What do you believe are baseline levels of complexity for an assembly

[Ni200]

I am attempting to establish some baseline expectations for expected lead times the company I am working for. I know a we can get lost in the weeds of how complex a singular part is that may affect the whole, but I really am attempting to keep things as simple as possible.

One of the parameters that may shed some insight onto the base lines I am proposing are the following: I work in an R&D environment which is heavy on the Research element, and secondary to the development. We have many fresh scientists and only 2 mechanical engineers which are also fresh, but very bright fellows. We have very little in standardization which is why I am trying to establish some now hopefully with some extra help from you fellows. Lastly our Design reviews are a bit of a joke, coworkers in the same project will "review" each others work (but that is a separate tangent).

I would be very curious to find out what you guys think of this proposition and what I may want to really include/exclude.

For Simplicity and Clarity I propose 3 levels of complexity for a project that may run through the shop.

Simple: that being a maximum of 2 components that have interaction. This would include no features that are beyond very basic DFM parameters (Think protolabs design guidelines). Tolerances can work from all parts being block dimensions.
(quote for each part is sufficient for baseline estimate)

Semi-Complex: That being a maximum of 4 components that have interaction. A maximum of 2 features may be beyond the basic DFM parameters. Some GD&T maybe required to ensure fit. Maximum of 3 tolerances may be beyond basic block tolerancing to ensure intended interactions.
(a minimum of 1 1hour meeting is to take place to ensure features do not conflict and are necessary before any quoting of job can begin)

Complex: That being a maximum of 7 components that have interaction. Features may be beyond basic DFM parameters. GD&T is required to ensure fits. Tolerances maybe beyond block tolerancing to ensure intended interactions.
(a minimum of 3 1hour meeting is to take place to ensure features do not conflict and are necessary before any quoting of job can begin)

Justification for each are as follows.
Tolerance is a fundamental aspect of machining and assembly. When tolerances can remain basic with intended interactions simple assemblies can be created quickly and easily, but as tolerancing becomes more specific and constrained (think limits and fits) the complexity of the whole assembly increases - not in a linear fashion but in exponential.
GD&T is the next step in fundamentals when it comes to machining and assembly. When GD&T requirements are applied to a part the whole has to be remembered. The more GD&T that is applied to features to ensure fits to more we have to do in the shop to ensure when a part is complete it will interact with its intended mating feature.
Number of Parts as this grows so does the level of complexity. once again this dose not grow in a linear fashion but in an exponential one. More analysis should be taken before processing will even begin. The amount of time and money to create a 7 component assembly for it to need rework or to flat out fail is exponentially higher than an extra 3+ hours over a weeks spread hammering out the finer details of an assembly.

Again trying to keep it as simple and straightforward as possible.

What do you guys think? Would this be an okay starting point for shop expectations? anything you think I should add or really hammer home?

 

[cngbrick]

It's difficult to comment without knowing your industry, costs, processes, precision levels, materials, how this relates to other aspects of your projects, risks, number of parts in a run, etc. Can you tell right away if your parts will work when you get them on your desk or will you only find out when your assembly landed on Mars? Context is important in your decision process.

 

[mhajicek]

Severity of failure also plays a big part. Is it a kids game, or an implantable medical device?

 

[AndyF]

Unless the goal is to either sit around a conference table or teach your engineers and scientists how to properly specify parts for your shop, I don't see why you need to schedule a minimum meeting length.

I like the idea of a progressive set of hurdles as assembly complexity increases and would look for ways other than a meeting to show that this analysis had been properly done before drawings are submitted for a quote.

 

[BoxcarPete]

Unless the goal is to either sit around a conference table or teach your engineers and scientists how to properly specify parts for your shop, I don't see why you need to schedule a minimum meeting length.

I like the idea of a progressive set of hurdles as assembly complexity increases and would look for ways other than a meeting to show that this analysis had been properly done before drawings are submitted for a quote.

Yep, meetings ain't necessarily the best way to get this done. I know a guy who was given the task of tightening up the assembly of a newly designed crash test dummy, which was constantly plagued with build errors. As you might expect, this was something of a big job. He told his boss, "I'm going to need some time to do this," and boss replied, "You have it, but it needs to work when you're done."

Three months later, after reviewing each individual drawing and making the minimum number of design changes, altering only tolerances and GD&T wherever possible, the newly tuned up dummy was built up with seven build errors, down from an average of over 170.

Moral of the story is, sometimes many cooks in the kitchen will not be as good as one dedicated person going through each component interaction all on his own, even (perhaps especially) if he had not done the original design. On the other hand, many eyes may catch flaws that could be missed by an individual no matter how long he stares at the sheets, so a group review is warranted as well.

 

[jccaclimber]

Doesn't really help in generic terms, but I've had good luck instituting specific rules/questions around things that have caused problems in the past. Some examples:

1. "Will this part work if any dimension is missed by more than 50% of the listed tolerance? By more than 100%? More than 200%?
If it won't work at a 50% miss I want to know the inspection plan, both at incoming and at the machine shop making the part.
If it will work at 200% then it had better not be tighter than our generic block tolerance, and in some cases should be looser.

2. We had an issue with engineers specifying a different size fastener for every part, many really small. Our standard screw is an M3 pan head. Anything smaller requires justification...per screw size you want to shrink by. Anything larger gets a free pass if it's already used in the same assembly, but requires justification if not.

 

[DDoug]

Sound like you need to hire a well aged (semi-retired) producibility person.

 

[Comatose]

Does the CAD package you use not have a stackup analysis? Do they not know how to use it?

Seems like throwing arbitrary levels onto things is obfuscating the real problem. I can design a two-element assembly that will be impossible to make, or a thousand-element assembly that you don't even have to measure.

 

[mhajicek]

Yeah, as said above, meetings don't really help. Meetings help management feel they're in control, but the engineers get stuff done either by themselves, or by direct communication at the moment information exchange is needed. Back when I was employed by others, I've had product development situations where the cycle is to spend the 2 hour Wednesday meeting convincing my boss to let me implement the solution I have, then modify my solution to meet the bosses demands, then attempt to implement on Thursday. I encounter a problem with the bosses demands, call or send an email, spend the rest of the day trying to explain the situation, then get told to wait for next Wednesday's meeting when I can show them the parts in person. The project then sits idle for five days each week.

I've also seen plenty of what I think is the worst enemy to proper design: pressure from the C suite to release too soon. The engineers want to make sure everything is right before release, but management doesn't give them the time and resources to do so. At the project launch meeting, the engineers estimate 18 months to do the job right. Manager hammers them down to 12 months, then six months, promising to hire more help and spend whatever it takes. Then those resources never show up, but the deadline keep getting cut shorter, and at four months there's pressure to release right now. All the right product development procedures in the world won't help if you're not given the time and resources to follow them properly.

 

[Ni200]

Thank you all very much for the suggestions and input.
I ended up scrapping the idea of required meetings and instead asked for a stack up analysis, a process control plan, and an independent review before release. I hope this will make strike some balance between process flexibility and adherence to the process itself.
Since there is no document at my company of what a manufacturing process control plan I have also tried to create a very rudimentary template of one to try to hammer home the need for better planning before taking steps beyond the design phase.

It's difficult to comment without knowing your industry, costs, processes, precision levels, materials, how this relates to other aspects of your projects, risks, number of parts in a run, etc. Can you tell right away if your parts will work when you get them on your desk or will you only find out when your assembly landed on Mars? Context is important in your decision process.

Sometimes I cannot tell what is going to work because I am not given foresight of what the true assembly will be. I work in the environmental energy sector. I will typically get a drawing with minimal details or notes, and told it will just bolt to some bracket, then a week later the engineer will come by to state the part I made will not fit the xometry part they ordered and can I please modify the unit. Oh then a week later it is another couple components that wont fit together because a while doing a braze the parts shifted in the oven. I guess it is no fib because that assembly will finally then be bolted to some bracket .

Unless the goal is to either sit around a conference table or teach your engineers and scientists how to properly specify parts for your shop, I don't see why you need to schedule a minimum meeting length.

I like the idea of a progressive set of hurdles as assembly complexity increases and would look for ways other than a meeting to show that this analysis had been properly done before drawings are submitted for a quote.

Agreed and scrapped that from the proposed document.

Does the CAD package you use not have a stackup analysis? Do they not know how to use it?

Seems like throwing arbitrary levels onto things is obfuscating the real problem. I can design a two-element assembly that will be impossible to make, or a thousand-element assembly that you don't even have to measure.

There is one. Engineers have begun to use it now. I work for a startup that has very bright, but very green engineers/scientists. While I myself am not great explaining the in/outs of proper GD&T. The company had no standard or documentation whatsoever of how to create a drawing or how to apply tolerancing. It has been a long road of self education and also educating the company. In prior slides I mention that this may not be the cure, but we do need to start somewhere.

Here is the altered proposal, what do you guys think of something like this. Please remember I am treating this as more of a reference point to set some sort of standard for us to orbit. I do expect this to change over time once/if implemented.

• Simple: that being a maximum of 2 components that have interaction. This would include no features that are beyond very basic DFM parameters. Intended interactions work with components need no secondary processes (component goes from machining, to welding or vise versa for completion).
  • No required prelim meetings
  • Stack up analysis
  • Minimal manufacturing process control plan information
• Semi-Complex: That being a maximum of 4 components that have interaction. A maximum of 2 features may be beyond the basic DFM parameters. Components may need secondary processing (component goes from machining to welding or vice versa for completion).
  • One preliminary meeting
  • Manufacturing control process plan (what components will be fabricated, machined, or purchased, how many expected operations will be needed, and in what order shall they be done)
  • Stack up analysis
  • Independent review from someone outside of project’s team
• Complex: That being a maximum of 7 components that have interaction. Features may be beyond basic DFM parameters. Components may need secondary or tertiary processing (component goes from machining to welding to machining for completion).
  • One preliminary meeting
  • Manufacturing process control plan (what components will be fabricated, machined, or purchased, how many expected operations will be take place, and in what order shall they be done)
  • Stackup analysis
  • Two independent reviews from people outside project’s team

 

[cngbrick]

It looks like you are making some progress.

I notice that you don't explicitely identify a place for feedback from your vendors in your evaluation process. Considering that you are working with a team that although bright has little experience in manufacturing I would suggest that you try partnering up with a few shops that have proven they can respond to your needs and provide good quality parts. You may have to pay more (but not necessarily so) than an outfit like Xometry but building a rapport with a few trusted vendors that know the time they invest in helping you will be rewarded by getting your business will more than offset the few dollars you may save by shopping around. Your team then learns from the feedback provided by experienced and typically very practically oriented contributors.

RT

 

[Plane Parts]

As a job shop that charges hourly I welcome the most complex systems imaginable.

 

[hanermo]

To OP Ni200 --
Bright and or brilliant engineers do actually not relate to great results in terms of cheap to produce and high consistent quality.
They are what you need for a one-off prototype to jupiter with no time or cost constraits.
Or a great new thingy, for somethingy.

WHEN and if You need to make very good parts, very efficiently, in either medium or high volume, what you need are either experienced shop owners or sometimes experienced volume parts manufacturers.

For example experienced auto parts engineers are generally not what You want. They have a terrible track record in making very good parts, very well, at good cost.

Solution:
IF I personally wanted to make great new thingies, that we had designed and produced at some high cost a working prototype of, I would ask my nearest local best jobshop for 10 days of product analysis and manufacturing advice -- and offer to pay 2000-20.000$ for it, around 1000 $ per day.

Examples:
Flextronics is the best volume manufacturer in the world.
They can do 56 machining ops for about 5 $ per part total.
WITH 2$ per part gross margin.
About 0.07$ per machining op.

Anyone here (generally) can make one flextronics iphone carcass with 56 ops for about 2-3000 $ - 10.000 $ depending if the speaker holes need to be lasered or reamed and post process and so on.
Typically jobshops need 1-2$ per machining op.
I believe I could not match flextronics, even given a 400M$ advance, 3 years, and a 100M parts count order for 700 M units eventually.
I might possibly be able to make the parts for 12-20-30-50$, each, given the need for 50 ops, but no way for the 4$ they need to cost them at.

Note that this is for qty 100 M parts per year with 10.000 workers, and one can buy 10.000 VMCs and or stuff to make them.
One line, one product.
5B$ per year for 700 M units made over 3-5-7 years.

SO, Your bright new engineers can make the thingie for say 3000$ once.
And for say 300$ per piece in qty 100.
And get produced for say 80-180$ each qty 1000.

A very good jobshop or contract manufacturer can drop your 80$/piece cost down to 30$/qty 1000, 14$ qty 20.000.

Example:
A car gearbox, performance sedan, is about the hardest thing to make in volume.
It is as accurate as a precision lathe, each spindle, runs 0-1000 rpm in insane 2 secs accelerations, is subjected to 20.000 kgf forces and acceleration and deceleration and needs to last about 5000 hours with no parts changes, both hot and cold climates.
With about 3-4 spindles running about 56 high precision gears/bearings, usually cycloidal and complex gears, hardened surfaces, and very fast and smooth gear-train switches as routine, sometimes several per minute for upto hours on end.

Oh, and You need to sell them for 3000$ each qty 200k per year, with insane quality guarantees for no general failures for 5+ years.

Like a hardinge 25.000 $ precision lathe, but 200 hp instead of 3 hp,
3-spindles of, aka 3 lathes,
subjected to insane peak loads, impact loads,
lasting 10-20 years often at great use,
for 1/10 the money.

--
My point is that making good things really well and cheaply and so they last is a Very Very special expertise.
You can hire that expertese really cheaply (20k) for really great improvements.

There is a near zero probability of initially having that expertese, which is easily achieved over 20 years for 2-4M$/yr producing ever better widgets.

 

[BugRobotics]

Yep, meetings ain't necessarily the best way to get this done.

x1000