In my last column, I presented my philosophy of troubleshooting oven heating systems -- a set of guiding principles that have worked over the years. This time, I'll dig into them a little deeper. By the way, if you haven't already noticed, these principles are general enough that they should apply to just about any kind of industrial machinery.
Divide and Conquer: Break the problem up into smaller and smaller pieces until you can isolate the cause.
Whether things are running right or going totally haywire, events follow certain prescribed sequences, and your first task is to find out what that sequence is. With systems like electrical controls, those sequences are built into the system, and a schematic provided with the panel may tell you all you need to know. If you're troubleshooting a process-quality problem, you're going to have to reconstruct the flow of process steps, and that may involve more than the oven. Pay special attention to those forks in the road, where the sequence takes a different path if B happens instead of A. If it helps, sketch the sequence out on paper.
Once you have your sequence map constructed, locate the point where the trouble first appears. The cause probably is somewhere between there and the start point. To reduce the possibilities, go back about halfway upstream and determine if the problem, or its source, is there. If it isn't, you've got to focus on the steps between that point and the place where the trouble appears. If it is there, you've got to go farther back to locate the source.
Once you've determined that the problem is originating within a certain block of steps, find the approximate midpoint of that block, and see whether the problem existed before or after that point. You're closing in now. Once again, break into the middle of the remaining suspect sequence of steps and look both ways for the problem or its cause. Keep doing this until you're down to a single step, and then you've apprehended the culprit.
Twenty yes-or-no questions to isolate one word out of 80,000? Sounds crazy, but here's where the power of Divide and Conquer really shows. For your first question ask, "Is the first letter of the word between A and M?" Boom! 40,000 possibilities eliminated, no matter which answer you're given. Let's say the answer is yes. Next you ask, "Is the first letter of the word between A and F?" You've just split the half in half again. If the answer is no, you know the letter lies between G and M, so break that string in half with your next question. Continue doing this till you've isolated the first letter. It should take no more than five questions. Repeat the drill with the second and third letters. You're now 15 questions into the game.
Take the dictionary and look at the list of entries beginning with the three letters you've identified -- usually, it's surprisingly short. Find out which letters straddle the midpoint of the list, and ask your "Is it between...?" question again. Continue splitting the list with the remaining questions. It rarely fails.
The point of this exercise is that if you can isolate one word out of 80,000 with 20 well-placed questions, you should be able to troubleshoot a process with a lot less. Very few processes have this many possibilities.
There are other uses for this exercise. If you've ever longed to be the life of the party with your wit and cleverness, you can troop it out, or if you're raising teenagers who are convinced their parents are terminally stupid, you can pop this one on them.No guarantees it will make you socially desirable or an icon of wisdom to your kids, but it sure does work for troubleshooting.