A gentleman (Discoducky) on the TM forms had this post on manufacturing ramp up that I thought was pretty damn cool. It seemed relevant.
Here ya go, you asked for it...
It's not going to be linear if they do "quality gates" in that they will build a small # of units, put them under test to failure, examine the results, approve change orders, build the same # of units again and do it all over again until the parts meet production spec's. Then there is also "reworking" online and offline to consider as well as an iterative break/fix cycle of stepping through the gates over and over. I've brought in manpower to handle this for simple reworking. It works but only in a pinch for small amounts of time.
Oh and I should explain quality gates: This is where you build say 5 units as fast as you can (aka your ideal line speed) and if they all pass and you built them in the ideal amount of time that is said to pass QG1. Then you build 20 and pick 5 to examine at random, if these pass then you pass QG2. But if you find one flaw, then they all must be rexamined and you go back to QG1. If you pass then you build 100 and examine 5 again, at random. See the steps? Once you do this a few times it becomes second nature because you get solid quality at scale without over burdening Quality Control. Yeah Statistical Anaylsis! But it really sucks when you miss something and have to scrap scale quantities of anything (I have stories...). This process nearly elliminates that risk.
These are very effective in finding MTTF issues as well as throughput/scale issues in material, workmanship and complexity of assembly (picture 3 types of connectors, screws, bolts or whatever that are different but look identical except for their part number). That sucks for the person installing it and being measured on their quality and speed, not to mention the QC person.
Anyway...If they *did* it this way *and* the VIN00023 is a product of that process then I'd say that they are still at QG1 and are iterating there. Once they pass QG1 we'll know it since they'll really want to show it off since this will be the first unit that can be driven, beaten, used and abused by anyone with a drivers license. I'll be worried if we don't see that prior to April.
For computers my typical quality gates to ramp were: 5,20,50,1000,100000 (Inspect 5 at random). Each representing a week's worth of time. So if you find a flaw at QG5 and you need to go back to QG1 (depending on the flaw) then it could cost you dearly. This rarely happened but did more often for new lines.
If TM did Quality Gates I'd expect more like: 3,10,20,50,400 (Inspect 3 at random, but usually first through the line, last and shift change)
So TM should be able to ramp up to full production without issue in 5 weeks, but that won't happen I'm guessing. It will take 5 to 10 cycles to hit 400. That should take about 4 to 6 months (8 to 12 weeks) assuming all the other things I pointed out go smoothly.
In order to hit their goal (6000 by years end) and if I had to guess it would be this:
QG1 needs to pass by the end of Jan at the latest with at least 3 perfect production cars built at scale speed.
QG2 in Feb/March with 23 (accounts for 7 cycles of break/fix)
QG3 in April with 100 (accounts for 2)
QG4 in May/June with 500 (accounts for 1)
QG5 in July/Aug with 1000 (accounts for 1)
He also had this to say, actually earlier in the thread, but it's reasonably separate:
I have a hard time believing the 6K number only since I've had to bring up several computer manufacturing lines. And imagine that vehicle manufacturing lines are mostly just bigger and have more complexity due to the nature of the product (like NHSTA testing).
If they really did start on Oct 17th and had all the raw parts then *maybe* since VIN00023 was at Bellevue I expect this was built in Freemont, but did not have production parts nor paint finish. So I expect that they are building one or two a week and working through bugs. This could go on for at least 4 to 6 months depending on the level of change requests that go through. So April is the month of no return, since any hard tool change (any change that is necessary to reach production quality and scale assuming no redundancy) and after this time will mean they miss July deliveries.
The only way they can do 6K cars by this time next year will be:
0. If they did start using the robots on production spec parts on Oct 17th then maybe since each and every part will need to have tweaks as well as the machine or process bringing the part to production spec's. When I build schedules each piece is assigned a risk value based on critical variables, then floated to ensure it doesn't hit critical path inside of a certain tolerance.
1. They don't have any supply chain issues on critical raw elements or components. Can't stress this enough as shortages will stop the line. Several risk mitigations need to be in place like pre-pays, just-in-time contract stipulations and supplier quality assurance guarantees.
2. Robot tech's on site. Those shinny new red robots are going to have issues when they start building. Techs need to be onsite for programming issues and general maintenance during ramp.
3. Continuous process improvements in place to ensure that TM assembly line techs have the ability to provide feedback to engineers so the line throughput can be improved day over day.
4. Iterative BOM cost reductions. If TM isn't focused on this from day 1 it will eat into their margins faster than the product comes down the line.
5. Software burn in. It's not what it sounds like, but is a way of thinking to run diagnostics to ensure MTTF are found for each unique product.
These are just a few but usually the most important for a new line.