Deming Must Be Wrong

Gary Wagner, General Manager (U.S.), Ophir Photonics

W. Edwards Deming said, "If you can't measure it, you can't control it," and "You cannot inspect quality into the product; it is already there." Instead of trying to inspect-in quality, focus on adding value once you're assured that the previous process was done right and to specification.

Since the mid-1980's, tens of thousands of machine vision camera systems have been installed along manufacturing lines and in assembly stations. They can be found in every conceivable industry where there are volume users - automotive, pharmaceutical, medical devices, electronics, aerospace, and, at this point, almost every industry that makes something. Why? Because subsequently adding value to a good part reduces the cost of goods, making the manufacturer more money. In fact, machine vision is one of the reasons the U.S. regained its world position producing quality products at reasonable prices.

I've been in the machine vision business for over 15 years solving manufacturing quality issues, and measuring and controlling manufacturing processes. A few years ago when I moved to a company that makes products for measuring the quality of a laser beam, I assumed the manufacturing industry would be a natural for these products. We make instruments needed to measure the various characteristics of beam quality and consistency…in other words, a machine vision system that measures and manages the quality of a laser beam.

So I asked our sales people how much we sell into the industrial manufacturing world, and the answer was shocking. Except for the guys that are under FDA scrutiny, almost none. I said, "What are you talking about? There are all kinds of lasers being used in manufacturing of high precision, high reliability parts in the aerospace, automotive, and electronics industry, to name just a few. They need constant precision and consistency, which definitely isn't synonymous with a laser over time." The answer was, "Might be, but they don't think they need it. We've tried for years to get them to understand the need for periodic or on-line measurement, but with very little success."

Not surprisingly, my reaction was, "OK, but I know something about these people and their needs and requirements, so let me try." What I discovered is that nobody thinks they need to periodically measure their beam because:

We've never measured it since the day it was installed, so it is probably OK, We have periodic service maintenance and we're sure those guys measured it then, or When we start making bad parts, we shut down the line and call the service tech. My head started spinning, taking me back to the early 1980's when these were the same kind of initial answers we got when talking about how machine vision would help make: 1) Consistently high quality, 2) Less scrap, and 3) More satisfied customers. We got our message through. What's happened since then?

Machine vision users are process engineers, the same users of laser systems. Why do they not see how beam measuring equipment can help?

Maybe they aren't seeing manufacturing inconsistencies from their laser? Wrong… It doesn't take many interviews to find out that the laser process can indeed make bad/inconsistent parts or that it may be shut down at unpredictable times for maintenance. So that can't be the reason for denying the need to measure.

We ask how are you measuring the laser now? They reply, a) Burn paper or b) Acrylic burns. These answers prove my point. Let's start with burn paper. I show them that burn paper is one-dimensional; the resultant burn is either burned or not. There is no spatial depth to a burn. Yet the power cross-section of a laser beam is 3D, not 1D. All of the information on how a laser is going to cut or weld or process is in how the power is shaped. In addition to its limited spatial depth, the burn paper is not dimensionally accurate. The burn only happens where there is enough power to burn; measuring it for useful information about power density is useless.

Let's address the second issue: acrylic burns. I hear, "I use acrylic plastic to make a 3D representation of the profile." Wait! Got you here…burning acrylic plastic produces carcinogenic fumes! I figure any big company that has to worry about OSHA compliance will not want that happening in their shop. But the reply is, "No big deal. We've got fans that vent it to the outside and the operator is safe." I'm thinking, you've got to be kidding! And from the little shops without fans, I hear that we only do it once in a while so the operator isn't in the fumes very often.

Does this mean you don't care about safety?

Let's talk technical, then. An acrylic block can't be made with only one laser pulse (if you are pulsed) or it has to be over a few seconds (if you are CW). For some reason, users think that the first laser pulse looks like the 100th pulse or the first millisecond CW power distribution looks the same as the power distribution at the end of the first second or at the end of the 2nd second.

But that is absolutely not true! Especially when the initial laser output is compared to having the laser on for a period of time. Let me at your laser with a camera or slit-based profiler and I'll show you.

This is in fact how we sell profilers. Users have perfect performing lasers…right up to the time we show them what they have with an on-site demo. This sells more profilers than any other method, except to industrial users. They see it and say, "What I've been doing of all these years seems to work, so why would I want to change?

Here's why. By knowing certain laser performance characteristics, you can do things you haven't been able to do until now:

• Predict the future, just like a machine vision system can. Knowing certain measurements about your lasers beam, periodically and over time, you will be able to predict, let me say that again, predict, when you will need periodic laser maintenance. Your down time can now be scheduled, not a catastrophic line shutdown.
• If you are using the same laser for multiple processes, you can optimize beam adjustments to each process.
• Over time, optics get dirty and dirty optics cause beam delivery and beam characteristics to change. The same laser parameters don't work with dirty optics. So you say, I change out the optics when this happens. Again, you are thinking binary. What about the subtle differences from day one to the change out day. How does that affect making consistent product? Beam delivery parameters subtly change over time. Right up to the time you change out the optics. Change is inevitable. You need to measure to know when that change reaches an unacceptable point.

The next objection I hear is that there might be something to this electronic monitoring of the beam quality, but they've heard that it is too expensive for the shop. Now we have it out in the open. Yet another objection we also heard about machine vision -- too expensive to make quality product. So we go through the ROI.

We take the value of the product, number made per unit of time, cost of scrap, how many do you make before knowing there is a problem, cost of unscheduled down time, etc. The standard ROI stuff. Our ROI calculator can quickly and easily predict what the payback time will be.

But let me be give you a starting point. If you have more than one laser in production and you are welding with it at least one shift and your part is precision in nature, the payback will be 16.7 months or less.

Too big, too small, doesn't quite fit, operators are used to the old process, we don't use PC's on the floor, too complicated, doesn't give the same information as what we are use to, and on and on. I've heard it all. Sigh.

Does this mean Dr. Deming's "measure it and control the process" isn't right for most manufacturers that use lasers? Definitely no. But manufacturers first have to open their eyes to this lost opportunity for improvement. Especially now, when manufacturing needs all the help it can get.