Almost all systems install a protective device to prevent shocks which occur suddenly from outside or inside. A typical example is transformers with a surge absorber and computer systems with a C.V.C.F. device. It is extremely serious If the device has failed when a shock occurs. To avoid such an unfavourable situation, we need to check whether the device is good or not. This paper considers a system with the device which fails by shock or by myself. The device has a failure time distribution F(t), and shocks occur o;t randomly in time, i.e., according to an exponential distribution (1 - e^<-αt> ) . If the device is good, then it prevents shocks with probability 1 - p, other-wise a system becomes failure by any shock. We inspect the device at periodic times kl (k = 1, 2, ...) under these assumptions. Costs c_l and c_2 are suffered for failures of both system and device and of only device, respectively, and c_3 is for one inspection. Then, using the usual calculus method of probability, the expected cost rate is [numerical formula] where F^^- ≡ 1 - F. It is difficult to compute an optimum inspection time l* which minimizes C(l;p) . In particular case of F(t) = I - e^<-λt>, we discuss an optimum inspection policy for 0 ≦ p ≦ I . We finally show figures which give an optimum time l* and its effect [C(∞;p) - C(l*;p)]/C(∞;p), for two cases where the mean failure time (1/λ) of the device and the mean interval (1/α) of shocks are changed.
View full abstract