this was the difference between common and special causes. Let’s look at the lesson plan as we go further into the topic of control charts. Here you see number of control charts listed. We will go through each of these in the next few videos.
But what I want to emphasize here is when you are monitoring a process, you are monitoring that process based on the data which you are collecting. In case of my traveling from home to office, this data was number of minutes it takes from my home to office in case of rod which we were making. The size of the rod in millimeter is something which we are controlling. If we are looking at a book and finding out the number of errors on each page then that is the data which we are measuring. So whatever data you are monitoring, that data could fall into two broad categories. And we have talked about this earlier in statistics as well, but let’s understand that that the data which you are controlling could either be a discrete data or this could be a continuous data. The example of continuous data is something which you are measuring. The time which you are measuring the dimension of the rod which you are measuring, these are continuous data. On the other hand, discrete data is something which is counting. So if you take a book look at number of mistakes on each page you are counting here that will be the discrete data. So at this stage I just want to have the discussion on these two types of data. And depending on the type of data, there are different control charts which we will be learning. And then at the bottom of the slide if you see there is a statement that N is the subgroup size. So there is something called as subgroup control charts.
And there we looked at this flow diagram where we talked about discrete data and continuous data and if you look at the bottom here, it says that N is the subgroup size. What is the subgroup size, what is the rational subgroup size? This is something which we’ll be talking in this video. So when we make a control chart and control charts such as x bar chart and we will talk about X bar R chart later. So what we do in X bar R chart is from the production line. Let’s say we pick five items, five rods which we were producing, take the measurement of those five items and take the average of that and we plot that average on the control chart. So let me quickly draw a control chart here. This is the mean, this is the upper control limit and this is the lower control limit. UCL is the upper control limit, LCL is the lower control limit and this is the mean. So I take five sample and take the average of that and I plot this average here. Then after some time, let’s say after 1 hour I take another five samples from the production line, take the average of that and I plot it here and here and here.
After each 1 hour I do it less like this. And one time let’s say this average goes above this limit of UCL. So this is something because of special cause. Anything which is between lower control limit and upper control limit is because of the common cause of variation. Now, when I say that I pick five items from the production line and take the average of that, those five items are subgroup. Subgroup is a snapshot of the process at that time. So when I pick five items from the production line that basically represents the production at that particular time, then I take another five items as subgroup next hour, that is the snapshot of the next hour.
And when I say I pick five items, why don’t I pick three items in the subgroup? Why don’t I pick 20 items in the subgroup? This is what is the point of discussion here, what is the rational or the reasonable size of the subgroup and what is the basis of deciding that subgroup size? This is what we are talking here. But one important thing which we need to understand here is that subgroup is a snapshot of the process and what does that mean is that when you take items as a part of subgroup, you need to take all those items at that time only. When the production is being done at that time only you pick five items. Don’t pick those five items from the production bin where all the old items, the items which were produced yesterday, day before yesterday, don’t pick those items from that bin, pick these five items from the production line because what you are using these five items as the snapshot of the process at the time.
So these items, these five items in the subgroup must be taken together in time and still should be independent of each other. So anything, whatever you do in statistics is you make sure that the items which you are selecting are random and independent of each other. So after 1 hour you just pick five items randomly from that production. So that was about the subgroup. Now the variation in the process could be of two types, variation within the subgroup and variation between subgroup. Earlier what I was showing you a control chart, this is my mean and this is my upper control limit and this is my lower control limit. When I picked five items and took a mean of that and I put that mean on the control chart. So that mean is here, let’s say.
But this mean is the average of five items. So let me put those actual five items here in red. So those five items would be something like this, 12345. And the mean is shown in the black. Next time I take another sample after 1 hour and these five items are actually here, 12345. And the mean of this is somewhere here. Then after some time I take another five items and I take a mean of that and I plot that mean here. So mean is this one. But then those actual five items, the actual five dimensions let’s say, are these 12345. Now, when we talk of variation, there are two types of variation here and this we have talked in anoa and this we have talked in two sample t test. Also, if you remember, there are two types of variation, variation within subgroup. Variation within subgroup is this variation. This is within subgroup. This is within subgroup. This is within subgroup.
And then there’s a variation between subgroups. Variation between subgroup is the variation between means of these. This is between. So these were two types of variation which you have. So in control charts such as x bar R chart, what you will have here in x bar chart is let’s say this is x bar chart. You are not plotting individual values. What you are plotting is the mean. So whatever variation you see here is because of variation between subgroups. So this is variation between. And when you plot R chart here below this and you will see that when you draw R chart in R chart, what you are doing is you are drawing the variation within subgroup because R is the range. What is the range, this range in the red, this is what you draw here. So in x bar R chart, the X bar chart or the mean chart will show the variation between subgroups and the R chart will show the variation within subgroup. And as you go further into the calculations of upper control limit and lower control limit, you will understand that these upper control limits and lower control limits are driven from the range and this is driven from the variation within the subgroup.
So if you have too much variation within subgroup or the range is too high, your upper control limit and lower control limits also would be wider. So what does this mean is that if you have too much of variation within subgroup then your control limits will be too wide. And if your upper control limit and lower control limits are too wide, then any fluctuation which is happening even because of special cause will not get detected by control charts. So that is the reason you want your subgroup size to be so optimum that you don’t get this subgroup which is leading to too much of variation. What you want to have is a subgroup from a single stable source.
Now, what is the best or the most optimum subgroup size? The most commonly used number here when it comes to subgroup size is five. So you have subgroup size as five, which is the most commonly used number. And if you want to go for a higher subgroup size or lower subgroup size, you need to understand that that when you have a small subgroup size, the meaningful shift might get undetected. And when you have a very large subgroup, your subgroup size of 1020 or something, then even the smallest change in the process shift will lead to false alarm. Now you would ask that why I’m making this statement. To understand that, let’s go back to the Central Limit Theorem and normal distribution. Any process which is having normal distribution will be something like this. And this is my minus three sigma and this is my plus three sigma. This is the thing which we use in control chart. So what we do is, let’s say if I have a subgroup size of one, I just pick one item and I plot that if I pick one item, then there is a 99. 97% chance that this item will be in plus minus three sigma. So I pick one item, this might fall somewhere here, I pick second item that might fall somewhere here, the third item here, here and so on, most of the items in the middle, then some items near the tail. This is what will happen in control chart also. So if I put this in terms of control chart, UCL, LCL, most of my items will be near the center and then there are some items which are near UCL and LCL.
Now, instead of one item in the subgroup, let’s say if I have four items in the subgroup and I draw the mean of that on the control chart. And here you need to remember that what we learned in Central Limit Theorem is that if I take four items from the production line, take the average of that, the average will not follow this particular distribution. The average of these four will follow this distribution which is much narrower because here, sigma x bar will be sigma divided by square root of n. And when I say picking four items, that means sigma divided by square root of four is equal to two.
So here in this case, the sigma will be half. What does this mean is that instead of one item, let’s say if I have four items and I draw a mean of that, the control limits for that will be something like this, half of this. So all my items will be in this range. So if I have bigger subgroup size, that means my control limits will get narrower. And when my control limits gets narrower, then if there is a smallest shift in the mean, that will immediately show that things have gone out of control limits because my control limits have become narrower. Now, when I have a larger.
Popular posts
Recent Posts