Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com Reliability Characteristics for Two Subsystems in Series or Parallel or n Subsystems in m_out_of_n Arrangement (by Don L. Lin) 1 Introduction This is for reliability of two subsystems in series or in parallel. Also discussed is the system consists of n identical subsystems in parallel, the system is declared as failed if m or more subsystems fail (the m_out_n case). The system characteristics such as “system failure rate”, “system Mean Time Between Failure”, “system availability and unavailability”, and “system mean down time” are derived. 2 Connection in series This section describes a system consists of two (non-identical) subsystems in series. 2.1 System Failure Rate For just one subsystem, the failure rate is λ1. The probability of failure in dt is λ1dt. For two subsystems in series, the probability of failure in dt is (λ1dt + λ2dt ). The system failure rate is thus (λ1 + λ2 ). λseries = λ1 + λ2 The reliability function is R(t)=exp[-(λ1 + λ2)t]. 2.2 System MTBF From the exponential form of the reliability function, it is obvious that MTBFseries=1/(λ1 + λ2 )= MTBF1 MTBF2 MTBF1 MTBF2 2.3 System Availability and unavailability For the system to be available, each subsystem should be available. Thus, Aseries A1 A2 687288856 Page 1 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com Conversely, the unavailability is UAseries 1 Aseries 1 (1 UA1 ) (1 UA2 ) UA1 UA2 UA1 UA2 2.4 System Mean Down Time for Repairable subsystems If two subsystems are both repairable, one with mean down time MDT1 and the other MDT2, what is the mean down time for the two subsystems in series? At any instance in time, the system is in one of the 4 states: both subsystems functional, only subsystem #1 is non-functional, only subsystem #2 is non-functional, both subsystems are non-functional. The last 3 cases are responsible for the system being non-functional. It is assumed that the 4th case has negligible probability. Given the system is down, what is the probability that it is because the subsystem #1 is non-functional? It is obviously 1 1 2 . Since subsystem #1 needs MDT1 to repair, the repair time associated with repairing subsystem #1 is then 1 1 2 * MDT1 A similar expression is true for subsystem #2. Summing them up, one gets MDT series MTBF1 MDT2 MTBF2 MDT1 MTBF1 MTBF2 3 Connection is Parallel Here the two subsystems are repairable. The mean down times are MDT1 and MDT2. 3.1 System Failure Rate If the system just consists of subsystem #1, then the system failure rate is λ1. The probability of failure in dt is λ1dt. Adding subsystem #2 in parallel, the probability for system failure in dt is λ1dt scaled down by the probability that the subsystem #2 687288856 Page 2 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com is in the failure state. The probability to find the subsystem #2 in the failure state is MDT2 1 given by . Assuming MDT2 MTBF2 and using MTBF2 , 2 MTBF2 MDT2 the scaled down failure rate for subsystem #1 is then given by 1 2 MDT2 . Likewise, the scaled down failure rate for subsystem #2 is 1 2 MDT1 . Consequently, parallel 1 2 ( MDT1 MDT 2 ) 3.2 System MTBF Taking the approach that the inverse of the failure rate is MTBF (true for exponential distribution), one gets MTBF1 MTBF2 MTBFparallel=1/λparallel= MDT1 MDT2 It is noted that if the two subsystems are not repairable, then the MTBF for the parallel case is the sum of the individual MTBF’s. 3.3 System Availability and unavailability For the system to be available, either subsystem should be available. Thus, Aparallel A1 A2 A1 A2 Conversely, the unavailability is UAparallel 1 A parallel 1 ( A1 A2 A1 A2 ) (1 A1 ) (1 A2 ) UA1 UA2 (1) 3.4 System Mean Down Time for Repairable subsystems From the definition of Unavailability MDT MDT MTBF MDT MTBF one can get the MDT for the parallel case by using Eq.(1) above. UAparallel 687288856 MDT parallel MTBF parallel MDT parallel MTBF1 MTBF 2 MDT1 MDT 2 Page 3 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com UA1 UA2 MDT1 MDT2 MTBF1 MTBF2 Consequently, MDT parallel MDT1 MDT2 MDT1 MDT2 4 M out of N Parallel Subsystems If a system consists of n parallel, identical subsystems and the system is down if there are m or more subsystems down, what are the formulas for system failure rate, system MTBF, system availability, and system mean down time? 4.1 System Failure Rate If the system just consists of subsystem #1, then the system failure rate is λ. The probability of failure in dt is λdt. To have a system failure, we need to have other (m1) subsystems in the failure state. The chance that any one subsystems is in the failure state is given by MDT/(MTBF+MDT), or (MDT/MTBF), if we assume MDT<< MTBF. To find (m-1) subsystems in the failure state, the probability is MDT m 1 ( ) . There are n 1 C m1 ways to group (m-1) subsystems out of (n-1) MTBF subsystems. Also, we can choose any subsystem to be the #1 subsystem in the analysis. Putting all together, one has m _ out _ of _ n ( MDT m1 n! ) n1 C m1 n m MDT m1 (2) MTBF (n m)!(m 1)! This is the failure rate for exactly m subsystem failures. The failure rate for more than m subsystem failures is going to be smaller by a factor of ( MDT ) . For a consistency check, we consider n=m=2. This is a system consisting of two parallel, identical subsystems. When m=2 subsystems fail, the system fails. This is what discussed in Section 2. And Eq.(2) for this case is system 2 (2 MDT ) which agrees with the formula in Section 2. 687288856 Page 4 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com 4.2 System MTBF Taking the approach that the inverse of the failure rate is MTBF (true for exponential distribution), one gets MTBF m MTBFm_out_of_n=1/λm_out_of_n= n! MDT m 1 (n m)!(m 1)! . 4.3 System Availability and unavailability For the system to be available, at least (n-m+1) subsystems should be available. Thus, Am _ out _ of _ n n n! A i (1 A) n i i n m 1 ( n i )!i! Using the following equality, n 1 [ A (1 A)] n i 0 n! Ai (1 A) n i (n i )!i! we can rewrite the availability as nm Am _ out _ of _ n 1 i 0 n! n! A i (1 A) n i 1 (1 A) m (n i )!i! m!(n m)! And the unavailability is given (again, for MDT<<MTBF), by UAm _ out _ of _ n n! UAm m!(n m)! 4.4 System Mean Down Time for Repairable subsystems From the definition of UAm _ out _ of _ n 687288856 MDT m _ out _ of _ n MTBFm _ out _ of _ n Page 5 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com one can get the MDT for the m_out_of_n case by using in Sections 4.2 and 4.3 for UAm_out_of_n and MTBFm_out_of_n. Consequently, MDT m _ out _ of _ n 687288856 MDT m Page 6 of 7 Aurora Consulting Engineering LLC 609 Highview Ct, Neshanic Station, NJ 08853 (908) 369-3297 www.AuroraConsultingEngineering.com 5 All Formulas in One Table Two subsystems In Series ( is failure rate ) Two subsystems In Parallel System Failure Rate λseries = λ1 + λ2 parallel 1 2 ( MDT1 MDT 2 ) System MTBF MTBFseries= System Availability (A) MTBF1 MTBF2 MTBF1 MTBF2 Aseries A1 A2 System Unavail- UAseries UA1 UA2 UA1 UA2 ability (UA) System Mean MDTseries= Down MTBF1 MDT2 MTBF2 MDT1 Time MTBF1 MTBF2 (MDT) 687288856 MTBFparallel= MTBF1 MTBF2 MDT1 MDT2 Aparallel A1 A2 A1 A2 m _ out _ of _ n n! m MDT m1 (n m)!(m 1)! MTBFm _ out _ of _ n Am _ out _ of _ n 1 UAparallel UA1 UA2 MDT parallel n identical subsystems in parallel; system fails if m or more subsystems fail. (m_out_of_n) UAm _ out _ of _ n MDT1 MDT2 MDT1 MDT2 n! (1 A) m m!(n m)! n! UAm m!(n m)! MDT m _ out _ of _ n Page 7 of 7 MTBF m n! MDT m 1 (n m)!(m 1)! MDT m