The changeover wheel is a visualization of a good changeover sequence. In this series of posts I will go deeper on how to use such a changeover sequence in planning your production sequence. The concept itself is simple, but there are still some pitfalls in using it. This first post looks deeper at generating a first sequence. My next post will then optimize the sequence, where we will also learn why you probably should not shoot for the perfect solution, but merely for good enough.
I have written about changeover sequencing and its prioritization before. In this post I will go into more detail on the changeover wheel, an illustration of the changeover sequence. In my previous posts I used the example of ice cream (since I like ice cream), hence let’s continue with the ice cream example, but this time let’s make it a bit more complex. Let’s assume you have a machine that can make different types of ice cream. There are three different variables that can be influenced:
- The waffle: Is it a cone, a cup, or a sandwich?
- The flavor: Vanilla, raspberry, or chocolate
- The size: Small, medium, or large
This gives us, for this example, a total of 3x3x3=27 different products, assuming all combinations are actually produced. In industry, you often find similar systems with multiple variables that influence your product. The product could even have more variables. It could require two different tools to change, using different diameters of wire, or using different glues or different components. I’ll stick with the ice cream example here since it provides a very nice visual.
How to Create the Sequence
Now you have to think about the sequence. You should include all products that you may produce on this system. The goal is to have a sequence with the least total changeover time that includes all products. Usually, we measure the time since it is easiest, but what we really want is the cost. In some cases, looking at cost may lead to different sequences than if you look only at time. For example, changing the flavor takes less time, but it does waste some ice cream that is washed down the drain. Is this wasted ice cream more relevant than the longer time for the waffle changeover? It could be. In our example, it is not (merely to keep the example simple). But your changeover may not be as simple as this ice cream example here.
Now you could take all products and arrange them in a sequence. However, this is cumbersome, especially if you have many different product variants. I find it easier to go by variables. Take the most cumbersome type of changeovers, and arrange them in a sequence. In our example, the most cumbersome change is the change of the waffle type. You would have to remove a mechanical tool from the machine, add the new tool into the machine, adjust all the settings, bring the tool to the right operating temperature, and you are ready to go. Hence, we first sort our product groups by the waffle type.
The sequence of the waffles in our example does not really matter. Changing from a cone to a cup and vice versa is very similar. Changing from a cone or cup to a sandwich or vice versa requires more time. But since we have only three waffles, the sandwich is always preceded or followed by a cone or cup, and hence the sequence does not really matter. If we would have two different sizes of sandwiches, however, we should keep the sandwiches together. Again, the goal is to have a (in our case, waffle) changeover sequence with the least total changeover time.
The second most complex variable in our example is the flavor. Changing the flavor requires exchanging the tanks with the ice cream and also requires cleaning of the machine. Here the sequence matters quite a bit. In many changeovers involving color, the change is from light to dark. A speck of vanilla ice cream won’t be noticeable in chocolate ice cream, but a speck of chocolate ice cream will stick out quite a bit in vanilla ice cream. Hence, going from light to dark often requires less cleaning effort. Therefore, we would like to always change the flavors from vanilla to raspberry to chocolate. After chocolate we have vanilla again, although this changeover requires more cleaning effort.
Finally, there is the size of the ice cream. In our example, this is the easiest variable to adjust. We merely change a setting in the program on how much we have to extrude. The changeover time is not quite but close to zero, and it is our least important variable. It also does not matter if we change from small to medium to large or any other sequence, as it will make no difference whatsoever.
The entire changeover wheel could look like something below. It starts with a small vanilla cup (at the top), followed by medium and large vanilla cups before changing to raspberry, etc.
Overall, we sort our variables by their changeover duration, and start with the biggest factor first (in our example, the waffles). Within this biggest factor (waffles), we then sequence the next biggest factor (in our case, flavors). We go through all variables, until we get to the variable that have actually a changeover duration of zero (in our example, the size is close, albeit not quite zero). Within the sequencing for each variable, there are also different situations. The sequence could matter (like the flavors in our example from light to dark), or does not matter (like the size in our example). This will give you a changeover sequence (or a changeover wheel), but – attention! – this is only the first step. In my next post I will show you how to optimize the sequence. Even if the changeover wheel above looks very neat, there is still a lot more potential for improvement! Now, go out, create a changeover sequence, and organize your industry!
PS: This blog post was inspired by a master thesis by my student Milena Oberle: “Entwicklung eines Steuerungskonzepts zur Stabilisierung volatiler Auftragsfolgen einer variantenreichen Mischfertigung unter Berücksichtigung der zukunftsweisenden Erfolgsfaktoren einer Smart Factory“, Hochschule Karlsruhe, 2022.
Overview of Blog Post Series
- A Few More Turns on the Changeover Wheel – Part 1: Creating a Sequence
- A Few More Turns on the Changeover Wheel – Part 2: Improving the Sequence
- A Few More Turns on the Changeover Wheel – Part 3: Prioritizing Jobs
- A Few More Turns on the Changeover Wheel – Part 4: Options for Prioritizing
- A Few More Turns on the Changeover Wheel – Part 5: Frequently Asked Questions
5 thoughts on “A Few More Turns on the Changeover Wheel – Part 1: Creating a Sequence”
OK……..but without knowing the indiviual batch-size and the total volume in a given time, as well the time it takes to change between the different “products”, how can we define the sequence?
You have a good start on the information needed, Gerhard. Same thinking on lot size as we have with Tak time. Pace is King. We must satisfy the customer demand so, our replenishment speed must be consistently achieved. Like in dedicated environment, we set takt time to set the production pace to meet the customer demand. We must establish the lot size based on number of C/O per day and the total number of products we run on that specific asset. With the principle of “Every Product/Every Day” we can determine lot size based on Total number of SKU’s divided by the Number of C/O per day. We get the total number of C/O per day based on how much time we have for Changeover in a given production day. We start with our TPS Guideline of “10% of available time is used for C/O.” Sometimes I use 10%, sometimes 8%, Sometimes 15%. We must consider the business we’re working in to determine our time available for C/O (Of course we consider EOQ). EX: if we can only do 4 C/O per day and we run 12 SKU’s. We have a lot size of 3 days. Set the pattern to run all parts in the specific asset to complete the pattern every 3 days. Now, you can see that if we can do 12 C/O per day, and we run 12 SKU’s, we can do every product / every day. Conclusion: reduce C/O time to reduce lot size. I hope this helps
Hello Gerhard, this first post (of a longer series) looks only at the naked sequence that would be best if you have to make all products. It does not (yet) look at what jobs to actually do. This I will talk about in the upcoming third, fourth, and fifth post of this series.
Christoph, Thank you for another new Blog. I look forward to following the “Creating a sequence” posts. I have a background in Mathematics more specifically Discrete mathematics and its applications. Your simplified explanation of “weights” on each combination was great. I immediately started thinking about the “graph” you would get from the ice cream problem.
Hi Justin, my next post refines this a bit. Finding a good sequence starts with these “weights”, but often there is more complexity to this in the real world. Plus, the number of ways to arrange even a small number of options (product types) in sequence quickly becomes astronomical. More next week 🙂