Money Only Gets You So Far When It Comes To Buying Time
No, this isn't a misplaced post that belongs on a self improvement website trying to convince you that "nobody on their death bed has ever looked back on their lives and wished they had worked harder" (although that's probably true, too). In this case, the time I'm referring to is cycle time, and how I believe you should think about it in your design process.
Before we go any further, it is very important that you understand what cycle time means. There are a bunch of terms and variations on the terms that get used interchangeably but mean very different things and confusion between them can cause poor and expensive assumptions to be made. I have tried to cover some of that in the post you can navigate to here.
Back to the subject at hand…..I maybe can't even count the number of station cycle time charts that I've looked at over the years in design reviews that magically end up right at the overall contractual cycle time. It would seem that a lot of times, we have been given a concept by either the customer or the applications group, we jump into design thinking mainly about the required tooling and motions to do what the station was quoted to do, we come up with a station that is robust, easily maintained, maybe even very cost effective, and in all respects, would be considered an excellent design, but when it gets out to the floor, it is 20, 30, 50% over cycle time even though the cycle time chart shows that all should be right with the world. I think this happens because at some subconscious level, we mold the cycle time chart to fit the design we have developed because of the blood, sweat, and tears we have invested and the fact that the PM is asking every ten minutes when we'll be ready to release. I'm not talking blatantly cooking the books, but shave a second here, shave a second there, be optimistic on some process times, etc until we honestly believe we're ok. So I'd like to propose a slightly different way to look at things. It's not foolproof and maybe it's not even all that revolutionary, but I do think it can reduce the number of times that you end up with an over cycle station and could even streamline your design process.
The paradigm shift is simple to grasp, maybe a little harder to talk yourself into doing: start with time, not mechanics. I’m sure that could seem counterintuitive since you can argue that the layout and the mechanics determine the time, so how would you be doing anything but guessing without having a design first? I would counter that in most situations, more than half of the available cycle time is governed by things that are either pre-established by the laws of nature and/or currently available technology, or are times that are common to all stations in a system and therefore cannot be altered by your individual station design. So if you already know more than half of the cycle time before you draw the first sketch or model the first piece of tooling, all it takes is a mental shift to “how do I use the time that’s left the most effectively?” instead of “how much time does what I’ve designed take?” and use that cycle time analysis to actually drive your design decisions (where to place things, what technologies to choose, etc). The tools to do this are simple, fast and cheap….a scrap piece of paper, a quick spreadsheet, or my favorite when you’re just learning this method, a pad of post-it notes. Here’s the approach I’d recommend:
Identify all “process” times. These are screwdriving times, leak test times, pressing times, manual assembly times, etc. Anything that is basically at the mercy of technology or customer product design. These are the ones you need to put in serious effort to determine. You can leak test some parts in 4 seconds. Some take 3 minutes! Screwdriving times obviously vary depending on pitch and length, but even material, screw type, and torque spec have an effect. Guesses in this category can be very bad because they are usually the highest dollar components of your station and the whole station design typically revolves around the processes, so ask people. Do math. Google extensively, even prototype or do mock ups. And once you have those times established, add 20% and resist the urge to tweak the time down when you find out you’re over cycle later in the process unless you can find concrete examples where the new number is valid. No cheating here. It will bite you.
Identify all times beyond your control. On a conveyor system, this would be the pallet exchange time. On a dial, the index time. It could be bar code reader time or RF reader time (varies somewhat significantly depending on whether it’s read/write or read only). Another biggie in this category is data exchange with an upper level computer. Some companies have SCADA systems and require positive permission to work on a part and positive acknowledgement of the completed work before releasing the part. This can take TWO TO FOUR SECONDS (yes, that seems crazy, thus the all caps). Again, don’t assume a half second and move on. Get confirmation in writing from the customer. Document each of these times on a post-it note or list them out. What is left is basically time to move parts around or move tooling/process equipment into position and home, plus “overhead” stuff like cycle initiation, getting clear of a light curtain, walking, twisting, turning, bending for operators if it’s a manual or semi-auto station. These are the places where you will most effectively be able to use station layout and different technical solutions to impact time.
At this point, it’s good to capture what are called non-cyclical times that will be considered as part of the takt time during an extended run, but you may not be thinking about during design. This could be removing layers in a skid or pallet/box exchange for an operator, or tray exchange time on a tray handler in a robot or hard automation cell. Be realistic in your estimate. Exchanging a pallet can easily take 30-60 seconds, and many companies use their operators to make that exchange. Again, put all these times on a post it note or list them, but don’t panic yet…. a lot of non-cyclicals can be masked by other parts of the sequence or you can do mini-buffering to eliminate the effect on the overall, but they definitely need to be considered and listing them now, early in the process, will help keep them on your mind.
Start sketching out possible solutions for the station, most likely starting out with the concept that was sold (see post on hand sketching for my take on that topic). Once you have two or three possible layouts sketched, start putting rough distances on the part transfers and operator movements. Describe the motions. Are they pneumatic? Servo? Hydraulic? Pneumatic motions are still the cheapest even with servo prices coming down, so you should probably assume those to start with unless there is an overriding reason not to (high model mix, customer spec, your company standards, etc). There are rules of thumb for speeds…1-2 inches/second for pneumatic and over 10 feet/s for belt driven servo, with almost anything in between, but it varies significantly depending on the masses, forces, and strokes. Obviously, every motion has to accelerate to start, and will likely have a program function or shock absorber to slow down prior to stopping. So, even though an air cylinder may move 1”/s, it won’t be moving that fast at first or towards the end, and that effect on average speed is more pronounced on short strokes, so you really have to take the time to figure it out or start out really conservatively. Assign a time to each motion.
Lay out everything (processes, part movement, overhead times, etc) sequentially and add up all the times. If you are already under cycle time by 10-15%, no need to go further on optimization at this time, even if the station might ultimately be programmed to do some things simultaneously. But then you need to think about the non-cyclical times and make sure they don’t throw you over the limit.
If you are over 85-90% of the target cycle time, start evaluating each step to see if it can be done in parallel with something else. This is one place the sticky note method works very well because you can move the squares around without a lot of erasing and redoing. You are looking for ways to make things happen in parallel that just involve re-sequencing the elements of the solution you sketched, not adding complexity or cost at this time.
Still over? Time to get more drastic. This is where you have to start thinking about substituting technologies. Would a servo slide be fast enough to get you within cycle time instead of that pneumatic slide? Do you need to auto feed and present that part so the operator doesn’t have to twist, obtain, orient, twist and place? Can a part be kitted onto a post or nest in a time when an operator isn’t currently being utilized in the sequence so that it can be automatically handled in parallel with another operation or the manual motions during the critical path time are simpler? Don’t forget to step back and evaluate the layout….would an alternate layout allow for shorter strokes? Less wasted movement? A lot of cycle time can be eliminated just by proper arrangement of the station, but you do need to consider maintenance access and complexity and balance all of it. Anything in this category should be discussed with the PM and maybe others because here’s where you’re talking about changing basic content or adding serious money (see my post on the true cost of adding things here).
No luck yet? One thing that is overlooked pretty often on multi-station systems is the fact that no line is ever conceived in a perfectly balanced manner. Look at the processes you have on your station. Are any of them easily cleaved off and given to another station? Some parts of an assembly have a very particular order that logically falls within your sequence and would require serious re-balancing to move. Others can be done literally anywhere on the line. In between are things that could maybe moved upstream but can’t be moved downstream. Make a list of all the processes that can move, then look to see if anything on that list has totally unique tooling/process equipment that isn’t shared with other steps on your station. Those are the least expensive to move. If the step is movable, but shares equipment (a DC screwdriver is a common example), you can move it, but it will cost you the amount of money it takes to duplicate the shared equipment, however, this may still be the lowest cost overall solution to the cycle time problem. Now go to the other designers and the project manager and talk about what can be done. I struggled whether to put this as step 7 or 8 because it is a cost/benefit analysis, so this may make more sense than going to all servos to gain speed, for example.
If you can’t get there with all of the above, you’ll have to start double tooling (working on two parts at the same time) or changing basic chassis/technologies from what was quoted, which is of course, a major change so it’s time to raise the flag and get help so that everyone is working on the problem together. This is basically the nuclear “do over” option, but still better than finding the issue once an over-cycle solution is built.
Once you have picked a final solution that appears to satisfy the cycle time, add in all the non-cyclicals that you identified in step 4 and make sure you can account for those with parallel time (ie, the operator changes the skid while the station is picking and placing and pressing the bearing so that they’re back in front of the station without any lost time) or by using buffering. The buffering thing can be a little tricky, but it’s worth mentioning. If I have a 10 second cycle time and a one minute task to do every 250 cycles, the math says that I have to artificially lower my target cycle time to 9.76s (10s cycle time - 60s/250 parts), which doesn’t sound like such a big deal. BUT, that minute also represents 6 parts of throughput at the 10s cycle time. Those parts have to have a place to sit for the downstream station to consume them because this station is making them faster than 10s AND there has to be room for the 6 parts upstream for when the operator is off doing the one minute task. Chances are, the buffering will be more or less split between upstream and downstream, but to assure you never affect throughput, you need to account for all of it in both places. If your calculations send you back over (or 90-95% of over) your target cycle time, it’s back to steps 7-10 to find some more time.
If you’ve followed the plan, NOW is when you’d start modeling! I know, it seems like a lot of guessing without any “real” solution in place, but actually, you will be much closer than you think and going through this exercise really will guide your design. By this point in the process, you’ll have made several variations of sketches, have rough distances and technologies worked out and have done a fair amount of research and experimentation to make sure your times are solid. That is the true design part of designing. Putting the mechanics in place via modeling is actually secondary to figuring out the right solution. It’s the same argument I use in my sketching post…..spending a couple of days exploring options and uncovering roadblocks is almost always a good investment over just jumping in and modeling based on what someone gave you or your first thoughts. I would even argue it takes no more total hours because you will be modeling with purpose and confidence that the final solution is sound. But, the process isn’t really finished here. Once you do start modeling, you have to check your time assumptions as you go, which, since you’ve spent so much time establishing the assumptions up front, doesn’t take long. Now you are most likely making small adjustments in limited areas of your design rather than backing up to make a major change in direction. A word I use often is iterative, and it applies here, too.
It takes a while before you gain the experience needed for your timing estimates to be accurate. Don’t let that discourage you. This process is a great mental exercise even when it’s not perfect. Just make sure you review your work with with someone that has the experience so that early lack of accuracy doesn’t cause you to make bad design decisions. But, keep researching, keep experimenting, keep watching equipment run on your floor and you will become good at it. When you get good at it, it will turn you into a faster and more capable designer, so it’s worthwhile. And hopefully, it will keep you from being the designer that gets paged out to the floor with the always scary “and just what were you thinking here?” question.