Advice from a manufacturing engineer


outlier

Still Fresh
Joined
Aug 28, 2010
Messages
15
Hi all, this is my first post after lurking for a long time so get comfy it is a long one.


I work in a machine shop as a manufacturing engineer and I have some experience with lean manufacturing (a.k.a. the toyota production system) strategies. After watching the "Day in a life of..." video I have some suggestions about reorganizing your assembly line.


The big thing I saw is that you process units in batches. This kind of approach leads to a lot of extra handling and wasted movement especially since the actual work seems to be taking place in a separate room and then the in-process units are brought to that long table for staging. Having so many units in work at the same time is one of the "don'ts" of a lean approach since every device will spend most of its time sitting idle and you can expect to pick up and put down every device for each assembly step.


You can save a lot of effort by using a serial rather than a parallel approach. As a general rule you want individual units to start at one end of the assembly line and continue to completion, ideally without stopping. You can accomplish this by having all of your workers seated at the table where you had those dozens of partially assembled devices. For a broad example, suppose you divide up the process into a few work stations:


Lid assy -> Base assy -> Flashing -> Testing -> Packaging


Each station should have all the necessary components and tools close at hand and the work should be divided up so that each stage requires the same amount of time to ensure steady work flow.


I hope this does not sound like I am trying to stroll in and run the place, but I do think it would be helpful especially now that you are expecting the bottlenecks with board and case production to be cleared.


If anyone from the team is interested just let me know and I will be happy to go into more detail.
 
Well, you may not be aware that having as many units as possible sitting waiting for PCBs at this stage is probably a good plan - this isn't a well balanced just too late flow, they are receiving different parts in batches and in the early days spent about a week sticking labels on to 1000 cases since they were unable to ship a single finished unit.


You will probably find plenty of people on the forums with manufacturing experience - and all would give a different suggestion if asked...
 
Thanks for your suggestions :)


We tried a serial setup first, but it didn't work as well for a few reasons.


First, there are a lot of steps involved with a serial setup - about 10 - 15, which all need a different amount of time. So passing one part to the next one leads to a bottleneck.


Aditionally, four people are assembling - not enough for a proper serial setup.


The second is probably the main reason:


We have the LCDs and case parts ready. Assembling the lids is most of the work and that takes a lot of time. Therefore, we prepare as many of them as possible until we get the next batch of boards. When the boards arrive, they only need to be put into the prepared case, tested and screwed together (which is a serial setup - the built units will go to the testing area).


So basically, we're preparing as much as we can to speed up the final production ones those parts arrive :)
 
Hi all, this is my first post after lurking for a long time so get comfy it is a long one.


I work in a machine shop as a manufacturing engineer and I have some experience with lean manufacturing (a.k.a. the toyota production system) strategies. After watching the "Day in a life of..." video I have some suggestions about reorganizing your assembly line.


The big thing I saw is that you process units in batches. This kind of approach leads to a lot of extra handling and wasted movement especially since the actual work seems to be taking place in a separate room and then the in-process units are brought to that long table for staging. Having so many units in work at the same time is one of the "don'ts" of a lean approach since every device will spend most of its time sitting idle and you can expect to pick up and put down every device for each assembly step.


You can save a lot of effort by using a serial rather than a parallel approach. As a general rule you want individual units to start at one end of the assembly line and continue to completion, ideally without stopping. You can accomplish this by having all of your workers seated at the table where you had those dozens of partially assembled devices. For a broad example, suppose you divide up the process into a few work stations:


Lid assy -> Base assy -> Flashing -> Testing -> Packaging


Each station should have all the necessary components and tools close at hand and the work should be divided up so that each stage requires the same amount of time to ensure steady work flow.


I hope this does not sound like I am trying to stroll in and run the place, but I do think it would be helpful especially now that you are expecting the bottlenecks with board and case production to be cleared.


If anyone from the team is interested just let me know and I will be happy to go into more detail.

We do sort of do what you are saying in a way, when they are all layed out like that 3-4 people do one step on them each until they are ready to be moved.


I do agree with you though, it is faster if we could move each unit down a production line but it would take maybe 15 people and we don't have a company that big yet - however the ICP might change that situation.
 
I completely understand that you have had to work with whatever you had on hand given the irregular deliveries so far, and pre-assembling what you do have is also a great idea. But what are your intentions for the medium/long term. Pretty soon you will receive a boat load of cases and a stack of boards from Texas. Once you have cleared your supply of pre-assembled parts how will your process flow adapt? This is the case that I am thinking about. I know MWeston had written up a manufacturing plan for the Texas plant to quote from, and maybe you have your eye on that for the bulk of your production. If that is so and everything is thought through already then my feelings won't be hurt. I just saw a chance to contribute some of my own experience.


And four people is enough to manage a serial work flow. You just can't be as granular with the steps involved (obviously you couldn't have 10-15 stations). If most of the work is in the lid then perhaps you need two people assembling them to support one person doing bases and one doing testing. In any case an approach similar to this is worth consideration.
 
This is why a video of a Pandora being assembled would be awesome. It would not only satisfy those like myself who would not take one apart, as well as allowing those more technically minded to fully understand the process. It would also help those who have questions about case room, mods, "why can't I put in my own LCD cable", etc. and generally be a good response to those questions as well.
 
Ah yes the good old TPS system. It could be a good idea in the future and to set a kan-ban system etc. It is still early though.
 
Ah yes the good old TPS system. It could be a good idea in the future and to set a kan-ban system etc. It is still early though.

Yeah, the general consensus seems to be that it is too early to change much. They can worry when mounds of parts are laying around. My offer to help still stands though. As long as I don't have to write any TPS reports!


Off-topic: you did the case design right? What CAD system did you use? I am a Pro/e guy myself and got the feeling you used solidworks. Just curious.
 
however the ICP might change that situation.

miracles.jpg



They certainly would. Miracles up in dis Co. Durham
 
I guess if there are only 4 people then you need to divide the time the various steps take so their overal workload is similar in duration.
 
Yeah, the general consensus seems to be that it is too early to change much. They can worry when mounds of parts are laying around. My offer to help still stands though. As long as I don't have to write any TPS reports!


Off-topic: you did the case design right? What CAD system did you use? I am a Pro/e guy myself and got the feeling you used solidworks. Just curious.
Good guess with SWX. 90% of the time you would have been right, but no I didn't use SWX believe it or not :p


The CAD system I started out in was Inventor (got it free from work as a "home licence", they legally allow one home licence per seat) but gave it up halfway through. There were many changes going on with battery sizes, nubs, various parts, and the case size itself changes slightly. The editing of feature dependancies, constraints, and feature history was taking too long. I switched to a little known system "KeyCreator" which is a direct modeler as opposed to a parametric history based one. I could then change any feature at any time in minutes no matter what order I created the feature. I didn't have to hassle around with constraints, editing "sketches" etc. I wish I used that from the beginning, I would have had the case finished much sooner.


To put in TPS terms there was too much "wasted motion" going on editing and de-bugging feature trees in Inventor :p


For the ICP I took the Pandora model, stretched, chopped, pulled it together very fast in the beginning (although there was allot of debating about cost of the clamp design by the team so I did a few versions only to come back to my original design in the end).
 
Last edited by a moderator:
Ah, thread necromancy at its finest.


That is interesting though since the big advantage of parametric modeling is supposed to be ease of making dimensional changes. I went from a hideous wireframe/surface CAD package to full blown parametric solid modeling and now can't imagine going back. Have you used Inventor for a while? I did get to toy with it briefly when we were evaluating different systems and it was not so impressive. Of course that was ~7 years ago and it has surely improved since then.


In the end though you have to go with what works.
 
Ah, thread necromancy at its finest.


That is interesting though since the big advantage of parametric modeling is supposed to be ease of making dimensional changes. I went from a hideous wireframe/surface CAD package to full blown parametric solid modeling and now can't imagine going back. Have you used Inventor for a while? I did get to toy with it briefly when we were evaluating different systems and it was not so impressive. Of course that was ~7 years ago and it has surely improved since then.


In the end though you have to go with what works.
Parametric modelers are good when you know what you are going to (or might) change before you change it. That way you can set up everything based on what will change. Something like Pandora that was in total flux was a nightmare. The thing you thought you might change and set up for didn't and the thing you didn't think would change did. So there was allot of back-tracking and rebuilding of part recipe history trees. If you wanted to move a feature you did before others that you wanted to stay the same you had to do all kinds of gymnastics. These things are no secret, it is a well known issue with history based CAD. Inventor is basically a clone of SolidWorks. Inventor has improved in 7 years, and so has direct 3D modeling from those hideous wireframe systems. I still have to use Inventor at work so yes I have used it for a while. Sometimes parametric saves allot of time, sometimes it is like dragging a ball and chain. Editing a series of interlocking inter-dependant 2D sketches instead of just grabbing a feature and moving it is not always efficient. It totally depends on what you do. There is a best tool for every job.
 
Last edited by a moderator:
... The thing you thought you might change and set up for didn't and the thing you didn't think would change did. So there was allot of back-tracking and rebuilding of part recipe history trees. If you wanted to move a feature you did before others that you wanted to stay the same you had to do all kinds of gymnastics...

Ugh, been there before. Thankfully most of what I do is taking old (pre-CAD) designs and model them up and so avoid most of that. Since we also do our own machining all of our CAM files update with any design adjustments (if I do my part right). But you guys would not have any advantages like that working on the pandora so I can definitely understand the switch.

...Inventor has improved in 7 years, and so has direct 3D modeling from those hideous wireframe systems...

A quick rant: What we were using was hideous even by the standards of seven years ago. AutoCAD outclassed it by far in terms of usability, functionality, and stability. It was such a relief when we switched. This is probably why I am so enamored with pro/e.

DaveC: Have you had a look at OpenSCAD yet? Looks pretty nice, and someone like you might be able to do amazing things with it: http://openscad.org/

It seems a little primitive. Being a free software nut myself I have looked for a decent FOSS cad system and have been pretty underwhelmed with what is available, but I would love to hear a second opinion.
 
It seems a little primitive. Being a free software nut myself I have looked for a decent FOSS cad system and have been pretty underwhelmed with what is available, but I would love to hear a second opinion.

I've seen it used to make, for example, a lot of the things on http://thingiverse.com/ .. I don't think its 'primitive' per se - a lot of great, complex things have been made with it - but maybe it takes a lot of code to get something happening.. you write a script to build your model, which is pretty fabulous from where I sit, but hey .. maybe that approach is not for everyone. I would really encourage you to check it out and see some of the things that have been built with it - there is a monster spider robot that was built with this which really kicks ass.
 
Back
Top