Cpk - Capability Statistics explained

[Music] welcome to complexity made simple my name is Paul Allen and the story of today's newsletter is capability statistics CP and CPK that was much misunderstood capability statistics so let's have a look at this thing now these two statistics help you to answer a very simple question cbcb care that you should be answering right at the beginning of your new product introduction process or indeed your problem solving process and it's this this question you should answer can we can we make it okay now I know for most of you you are answering this question but you are doing it informally you are doing it by observation only where CP and CPK can help you make a much better decision better decisions save you money so let's have a look typically what you're doing well you're going out to your process on you you're collecting some data by the way this is very important that this data is hands-off you are trying to observe the natural pattern of the process and then once you've got that data you're trying to evaluate two things essentially it's a random number generator we work out the average I have a dataset so process up to and of course what we want to know is they're quite one of the questions we've got answer does the average hit the target or get somewhere close enough that we're happy but what we're also gonna do look you've got to evaluate the variability how much spread have we got in this process does the spread fit inside the tolerances and the next thing you've got to ask so let's just put some tolerances on these graphing let's see where we land okay so there's the tolerances look now if you just use observation alone use your common sense and just go for observation what you're basically you know assess now is yeah I'm fitting in so the timers are meeting them meeting the target on a hit all's good with the world can we make it okay yes and away you go no timidly there's two mistakes that are being made here the first thing is this often the sample size so let's say you've taken 15 pieces over the process it's a pure sample size typically and you are saving money here this should be between 30 and 50 if you try and save money here you are saving just 15 pieces and 15 measurements when you want to bow to measure the product for 10 years 15 pieces saved here when you're gonna make a mistake and cause chaos for 10 years is not money saved yeah it's an idiotic decision don't be are fast make sure you take a proper sample size 30 to 50 it's a mistake number one is that you don't take enough data but let's have a look what's going on so let's say the underlying distribution in the process looks like this but some tourists he's on there let's say you're going to get a small defect rate we'll say 2% in the tables here even if you've taken by the way even if you've taken a good sample size so we've taken 15 you're going to sample 15 over there but even if you've taken 30 where do they gonna come from look at the shape of the thing you know these results out here these things are rare probably sitting in the tail and that's not one in a thousand data point to get one sitting over there where's your 15 or 30 to go to comfortable are they going to come from the middle so you take a sample where does your date the come from when it comes from the middle section where most of the results are what do you see by observation or by observation you see something that looks great and of course what do you assume are we gonna make you know okay we're brilliant at this let's turn it on because you turn it on what happens the two-percent kicks in and now instead of measuring an additional fifteen or thirty pieces that you might have measured here now you've got a measure every one because you made a bad decision using observed data now let's take a look at what the CP and the CPK are going to do and how they're going to judge this situation much much better so we take your data but we don't go with observation what we do is we use the data to make a prediction about data points we haven't seen yet how do we do that well we take the data and we calculate the standard deviation we calculate Sigma what is the standard deviation so useful to us because it has the power to predict the range we haven't yet seen how does it relate to this shape well like this any normal distribution its true range its predicted range he's always assumed to be six times a standard deviation and this is always going to be much wider than your observed values it's going to predict results out here that you haven't seen yet so once we have the predicted range instead of the observed range now we can calculate these two statistics so let's show you how these are calculated and it's very simple CP is known as the process potential and all it basically says is does my process potentially fit inside my tolerances so let's draw the distribution let's put some tolerances on the shape he basically says this how many times does the big blue thing fit inside those red lines now in this case the big blue thing fits inside there once that's a CP of one potentially which fit inside the tolerance and it's just width versus width the width of the tolerance versus the width of the process that is CP width by width potentially can we fit let's draw that along just to get a sense of what's going on okay how many times does the big blue thing fit inside the red lines well this time I'm drawn in helpfully so it fits inside there twice we have a potential equal to two okay no he doesn't care about the fact that it's off-center by the way CP is just a width of versus width but of course we need to take into account the fact that it's centered on on it's one of the questions we asked over here are we eating the target and does the variability fit well we asked the question does the variability fit with the CP there's some where we need to take into account did we hit the target or not and that's what the CPK is because what the CPK is CPK is the actual capability okay and he does take into account the center so let's put the center on this thing I dropped the middle of my results down okay now then the CPK just says simply this please look left please look right from the average whichever is the smallest figure and of course in this case we are looking left now because we are only going to look at half the tolerance this time I'm only gonna ask the question how many times does half a process fit inside that space so how many time does the black slice fit inside that space CPK fits in there once so potentially I have a potential of two I have an actual of one okay so that's CP CP k now this is a prediction yeah it's a prediction of what's going to happen it's not coming from the observed data so it's a much deeper understanding of what's going to happen and it's going to predict this 2% defect rate for you even though you haven't seen it all your samples could be defect free this is still going to predict that you've got defects now in all honesty do I got all the trouble to calculate these things these days there is a calculation which I haven't shown you for CP and CPK and there's a reason for that these days I use software and the software draws these pictures for me and to be quite honest what I do is I use my common sense and I look at the pictures so if you looked at this one what have you got to do to make this okay I've got to squeeze the variability in that's gonna take three months to achieve typically it needs a project and a team to achieve it if I look at this thing can I make it okay well I can but I need to Center it Center in it takes could take three minutes could take three days depends whether it's a setting on the machine or the tool maker having to make an adjustment to the tool maybe this is very quick this is hard yards you have to be able to see these pictures but once you can see the picture with the software use some software to do these calculations the software is predicting what's happened so the prediction the prediction of tomorrow for data that you haven't seen yet this is a much more intelligent decision CP and CPK saves you money and saves you an effort for the next ten years why wouldn't you use these things you are going to make much more intelligent decisions if you use CP and CPK four measures of capability if you'd like to know more about any of the concepts covered in this video or any of the other concepts covered in my my other tutorial videos then here's my latest book drink tea and read the paper covers everything you need to know about how to make sure that Six Sigma becomes world-class engineering in your company noise if you'd like to get in touch with me a little bit of help about lean a little bit of help with Six Sigma please contact me on the email below [Music]

Channel: Paul Allen

Views: 46,555

Keywords: Cp, Cpk, Capability, ProfessorCleary, Saravanan Kuppusamy, Advance Innovation Group, Lean sigma, Six Sigma, Pp, PPk, TPS, juran, shainin, shewhart, deming, Khan Academy, paul akers, gemba academy, Narendar Sumukadas, janam sandhu, MIT OpenCourseWare, Doug H, Cary Countryman, Jim Grayson, process capability index, cpk statistics, cpk calculation, cpk 1.67 meaning, cp cpk formula in excel, capability index formula, cpk process, Cpk - Capability Statistics explained by Paul Allen

Id: QH984PnwRDE

Length: 11min 47sec (707 seconds)

Published: Mon Mar 12 2018