Dear Gemba Coach,
I’m in manufacturing engineering and have been asked to design a lean line. I’ve participated to several kaizen workshops and have some general ideas about how to go about it, but I’ve also heard many horror stories of lines that were designed so “lean” they can’t run in practice. Any ideas as to how lean should lean be?
How lean is lean? Good point, and hard to answer without more specifics about what kind of line you’re looking into, but I do understand your concern. I have indeed seen lines designed to be lean and totally impractical to run on a daily basis.
What should a lean line look like? The focus of any lean line is its operators. Ideally, they should sow the components like a farmer sowing a field and all parts should assemble easily and naturally into a faultless product. In particular, this means that (1) workstations are designed to make assembly easy, (2) the product can flow from one station to the next without in-between accumulation and (3) the line should be flexible in mix as well as volume, with quick and smooth changeovers, and (4) mistake proofing devices should stop defectives without encumbering the line with over-complex systems.
From the operator’s point of view, let’s imagine a tray table fitted on your airplane seat for a business class dinner about ten inches deep and twenty inches wide. All components should reach the operator in that range so that hand movements can be within easy reach and avoid ergonomic strain. This, in turn, involves VERY small containers for components. A good sign is when you can’t find containers small enough in industrial equipment catalogues and need to go to Wal-Mart to shop for Tupperware.
The part itself should flow smoothly from one person to the next, so that it can be handed over hand-to-hand, as a baton in a relay race. This means initially having operator workstations placed next to each other – by no means a small feat with most semi-automatic machinery – and without any barriers to hinder part movement. One of the best lines I’ve seen had parts auto-ejected by machines and pushed along the line by operators on simple rollers, with machine cycles triggered by thin joysticks.
Ideally, machines on the line should be able to change tools in “one touch/one breath” – the idea is that the tool change occurs in the time it takes to draw one breath. Hardly realistic in most processes I know, but still a worthwhile goal to pursue. In any case, ease of changeover – by the operators themselves rather than specialized setters – is a key component for a lean line.
Finally, the line should be fitted with smart error-proofing devices (poka-yoke) to stop defectives from moving forward in the process, but without encumbering the line with hydraulics, cameras, and over-sophisticated electronics which will encumber the line more than anything else. I’ve lost count of production lines where operators have disabled the “poka-yokes” dreamed up by engineers just to be able to run the line.
3 Common Problems
If you add the usual lean criteria that the machines should not be wider than the part they make by more than 20% and that the spread of the “U” (in case of loading/unloading machines – if not a straight “I” line is better) should be less than 50 inches, you could design a very lean line indeed. The question is would it run?
Let’s look at the problem the other way around: why are existing lines so full of waste? The production lines I see on the gemba tend to suffer from three common ailments. First, equipment manufacturers propose large machines with lots of conveyors; second, manufacturing engineering tends to specify buffers in order to keep production running even when there’s a problem; and finally, both equipment vendors and manufacturing engineers tend to be partial to a lot of unnecessary automation.
To be fair on the first point, it’s hard to make a complex machine with a small window facing the operator. Usually, the equipment manufacturer (I’ve worked with some on their designs) will have a favored design on the manufacturing process, and then build the supporting processes around them, wrap the result with some solid steel and glass and then try to fit the machine in the line. Part conveyance is typically handled by internal conveyors or platters working like choo-choo trains. Designing machines to present a small window to the operator is quite challenging. Make sure parts can flow seamlessly from one machine to the next (for instance, safety lifeguards placed horizontally rather than vertically). Not to mention changeovers. All in all, the main effort in lean line design is not with the line design itself by co-developing the machines with the vendor so that they come up with the kind of equipment you need for lean.
The second issue is related to the first. Lean lines are “lean” because a lot of the variation has been taken out of the process, and buffers become unnecessary – and, in fact, a hindrance. But on the other hand, when the machines first arrive the usual level of variation is to be expected, and chances are that a lean line (without buffers) simply won’t work: where will the variation be stored when it occurs? To understand lean line design you have to relate every buffer to a specific form of variation in the process, in order to visualize that if you take the buffer out, you will have to reduce the variation at that step of the process as well. I recently had this experience with a large paint conveyor that was shortened progressively as the paint team controlled the paint process better, and no longer needed the parts on hangers for miles on end. However, if we’d started with the existing paint process and no conveyor, they simply wouldn’t have been able to run and would have had parts all over the place. The physical size of the line is often linked to specific variation points hard to master in the process – whether human (in the operator cycle) or mechanical (in the machine cycle itself).
Furthermore, in order to “facilitate” operator’s work, engineers tend to overdesign the automation on the line, with lots of hydraulics to pick up parts and place them, robots, visual captors and so on. Like the guy who ran out in the desert and sat on a cactus, when asked why they did this or that, engineers tend to answer: “at the time, it seemed like a good idea.” And, most of the time, it was a good idea, but executed with mechanical devices that add variation to the process rather than reduce it. The automation can be too long (causing the operator to wait for doors to open and close, for tables to rotate and so on) or untrustworthy (for instance, video-control processes or light captors that have a high rate of false-positives). It’s essential to make sure that the automation of the line provides the support required without hindering the operator’s movements, creating barriers to the parts flow or add unwanted variability in the line’s operations.
Line Design Implementation
All in all, designing a “lean” line is less of a challenge than having it run at target cycle time. In fact, you are quite right: it’s easier to kaizen a line to get it leaner than design it for lean at the outset, so here are two kaizen exercises we have learned to do with manufacturing engineers to improve line design.
First, construct your planned line with cardboard boxes and ask experienced operators to pretend to run it with real parts in the process. Run for about ten consecutive minutes, and then get the team together for kaizen. The point of the exercise is to (1) visualize operator movements and (2) visualize parts and components flow. In many cases, we’ll find that people are so focused on getting the process to work that they’ve given very little thought to these aspects of running the line in production. With the operators and with logistics people, you can design the supply side of the line while the machines themselves are still a black-box. If you invite the machine manufacturer in these sessions you will radically improve the design of the line right upfront.
Secondly, once you’ve spotted the main sources of variation in the technical process – which will need to be buffered – you can take your team to existing similar processes and run a kaizen event to improve them right NOW. By struggling and scratching your heads on how to improve the old processes you’ll understand a lot better what to do and not to do in the new ones. Also, you will train your team to react properly to the early problems your new machines will have during ramp-up. This is equivalent to training a sailing crew to respond to unexpected maneuvers before the race rather than once the “go” signal has been given.
You probably will be able to think of other instances where this applies: the general idea is to conduct kaizen before and during the design to help engineers correctly understand the problems they need to solve, rather than focus on what they’re interested in and let production deal with the rest as best they can.
Leanness is the outcome of a superior handling of flow and flexibility and reducing variation in any processes (since variation will show up as stoppages or buffers – waste). None of this comes naturally, so the design teams must be trained to fix these problems if they want to avoid them in their final design. To answer your question, the line can be as lean as your teams can run it. Leaner than that, it simply won’t work. We need to develop the teams before we can make the products, so the place to start is with a kaizen program for the engineers themselves. In one company, the rule-of-thumb we’ve come to is that a manufacturing engineer can only start her project on a new line if she has convincingly improved the performance of an existing old one.