Munkum

Apparently the suppliers told Tesla that the dates were impossible, but Tesla management didn't want to hear that and continued to pass on nothing but good news up the chain. It wouldn't surprise me if Elon only recently learned just how bad it is all going.

There is no way that they are using friction stir welding in a high volume line. It's a very slow process, visually it's horrible (big scar along the joint) and, though I've no experience with it I would have thought that pushing a spinning piece of metal at high temperatures through two thin abutting pressings would be highly unreliable.

It isn't even that it's hard. I mean, it's been done before, many times. But there are processes here with a stack of key control parameters that can and will drift.

Panels will not be consistent - they never are (and you'd be surprised at the variation). Rivets will tumble in feed tubes or get trapped in escapements. Vision systems will get confused by ambient light or reflections at certain times of day. Weld spatter will accumulate on tips, and frequent tip dressing will change weld parameters (timing, pressure, current etc). Adhesives will be effected by temperature, humidity. Batches of sealer might not be consistent. Weld timers will fail. Power supplies will fail. Flow drill screws will deform, go off centre and be a bitch to get right. Laser welder lens will accumulate weld crud. People will mis-load parts. PIA bits will come in below the required quality. Tools will get shimmed wrong. Robots will be badly programmed.

All of this stuff will take time to get right. It isn't impossible, but it is very difficult to get a line performing reliably. Tesla will be integrating a lot of new technologies, and at volumes that they are not used to, in an area with no serious history of car building, in a high wage state. I doubt they can get the people to manage the line and maintain this equipment.

Really not the case. Space X have built, um, not many rockets. Tesla are planning to churn out hundreds of thousands of cars.

The key isn't being able to do it (as someone else said joining metals is a solved problem), it's being able to do it with high quality and in a reliable way. Do you think Space X joining techniques are optimised for making 100k joints in a factory environment?

No, this is not a good plan at all, it's incredibly risky.

To get the car all road legal it needs to be crash tested - that is usually what the pre-volume builds are used for. If they fail (and they sometimes do, because you can't do a finite element analysis on an assembly as complex as a car in the variety of scenarios that you would crash test it) then you need to reengineer the car. This usually means panel gauge increases, sometimes it means additional reinforcement parts (boron steel inserts and brackets a lot of the time). If that happens the production could be delayed by months as they'd need to find a way to feed those parts to the line and modify any tooling that might be effected.

The body on an electric car is basically no different to an ICE car. The floor is flatter and thicker to accommodate the battery pack and no need for a tunnel to accommodate the drive train, but other than that they're about the same.

Tesla will almost certainly be using a combination of materials on the model 3; mild steels (cheap and cheerful, easy to press), aluminium (lightweight) and boron steels (strong). To join these materials you need a variety of techniques including resistance spot welding, aluminium RSW (altogether more difficult than steel because of the low melting point and high conductivity, and the fact that weld spatter sticks to the tip of the gun), riveting (for steel-Al joints), structural adhesives, flow drill screws, laser welds etc. All of these processes take serious time to calibrate and get right, and some are not all that robust (by high volume standards).

There are many quality loops with a vehicle launch. The tooling needs to be proven out. Tools get shimmed relentlessly to make up for panel variations - panel quality can vary from batch to batch within a press run. This is especially true for mixed material cars.

Believe me, these are not easy problems to solve and nothing I've seen and heard out of Tesla makes me think they've addressed any of these issues.

There is no way on god's earth that Tesla can ramp up production like that. I suspect the first cars that they sell will roll off the VP tools and not off the main lines anyway. With a lack of product maturation the quality of the first cars will be appalling too.

I'd bet that the guys at the coal face are wanting to kill their PR people right now for making promises that they know will get broken.

Disclosure: work in the automotive industry designing high volume production lines and planning vehicle launches.

I work in the automotive sector. Not interested in human-robot collaboration at all, I was wondering whether this would be a suitable alternative to PLC control of spot welding and handling industrial robots. Even though the processing power of these things is likely to be better than that of a PLC, am I right in thinking that the overhead involved with running a proper (albeit fairly stripped down) OS would make it slower than a dedicated PLC? Obviously it has its place (and my curiosity is most definitely peaked) but controls hardware costs pale into insignificance compared to the capital costs of everything else in a car plant, so we'd need a pretty compelling case to switch.

How does it benchmark against other modern PLCs? Could you give a ballpark figure for the scan time of a 10 robot cell for example?

Can these be programmed by any of the 61131-3 languages?

The more parts in assembly there are, the harder it is to control quality. You have to trust that the 3rd party supplier is meeting the required spec in manufacture and that the parts are damaged in transit.

In car assembly lines there are often gantries, conveyor systems (accumulating and EMS), trunnions and other high level equipment that prevents the use of track cranes. First choice would be to extract heavy equipment by manitou or fork-lift, crane is a distant second best. It sounds as though Tesla aren't considering how to maintain or repair their model 3 line.

I read that and have zero confidence that there will be a leap forward in productivity.

Density of automation equipment is all well and good, but it comes at a price of being difficult to maintain. If a tool or a robot breaks in some way (and any FMEA will tell you about the myriad of ways that can happen) then people will need to get into the cell and fix it. If you need to remove a robot, gripper or fixture - well, that's going to be difficult if you've landlocked everything by packing the machinery in so densely.

Vision systems for handling are great idea, but they are a nightmare to set up and to keep clean, especially in a welding shop.

Robots that repair the looms of robots around them. Fantasy, and it will continue to be fantasy for a good while yet.