Fluctuations on Continuously Moving Assembly Lines 1—Don’t Move the Workers Around!

A lot of industries use continuously moving assembly lines. This is common in automotive, but also in many others. The challenge is to manage the fluctuations. You cannot decouple using inventory; you can decouple only long-term fluctuations using capacity, while all other fluctuations are decoupled using time. This is the first blog post of a series digging deeper into how to handle or NOT handle fluctuations on continuously moving assembly lines. This first post looks at the effect when workers move among different stations.

Introduction: Continuously Moving Assembly Line

Fluctuation (or in Japanese mura) is one of the evils in manufacturing (along with waste and overburden). There are three fundamental ways you can fight fluctuations. You can buffer them with inventory. You can add more capacity to handle fluctuations. And, as a last ditch effort, you can use time to just wait out the fluctuations (albeit your customer may not like that one). I have an entire blog post on The Three Fundamental Ways to Decouple Fluctuations. Inventory is a popular way to decouple fluctuations. However, always keep in mind that it is also possible to REDUCE fluctuations.

However, especially for continuously moving lines, inventory is not an option to decouple fluctuations. If you add inventory directly to the continuously moving line, it will only take up space; it offers no benefit. (Note, however, that if you have a buffer that is NOT continuously moving, then you can again use inventory to decouple fluctuations. Toyota does this—see my blog post Evolution of Toyota Assembly Line Layout – A Visit to the Motomachi Plant for more information.)

Furthermore, capacity is difficult to do in the short term. You can speed up and slow down the line to adjust capacity to match (e.g., seasonal fluctuations), but it would be very difficult or even impossible to speed up and slow down the line to compensate for fluctuations of individual workers on the line.

Introduction: Andon

The speed of the line defines the time available for one station to complete its tasks. If the worker takes longer than permitted by the assembly line, Toyota uses the famous andon system where the worker alerts the system if there is a problem and eventually stops the line if the issue cannot be resolved and the work cannot be completed in the cycle time. There is a lot more detail (like a yellow cord for alerts and a red one for stops), and—most importantly—someone comes immediately and helps. This support is often forgotten in the Western world, and that’s one of the reasons why many andon implementations don’t really work. For more, see my blog post All About Andon.

Waiting Time on Continuously Moving Assembly Lines

As mentioned before, for decoupling fluctuations on a continuously moving line, your primary option is time (i.e., someone is waiting on someone else). Let’s assume you have a continuously moving assembly line. For simplicity’s sake, let’s also assume all workers but one have no fluctuations and will perform the work exactly during this available time—no more, no less. However, one worker has fluctuations. The probability distribution for the duration of the work is usually very one-sided. Most of the time, the worker will complete the tasks quickly with similar speeds. But sometimes a problem happens and it can take quite a bit longer. In statistics we call this a one-sided or right-skewed distribution. Here I use the Weibull distribution, which is commonly used, but the Lognormal distribution or even the Gamma distribution are also options.

If this worker does his cycle faster than the line speed, then the worker has to wait on the speed of the line. The worker could work more, but the line is too slow (for this cycle). If the worker is slower than the line, we again have a waiting time, but this time the line is stopped and has to wait for the worker to resolve the issue. This is illustrated in the image below showing the random distribution (the probability density function) for this one worker in comparison to the line speed.

DON’T MOVE THE WORKER OUTSIDE OF HIS STATION!

At Toyota, the line is STOPPED! In many Western companies, the worker continues to work into the next station if he is (randomly) slower than the speed, or picks up his work already at a previous station if he is (randomly) faster than the line speed. I have seen an example of assembly line workers working two or even three stations BEFORE their own station since they were faster than the line. Furthermore, this was not a random fluctuation, but the worker built up a time buffer to have a smoke afterwards.

When he came back, his work was maybe even overdue, and he continued working one or two stations AFTER his station to catch up again on the work, until he again was three stations ahead for the next smoke break. In sum, the worker fluctuated between three stations ahead and two stations behind on the assembly line, edging out a time buffer of five cycle times for his personal use, at the expense of his coworkers.

The workers who actually worked on the other stations were negatively impacted. The station was designed for one (group of) workers, and having another worker working there too made it crowded, and hindered the other workers. It even happened that some parts were assembled in the wrong sequence, increasing the work to fix it afterwards. The other workers could have been quite upset… except they did the same thing too. The one who should have been upset would be the production manager, because this moving back and forth created a lot of chaos and hence waste.

This is an extreme example of intentional moving around by the operators to weasel out a break. Luckily, this seems to be rare (at least in my experience). However, much more common is moving around to catch up in case of problems. I still remember very vividly the worker at the end of an overhead segment of an automotive assembly line. The car was again coming down to the ground to be worked on, but the worker from the overhead segment was not yet done. Hence, he continued to work underneath the lower and lower car, until the worker had to do a duck-walk to complete his task. This happened quite frequently.

Again, this is an extreme example of workers working beyond the range of their own station, but this happens way too often in way too many companies, creating chaos and waste. Sometimes even an incomplete or faulty product is moved on simply because the worker ran out of time. Overall, I strongly advise against this, as the benefit is rarely worth the chaos. Instead, you better keep calm and stop the line to resolve this problem.

I will talk more about these stops and their impact in my next blog post in this small series on fluctuations on continuously moving assembly lines. Now, go out, make sure your workers work only on their assigned stations and stop the line if there is a problem, and organize your industry!