Location
Cocoa Beach, FL
Start Date
5-4-1965 8:00 AM
Description
Highly compressed time schedules, now routinely applied to the development of adyanced military and space systems, generate the need to arrive at management decisions without waiting for completion of detailed data-gathering activity. When a number of like units are placed on life test, it is advantageous to draw tentative conclusions, as soon as possible, about the distribution of time-to-failure. Several techniques are examined whereby preliminary rough estimates can be pulled out, early during -the test. These estimates are then modified in real time as the test continues and additional units fail.
These techniques hinge mostly upon the use of order statistics. The distribution-free case is considered first. Limitations inherent in all non-parametric statistics make it desirable to employ distributional methods also. Special attention is devoted to the exponential, rectangular, and logistic distributions. Of these, the last is a very good substitute for the normal distribution, possessing certain peculiar advantages of its own.
Real Time Estimation of Life-Distribution Parameters
Cocoa Beach, FL
Highly compressed time schedules, now routinely applied to the development of adyanced military and space systems, generate the need to arrive at management decisions without waiting for completion of detailed data-gathering activity. When a number of like units are placed on life test, it is advantageous to draw tentative conclusions, as soon as possible, about the distribution of time-to-failure. Several techniques are examined whereby preliminary rough estimates can be pulled out, early during -the test. These estimates are then modified in real time as the test continues and additional units fail.
These techniques hinge mostly upon the use of order statistics. The distribution-free case is considered first. Limitations inherent in all non-parametric statistics make it desirable to employ distributional methods also. Special attention is devoted to the exponential, rectangular, and logistic distributions. Of these, the last is a very good substitute for the normal distribution, possessing certain peculiar advantages of its own.
