TRIZ (pronounced TREES) is a Russian technique for problem-solving. The theory was developed from extensive research covering hundreds of thousands of inventions across many different fields to produce a theory which defines generalized patterns in the nature of inventive solutions and the distinguishing characteristics of the problems that these inventions had overcome.
The u chart is used with a varying sample size where you are counting the number of defects in the sample. If you remember, the difference between a defect and a defective is this. A defect can be found on an otherwise acceptable product; whereas, a defective means that the whole item is unacceptable. For example, you might have some minor defects in the fabrication of the windshield. The windshield passes inspection even though there are some minor defects. Defects are still undesirable, but there are an insufficient number of defects to render the windshield unusable. If, however, the windshield is cracked, the windshield would be rendered defective, and totally unusable. With the u-chart, we are looking at defects – not defectives, and remember that were looking at a varying (as opposed to constant) sample size. If it was a constant sample size and we were evaluating defects, we would instead be using a c-chart.
Weibull analysis of failure data using the Weibull distribution is the most often used technique in reliability engineering.
The Weibull distribution can be used to model many different failure patterns and this, along with its relative ease of use, accounts for its extensive application.
There are two ways to make a bad part or make for an unhappy customer. First, if the centering of the variation in a process gets too close to either the upper specification limit or the lower specification limit, a bad part will be produced, or the customer will be unhappy. Secondly, if the variation in the process is centered, but the variation in the process is too wide, no matter how the team tries to center the variation, a bad part will be produced. The process variation overlaps either the upper specification limit or the lower specification limit. The X-bar R Chart monitors the process across time and reveals assignable-cause variation, namely looking out for these two conditions as they arise. The other value of using this chart is that if there is no assignable-cause variation, the process is to be left alone. If adjustments are made to the centering on a process that is exhibiting common-cause variation, more variation will be generated than it would had the process just been left alone. In other words, if the process is in statistical control, leave it alone. For more information, please view the lectures for both types of charts.
The XmR Chart is very similar to the X-bar R Chart in the rationale for its use. The main difference is that the XmR Chart uses individual data points; whereas, …
The Z-score table (also known as the standard normal distribution) is used for larger sample sizes, namely where the population or sample size is greater than n=30. The Z score table assumes a normally-distributed process.
The Z score table is used in various hypothesis tests including tests on single means, the difference between two means, and tests on proportions.
Because processes will randomly fluctuate over time and will often deteriorate in the long run, Motorola introduced the practice of reducing their calculated sigma values by 1.5 standard deviations to provide a more conservative estimate of the long term performance of the process.
