I find this story interesting as well as your reply. It sounds to me like the design system was broken. In some companies the lines and definitions on what people are supposed to deliver are so blurry that things get overlooked. Never quite so big a thing as this. but still, how often does quality put on their engineering hat and start to solve the engineering problem rather than validating the design. How often do engineers start talking to vendors because, it's just easier and then I don't have to deal with purchasing.
Great example here of what not to do. But I think the focus on corrections has to go back to the system and how it failed.
I'm guessing there is no quality control within the manufacturing process. And for something like this, which could be used as a safety device, I would think the manufacturer would have to have some on-line testing to prove the product is capable of delivering on the specifications.
jmiller, that's what surprises me--that a company manufacturing a safety device can get away with no QA procedures, or at least, not the correct ones that would have caught this. And I think your comment regarding compartmentalization of responsibilities and job functions might go a long way toward explaining how that could happen.
Ann, your comment indicates that you think QC personel actually test parts. Too many times process and paper work are the only things checked. I was involved with a major project concerning a bearing shield. The customer kept rejecting the parts until they were in a "line down" condition and a company engineer came to our site to solve the problem.
We personally checked over a thousand rejected parts, but could not find one which was out of tolerance. However, the incoming QC would not approve the shipment because the dimensional variation, although always within tolerance, was great enough that an SPC chart indicated the process was incapable of producing good parts. It was only when the customer's engineer got hold of his plant manager and explained that all of the assemblies sitting on his idle production line could be completed with parts that had been 100% inspected and found to be within tolerance that we were able to ship the parts.
My point, if the inspection procedure is flawed, it does not matter whether the parts were inspected or not.
Tool_maker, you made me laugh. Yes, I assumed that either people or automated machine vision/inspection systems actually test parts. Having covered the latter subject for a few years, though, I did hear a lot of stories about companies not doing a very good job of either designing good inspection procedures, or of using the data their inspection systems provided.
Rob, automated MV systems are always superior to the human eye in speed. Ditto for accuracy when programmed correctly. Not all MV systems are for production lines, and the type and function depends on where on the line--or where else on the floor--the system is and what type of inspection it's doing. Then variety can be surprising. That said, MV systems are often, if not usually, put in place to go with existing automated systems, so have often been an afterthought. But that's often for larger manufacturers. Some smaller manufacturers aren't very automated, if at all, but install MV systems of some sort to improve quality control.
That's good to hear the MV systems are being used for improvement and not just a matter of replacing humans with a less expensive solution. There seems to be a lot of automation replacing humans lately. It will be interesting to see how IBM's Watson does in medical diagnostics.
Rob, one of the main reasons companies, especially smaller ones, put off investing in machine vision is the initial expense and hassle. They are not necessarily less expensive than human inspection except at the large scale. The cost differential in smaller systems seems to depend to a large degree on how much money they save in product returns and wasted materials from defective products.
That makes sense, Ann. A larger company will realize the incremental savings and improvement much more quickly. This is part of a trend in U.S. manufacturing where automation and the cost savings that goes along with it is helping to make U.S. manufacturing more competitive worldwide. Newsweek has a cover story this week on the efficiency of U.S. companies.
Rob, a recent study says that a growing number of large US manufacturers are looking at onshoring: bringing production back to the US. Although this article discussing it highlights plastics, the study is about all manufacturers:
Thanks for that, Ann. In another interesting piece of news, the productivity data released yesterday for the first Q of 2012, overall productivity fell 0.5%. However, manufacturing productivity rose 5.9%. Manufacturing in the U.S. is on a tear.
Yes, given the advances in mnaufacturing, we would actually be in pretty good economic shape is we still had a housing industry and we weren't in the middle of massive layoffs of city and state workers.
Thanks for that summary, Rob. That's really too bad--heartening news on one hand about a really important trend, and not so happy news about the employment scene (and the ongoing mortgage scandal fallout).
I'm still optimistic, Ann. We're seeing a lot of innovation now. This time reminds me of the early 80s. We were roaring in high tech innovation while unemployment was still very high. Eventually the innovation created jobs, which created more jobs.
Once we get more jobs, the 20-something post-grad kids will move out of their parents' home and drive housing growth.
Rob, I see your parallel with the early 80s, but so much has changed since then, including way more people and a shift to lower-paying service jobs that I don't think there are nearly enough good, mortgage-paying jobs for all of us, younger or older, in manufacturing. Or did you mean something different?
Good point, Ann. While I see many economic metrics that are similar to the early 1980s, I agree that our distribution of wealth and the distribution of opportunity has changed dramatically from 30 years ago. That means fewer good jobs as a percentage of what is out there, particularly for young people.
On re-reading Tool_maker's story I realized that not only was that inspection process flawed, but it apparently left out information that only a human could supply. I wonder whether the automatic SPC process could be adjusted with the correct information, or whether it would always take a human's explanation. In other words, can we really automate everything in QC?
Ann. I am sure every phase of manufacturing comes complete with its own set of problems unique to that area. I work in the stamping industry and maybe it has more variables than other areas, but rarely do SPC parameters written for maching operations serve any purpose other than to frustrate stampers.
The problem in this case involved the grain produced in the raw steel when it was produced in a rolling mill. When you form the material with the grain it reacts one way and when formed across the grain it reacts another. The parts in question were round so the forming went in every direction possible. As a result points checked 90 degrees apart would have a wide variance, with one on the low side of the tolerance and the other on the high side. Both within tolerance, but with a wide enough difference that when subjected to the SPC procedure the resulting formulation flagged an out of control process that would yield bad parts.
Our eventual solution was to only check parts in a restrained condition similar to that in which they would eventually mount onto the earth mover. The biggest problem we encountered was convincing the customer QC head that his methods developed in machining raw stock and castings, were worthless when working with stampings.
There are many cases where sensors have been mounted in stamping dies to monitor and grade parts before they even exit the tool, but sometimes I think "hands on" is the only way possible to properly evaluate the product.
Thanks, Tool_maker, for that input. It sounds like either SPC software needs to be adapted a lot to individual industries and/or specific manufacturers, which may take too much time and cost in engineer hours, or that users might benefit from industry-specific packages, kind of like what's emerging in industrial robotics. I know from covering machine vision that the first is sometimes simply not done for the reasons given (although the "too much" may be due to perception or procrastination), and that the second has not been successful because the technology is used so differently by each end user.
@Ann- LOL, shame on you for not switching hats from Engineering to Sales/Mktg ! The tape is a major breakthrough because it does not need to have sides identified. Both sides of the tape have equal conductivity. This saves a lot of time, labor, and chance for error on the production floor. :)
We like to make fun of engineers, scientists and their quality departments. However, this could be a case of your young sales person not understanding a new product and trying to sell into the wrong application. Shielded cable needs to have a. Conductive sheath around the signal carrying cable. However, other products exist that are absorbing (not shielding) that can attenuate outside signals. Absorbers are isolated metal (iron or ferrite) particles imbedded in a matrix that is wrapped around the area or cable to be protected or even the area emitting the undesirable frequency.
This might be a case of "sold by monkeys" instraead of "Made by Monkeys".
Friends of mine, who worked in quality control, said they were always under pressure to push the stuff out the door. Flagging bad products impeded productivity. It would be up to the customer to notice any problem and chances are they would not.
I'd be listening to this with my jaw sagging at the thought this behavior might be typical.
Years later I was upgrading memory in a computer and found that while the computer had worked for several years (expiring the warranty) that it could not work with the higher density memory that the specifications said it could. Obviously it had not been tested, but they were still able to kick it out the door and make money with it.
One place I worked I later learned typically quoted capabilities about twice what the product was capable of. Silly me, I had performed some time/motion studies on one system I had installed and estimated it could work at twice the current speed. The Salesman shushed me when I told him my results because they had sold the system quoting twice that rate. That behavior finally came back to bite them when a client said, "Show me." before they would sign off on the final payment. I calculated the system could never achieve the contracted rates. But the blame was placed on our department, rather than the salesman.
It sounds to me like quality is reporting to the wrong person. From the top down, if quality is reporting to operations they will be forced to push it through rather than slow down or stop production making them do it right. The goal of quality has to be quality and quality only. And they should report through a different chain of command all the way to the top. That way the big guy can tell quality to focus on quality and operations to focus on making good products.
Funny how often the products we put out are dictated by the system that we choose to function in and the directions we are given from the top down.
Sounds like something that worked fine in the lab but was never tested under actual production conditions, e.g. winding onto a large roll. Plastic stretches; metal films don't.
It's surprising how little stretching is required to make metal coatings break--I had a fair amount of frustrating experience making thermal imagers out of metallized PVDF film before I switched to carbon-loaded ink. That stuff was stretchy enough to take a hard crease, and conductive enough for the job. (There's a war story at http://electrooptical.net/www/footprints/fpwaropn.pdf, and a bit more discussion at http://electrooptical.net#footprints.)
I came up against exactly that sort of problem during an analysis of why some sensors were not sensing correctly. The product used connections to a gold alloy plating on each side of a piece of material, except that the coating was nonconductive in that areq, while on some samples it had a high resistance. I quickly developed a test fixture to check the problem, and we discovered a sixty percent yoeild in the coating process. But we did not produce any more deffective parts.
A few years ago while developing a very sophisticated proof of concept prototype of a laser projection keyboard, I had mistakenly planned for conductive adhesive tape –a tape very similar to the one described here – which was to provide an electrical interconnect signal for one of the finger sensors of the assembly. Being a mechanical engineer, I neglected to consider the high resistance of the conductive adhesive, and the entire design concept was jeopardized because of my lack of experience.I had **assumed** that conductive adhesive would be conductive – and did not consider insufficient conductivity.
The tape was manufactured using Φ.002" conductive spheres potted and floating within a .0015" thick adhesive layer.Surely that would work.But it did not. There was nowhere near enough metal content floating in the adhesive, and the (poor conductivity / high resistance) rendered my entire design concept back to the drawing board.But the experience taught me much about laminates and conductive tapes.
Fast-forward now to is article, and read again, what the Salesman brought in: he said he had a new electrical shielding tape.From the content of the story, his company didn't make any claims about conductivity.
Point being, it really is the design engineers' responsibility to understand completely the characterization of all the elements going into a design.I wouldn't spend too much time slamming the quality of the tape manufacturer.If a component is not suitable for your design, MOVE ON, and find one that is.
A new service lets engineers and orthopedic surgeons design and 3D print highly accurate, patient-specific, orthopedic medical implants made of metal -- without owning a 3D printer. Using free, downloadable software, users can import ASCII and binary .STL files, design the implant, and send an encrypted design file to a third-party manufacturer.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This discussion will examine what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.