ARE YOU THE BLUNDERBUSS OR THE SHARPSHOOTER?

There have been times when I reflect on my previous life that I recall some of the techniques other engineers used to solve problems. I picture a "sharpshooter" as one who took dead aim at a problem then fired his high-powered rifle when the target was clearly in his sights. The engineer with the Blunderbuss shot from the hip with a strong dosage of buckshot as he aimed generally in the direction of the target. Perhaps both accomplished the same task; however, the blunderbuss didn't allow identifying which pellet(s) actually hit the target.

Troubleshooting can be a lot like that. For the person with the Blunderbuss, just start to turn the machine knobs and adjust like heck until the problem is solved. But then, what actually solved the problem? And what actually caused the problem?

There are many techniques used to methodically identify where the problems are and what to do about them. If you read any troubleshooting guide there is one fact that should jump out at you. I haven't found any problem that suggested just one solution. There's the rub. One needs to be smart enough to eliminate some of the possibilities and then home in on the solution. One helpful technique is the handy fishbone diagram. Using this tool one can follow a path toward possible solutions. Another more recent development is the Design of Experiments.

DOE is typically performed using computer muscle to determine what factors make the greatest impact on the process characteristics being studied. The basic approach is relatively simple. One should make one or a combination of machine adjustments and then record the output from the process. Sequences must be set up to perform the various "sets" of parameters until enough data is collected that the computer can determine which parameter has the greatest effect on improving the process. A major problem with this is that the ending computer description of the process can be very confusing and difficult to interpret. If one is lucky there will be one parameter that jumps out as the one to "correct". If not, there would be some trial and error involved - or another experiment executed.

Some textbooks attempt to show visually the results of a designed experiment. In many cases there are 3 dimensions applied to the data and still the data provide a cloudy picture. One technique for designing experiments comes from the Taguchi school of thought; however, I have seen much evidence that this technique doesn't provide the accuracy of a more detailed study, so you may not get as much steering.

Let's try picturing the earth. If the data produce essentially a globe worth of output and you destination is Australia, perhaps the inputs that result in an output of England is not where you want to be. Another set of data might put you in the center of the earth. Then, the best result might come from inputs where you wind up in New Zealand- not quite Australia, but pretty close. This is now your best starting point and if the process is not satisfactory, maybe another experiment is in order. Still you are way better off than where you started!

Jerry Golmanavich

Golubki@cox.net