single data point, there is no rational way to know whether this component has been tested for longer than MTTF or shorter. Therefore, the engineer is entitled to only 50% confidence that a second component will last for the same time or longer. If we define X as the value of A such that the probability of operability is equal to 1 ~PC, or Thus, a single test for 7 years without failure allows us to have 98% confidence that the MTTF is 1.24 years or longer. To have 98% confidence that MTTF is 7 years or greater, 39.5 years of testing must be done without a failure. For cases in which failures do occur during life testing, the situation of course gets worse. The general formula relating A to PQ takes the form of the Poisson distribution: where n is the number of failures which occur during testing. An alternative to severe component reliability and confidence requirements is to reduce the component requirements through modularity and redundancy. If a system consists of two identical components where only one is required to function to ensure operability, the system reliability is For 2?s = O.9O, Rc is only 0.68, dramatically reducing the component reliability requirement with the addition of only a single redundancy. For the general case in which there are n identical components with a reliability R and a failure probability of F=l— R, then the probability of r successes (or n~r failures) is equal to Discussion The case of thermionic reactors is interesting because two types of redundancy are required. In the case of short-circuit failure (caused by a short between emitter and collector, for example), a single failure will remove only one cell from the unit. However, in the case of open-circuit failure (caused by loss of cesium), every
RkJQdWJsaXNoZXIy MTU5NjU0Mg==