# how to handle FLOATING BOTTLENECK scenario?

Six Sigma – iSixSigma Forums Old Forums General how to handle FLOATING BOTTLENECK scenario?

Viewing 19 posts - 1 through 19 (of 19 total)
• Author
Posts
• #37120

Maharajan Chidambaram
Participant

Hello:
I am working on a project to identify the bottleneck as well as set up a control point in a continuous flow line of a window manufacturing company. The entire assembly line is manually operated. Therefore, there is huge variability in the cycle time at every station. We are looking at setting up a “control point” (i.e., a preferred bottleneck from a strategic viewpoint, to increase units produced per hour) that will set the pace for the line. From the initial time study analysis, I can see that there are 4 successive stations that have nearly the same processing time (with a +/- difference of about 5 seconds). As a result, we experience a floating bottleneck scenario. Could you please share your thoughts and suggestions on how to handle this situation ? rgds,
Maharajan

0
#108638

Mike Carnell
Participant

Maharajan,
Just a guess but they are within 5 seconds of each other on the average. Your bottleneck is probably not floating but you might want to look at the std dev. around that average.
When they are all the same time it is called line balance and that is a good thing.
Good luck.

0
#108644

Maharajan Chidambaram
Participant

Mike:
Yes, the 5 second difference is based on average. I have another question for you. When I performed the 1st time study, I had defined cycle time of a station as “the time it takes for an operator to pick up a part, process it, send it downstream and pick up the next part”. But this accounts for the idle time (like queue time, waiting time for parts etc) after the worker is done processing the part. I performed another time study weeding out all the idle time and calculating only the “effective processing time” (ie., time period it takes to process the part from the moment worker grabs one). I had different bottlenecks in the two cases. The line was nearly balanced based on the 1st time study while it was not in the other case.
So, which time study gives me an accurate result ? I would also like to know if throughput rate might be affected  (because of random variation in a station’s cycle time) because of a balanced line. This is more relevant in our company since every station is manually operated.
rgds,Maharajan

0
#108646

Mike Carnell
Participant

Maharajan,
The secondstudy is what we call entitlement. Nice number to know and a good target for the future. It doesn’t have anything to do with reality right now. Study one is the “as is” state. That is what you have to deal with.
Walk through the steps. 5S/Workplace organization and Standardized work will remove the people induced noise from the process. When you have that out you are left with the process variation. Those are projects and not necessarily SS projects. Once you have done the previously mentioned steps there will probably be some JDI’s, some basic control stuff and maybe a SS project.
The next steps are Kanban and JIT. Just doing it by the book.
This may sound stupid but you could go out to the line and sit beside each of the operators or fill in for them. If you can’t do that You need to just sit and watch them for a while. Don’t offer up advice – just watch. Ignore what you see for the first hour or so because you are going to bother them by being there.
Just my opinion.
Good luck.

0
#108648

K. Subbiah
Participant

Maharajan:
How do you define a bottleneck? Do you use the Takt Time as the target and compare the individual machine cycle time, for all the machines to the Takt Time? A bottleneck is one that has a cycel time greater than the Takt Time. In one of my projects, I used Takt Time, the average cycle time, and range of cycle time. Please break the work content of each station into tasks elements and time each task repeatedly (at least 10 observations). For that work station, find the 1) task element (s) that have high range. These should be your focus. Typically, automation would help.  Those with high average cycle time can be handled by floor arrangement, etc. Once you handle the high range tasks, do the time study again.Repeat these until all the stations have effective cycle times (average cycle time +/- 3 rbar/d2) that are under Takt Time. Rbar/d2 will work now as you have already taken care of the tasks with high range as far as cycle time goes. You should get a balanced flow. If you get bottlenecks at the ends of the line, you can handled it easily by adjusting the schedules of the downstream or upstream stations with buffers. If the bottleneck were in the middle, then line balance is the only remedy. Good Luck.

0
#108652

vpschroeder
Member

Sounds like you have read The Goal and are trying to implement it.
Your second scenario sound more correct.  The whole idea is find the bottleneck, and then make sure it is always working..an inventory or queue in front of it will ensure that.
since it is a manual process, you can increase throughput by making the bottleneck more efficient, according to the methods the others described..then you might get a new bottleneck, at which point, you repeat the whole process..

0
#108656

Dog Sxxt
Participant

First, you have to make sure you know how to conduct a realible time study. Do you how to rate manual pacing or motion? Some formal IE time study training courses will help.

0
#108666

John M
Participant

Maharajan,
Keep it simple. Firstly work out Takt Time for your production line. Customer Demand divid by Capacity = Takt
Once you have established takt time gain the cycle time for each station and see the balance of the line. A simple bar chart of each stations cycle time with a target of the takt time. This is a yamazumi or line balance chart.
This will show you where the constraint is. Next Map out each step of the process with the operators in each station identifying value added and waste. Show on your balance chart waste in red and value in green in each bar for each station.
You will now be  able to see how you can balance the line  by removing the waste (5C and Seven wastes) and redistributing work to balance the line as much as possible.
This willl give you continous flow processing. You should always plan  90% of work leaving 10% Buffer at each station. You then look at stdev to earn the right to remove buffer.
The station with the least amount of variation and closest to takt then can be defined as the pacemaker process due to the reliability and stability.
Where you cannot balnce the line due to a shared resource that gives variable delivery patterns you introduce KAN BAN Supermarkets. You work out the Variation in your lead times for supply and hold inventory infront of that station to ensure you always have parts.
Hope This Helps
John M

0
#108670

John M
Participant

Maharajan,
Just to add from your post I forgot to mention. Huge variations in the station due to it being a manual process just means you have not introduced standard operations. Once you achieve the previous steps of balancing the line it is important you standardise the work.
You should involve everyone in that station to look at each step of the process break it down into operations and identify at each step any quality or safety issues. Take photograhs of various stages also.
Get the operators to write the standard operation so they gain ownership on how this process needs to be run. This is not only an aid for the people in that station to carry out the work consistantly but also a tool for easy movement of labour.
If you  move labour into another station it is fully documented how to process the goods in that station also identifying safety and quality risks to them.
You should use it as a training aid. I always aim for each operator in the line having the flexability to be able to work proficiently in two other stations.
All The Best
John M

0
#108694

Maharajan Chidambaram
Participant

Hello Everybody:
Thanks a lot for all your suggestions. Yesterday, I had a chat with the manager about this issue. He is interested in having a control point (i.e., a strategic and preferred bottleneck station). For the stations that were nearly balanced, we would answer the following questions and choose the one that satisfies most of the criteria.
Is it feasible for the control point  (CP) to perform at the expected rate of x seconds/unit (current takt rate) ?  (‘x’ is demand based. The “CP” should be capable of meeting excess demand in future)
Can all stations other than CP work faster than the CP ?
During trying circumstances (expediting orders etc), is it be possible to have control over the CP when the expedited orders have to go through the CP ?    Can we measure the performance of the control point with sufficient ease?
Does it have a large capital expenditure? (If so, utilizing it for 100% of time will justify its cost)
Is it highly technical ?
Based on the control point station’s requirement, we need to make the other stations work @ a faster rate by combining work, reducing non-value added activity, creating work standards etc.
Is this approach alright ? I would like to know your thoughts.
I would appreciate if you can also elaborate on the advantages/drawbacks of having a balanced line as against one managed through a bottleneck operation.
rgds,Maharajan

0
#108703

John M
Participant

Maharajan,
Why would you want to have a bottle neck control point? If you cannot do anything with a bottleneck then it becomes your pace maker process.
Try Goldratt Theory of Constraints – The Goal and It’s not Luck. Drum Buffer Rope Maybe this will clarify your thinking. You really need to line balance to stop overspeed/overproduction which leads to all the other wastes!!
All The Best
John M

0
#108708

Mike Carnell
Participant

Maharajan,
John made some good suggestions for reading material. You can make this sound very complicated like throwing out things like rbar/d2 – it is an estimate of your std deviation which basically says look at what is causing variation in your operations. Don’t make your life difficult.
In most cases it doesn’t make sense to outrun your bottleneck. If you are in front of it you build WIP. If you are behind it you run out of work.
Like John said just balance the line. Keep it simple. Don’t worry about simulations – you have the best thing available – you have the actual line.
Just my opinion.
Good luck.

0
#108710

PB
Participant

Maharajan,
What does a window manufacturing process look like? Where is the start and where is the end? How many stations do the same work? Do these stations feed station only in front of it or do these feed any station in front?
Do you work by a set schedule? (50 windows per day) OR Do you have a floating schedule? Do you make custom windows? OR Do you make OTC windows? OR Do you have a mix?
Let us say you have 40 windows that you make every shift. That means you have 5 windows every hour that you make. (You would have a WIP of 5 windows when you start out in the morning). You could have 5 lines making 1 window every hour. You could have 2 lines making 5 windows every hour. The rate of manufacture will be much faster for the 2 lines. Within this 2 lines you could have one process where you may double up on the stations because 1 station by itself may be the bottleneck. If each line is making a different window then the resultant output may differ (for the shift).
Therefore, in order to see a clear picture, we need some additional info if you can.
PB

0
#108714

walden
Participant

Maharajan,
Echo Mike’s opinion on Theory of Constraints.  Want to recommend one of my own…
If you have already read “Learning to See” by Rother and Shook, get a copy of “Creating Level Pull” by Art Smalley.  Apply the concept of a Pacemaker to your production line.
I’m not implying bottleneck and pacemaker are equivalent terms (you definitely need TOC as well), but scheduling using a pacemaker is a practical way to accomplish your objective.
Cheers,
Chris

0
#108896

Leon Ortega
Participant

In late 80’s one system called pull system was really popular in manual operations specially in semiconductor assy.
The main principal is to set up your operation in line, and every operation become customer and supplier, So the first operation has the first step, them the second operation start demanding product for the second step and so on.
The benefit of this is that you only need to provide fixture to reduce cycle time in the operation inself to increase the speed of the line, so your cycle time will be more easy to measure. change over from one product to another can be more faster because you don’t need to stop the line. Constrains become only in the delay operations because you are building inventory on them. The goal is not to have any inventory,.

0
#130808

IE
Participant

Sorry I can’t offer any suggestions, but I am curious about the floating bottleneck situation.
How do you know the instant when the bottleneck has technically transferred from one station to the next?  The point in time when station B’s cycle time is longer than station A’s?  (Station A being the current bottleneck and station B being the future bottleneck)  More importantly, what can we do with this information?
Thanks

0
#130811

Mike Carnell
Participant

IE,
Why do you want to know the “instant” it changes?

0
#130828

IE
Participant

It seems like it would give you a different outlook on that station when it changes to a bottleneck, wouldn’t it?  For instance, you could re-calculate the potential output of the system once its change, couldn’t you?  I was thinking there would probably be other useful results, but then again, maybe I’m looking at this all wrong.  It was more of a theoretical question, but its always been something I’ve been curious about.
Thanks

0
#130883

Mike Carnell
Participant

IE,
If you have a well balanced line your bottleneck can shift constantly particularly if you have a process that is nort well controlled or suppliers that have problems. The minute there is an issue at one operation the bottleneck shifts. Basically there isn’t any real advantage to recalculating anything. It goes to whatever speed the bottleneck is running. If you look at the basic TPS flow you have several steps ahead of JIT/Kanban because successful execution of those concepts is contingent on a predictable flow. The same basic thought process is around the bottleneck.
Just my opinion.
Good luck

0
Viewing 19 posts - 1 through 19 (of 19 total)

The forum ‘General’ is closed to new topics and replies.