iSixSigma

Which Control Chart Should I Use for Output Per Hour?

Six Sigma – iSixSigma Forums General Forums Tools & Templates Which Control Chart Should I Use for Output Per Hour?

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #238175

    Erik 2018
    Participant

    Hello everyone,

    Two short questions.

    If I want to create a control chart where I measure the output each hour, should I just use an I-MR chart as my subgroup is an hour? However, output can range from 0 to infinity it is still discrete measure. So i’m not sure.

    The other thing is that my output per hour is not normally distributed. However, if I change it to “average cycle time per hour” it is. E.g. 60 output per hour is an average cycle time of 3 minutes. In that case I have to use an Xbar-chart isn’t it? I’m a little confused that just dividing our data lead to differences regarding normality and the use of another type of control chart.

    Hope to hear your opinion about this:).

    0
    #238179

    Robert Butler
    Participant

    1. Control charts do not require normal data – see page 65 Understanding Statistical Process Control 2nd Edition – Wheeler and Chambers
    2. When you have the kind of count data you have it can be treated as continuous data – see page 3 Categorical Data Analysis 2nd Edition – Agresti
    3. I-MR would be the chart of choice – you should check to make sure your data is not autocorrelated. If it is you will have to take this into account when building the control limits – Wheeler and Chambers have the details.

    1
    #238205

    Chris Seider
    Participant

    I would advise a run chart–you’ll probably get “out of control” indications most of the time. There are other things to track also besides output per hour but I’m sure you’re aware of this point.

    1
    #238208

    Mike Carnell
    Participant

    @erik2018 @Cseider @rbutler I think both Robert and Chris gave great advice which they always do.

    My question is why do you want to chart output? Output is a dependent variable. The hour is over and now you look at your chart and find out you are out of control. Now what? You still don’t know anything so you need to investigate meanwhile what ever made you run out of control is continuing to make you run out of control. I would think at the very least you would want a chart (I agree with Chris it should be a run chart not a control chart) on the bottleneck and I am not sure I would be plotting by the hour if you are running any volume. (Think about your Lean basics and takt time – output in relation to time)

    There used to be this thing in SS Y = f(X) where Y is the dependent variable and X is the independent variables. I am not trying to sell you Six Sigma but you can waste your time playing with a dependent variable all day long and nothing will change until you do some thing about the X’s. You will get those people that are going to push back and tell you how there are hundreds of variables that can effect it. That is a resistance to change tactic. They through out some outlier that occurs every time Halley’s Comet shows up.

    The most I have ever seen, leverage X’s, is 5 and that has only a been few times. Probably 3 is most common. Put in 3 run charts on independent variables so when it goes south you know where the issue is or one chart on the output and spend another hour figuring it out.

    You probably have some old time quality person or a MBB that thinks a Control Chart is the answer to everything. They aren’t. With the technology we have today the concept of “control” is more difficult to understand in terms of what to control than how to control it.

    Just my opinion.

    1
    #238350

    Erik 2018
    Participant

    Thanks for all your replies. At the moment we use a control chart just for analysis of a few weeks of production. Based on an histogram and a control chart we could determine out of control situations, variety within the control limits and track both back to certain X-s. You’re right that using output just for control without analysis is useless, even more when the desired output changes each hour. However, in this case it is the output on the bottleneck, which needs to produce a certain amount of products each hour. An hour lost on the bottleneck is an hour lost for the company (that’s what Goldratt learned me). In that case, I was interested in which control chart to use.

    Speaking of tracking X-s, what do you mean exactly if I may ask. E.g. we can regard cycle time as the dependent variable (amount of minutes per hour/desired output per hour), which is maybe a more useful performance indicator than output. What I was wondering, what do you regard as tracking X-s? Or do you mean tracking a subdivision of the Y. E.g. I can imagine that we track downtime as it could be an important part of the cycle time of a product. However, this still isn’t an X, but part of the Y (the same as waiting time is not an X but part of throughput time (throughput time = waiting time + production time + …). What would you track in those cases e.g. in a run/control chart?

    0
    #238357

    Mike Carnell
    Participant

    @erik2018 The difference between a dependent variable and an independent variable is that the dependent variable is simply a function of other things. Output is dependent on the bottleneck because output can never be greater than the bottleneck. Output doesn’t control the bottleneck the bottleneck controls the output. There will be other variable in a model of the line that have more influence than others. I would look at the top 3 constraints. If Operation A is my bottleneck and it runs at 10 per hour then you set up your line to run at 10 per hour and hope that actually meets the takt time requirement. So you are saying the volume varies and you work on the bottleneck so it can produce 15 but Operation B is maxed out at 12 then when you move to 15 your bottleneck shifts.

    There are two issues. First as long as your volume varies then the relationship of process cycle times across the entire line is important. The other issue is why is the volume shifting and can you level load . If you can do that then the control issue becomes much simpler.

    Now that we know your control chart is not being used for control Chris Seiders (@cseider) comment about a run chart makes even more sense. Plot two lines on a run chart – the target and the output (if you want to stay with output). It might make more sense to plot target output and bottleneck but that is going to have to take into consideration the time lag between output and bottleneck output. At some level it is a lagging factor. Understanding that could be extremely interesting as well.

    The changing volumes are driven by whom? Are you responding to a customer?

    0
    #238461

    MagicMaria
    Participant

    I find a lot of value in using statistical process control for many reasons, including, but not solely, for detecting when a process is not in control. The type of chart you use depends on a number of items, many of which you have considered. The one question that I have not observed addressed so far is what does your data look like? If all you have is the average for the hour the I-MR chart becomes your best choice by default. However, it sounds like you have the detailed data with the full variability available to use. In this case an Xbar-R (8 or less observations per subgroup) or Xbar-S (greater than 8 observations per subgroup) can be more powerful as it will take into account the full variation in the data when calculating limits.

    0
    #238462

    Chris Seider
    Participant

    @magicmaria I’d advise against control charts for process output off a line. As the surgeon once said, not everything can be fixed by duct tape. Paretos and other tools work better for this problem–my two cents.

    Just an amusing note to ya :)

    0
    #238463

    MagicMaria
    Participant

    @cseider All of those tools can be useful to identify where opportunity for improvement lies. None of these tools can fix the problem – they only point you in the right direction. Some of them may show the same opportunity in different visual format- in fact they should. A surgeon checks the basic functioning on the body before using other tools, but of course the other tools are important also. I probably wouldn’t use the control chart with the front line teams in this situation, but I like to see what it tells me. :)

    0
Viewing 9 posts - 1 through 9 (of 9 total)

You must be logged in to reply to this topic.