You are on page 1of 3

IEEE TRANSACTIONS ON RELIABILITY, VOL. 49, NO.

1, MARCH 2000

85

Generic Rules to Evaluate System-Failure Frequency


Suprasad V. Amari, Member, IEEE
AbstractFrequency of failure of a system with -independent components can be obtained from the system availability (unavailability) expression and failure and repair rates of the components. Although, Grouped Variable Inversion is an efficient technique to find the system availability, there is no convenient method to convert the availability expression obtained by this technique into an expression for system-failure frequency. This paper present generic rules to find system-failure frequency, particularly, when the availability or unavailability expression of a system is obtained using this technique. The rules are straightforward, and produce appreciably shorter expressions for system-failure frequency. Examples illustrate the simplicity and efficiency of the proposed rules. Index TermsDisjoint product, sum of disjoint products, system availability, system-failure frequency.

of products of component availabilities or products of component unavailabilities, or their mixed products. These methods apply only when is obtained using the I-E method [7] or SVI (a disjoint product method) [1]. Recent literature on reliability or availability evaluation of systems with -independent components [3], [4], [8] shows that GVI is efficient in terms of computation time and accuracy as well as in getting appreciably fewer disjoint products. Although, GVI is efficient for finding , there is no convenient method in this case to convert into . This is because, for GVI the disjoint products involve the complements of products of component availabilities or complements of products of component unavailabilities or both. Therefore, this paper proposes generic rules to convert obtained through a GVI technique into an expression of . II. STEADY-STATE FREQUENCY OF FAILURE

I. INTRODUCTION Acronyms GVI SVI Notation , label of component [failure, repair] rate of component system-failure frequency mean number of system-failures per time-unit Boolean variable indicating that component is is the complement of [working, failed]; steady state [availability, unavailability] of component ; Pr component is [working, failed] grouped variable inversion (method) single variable inversion (method) A. Assumptions 1) The system is composed of -independent components. Its structure function is -coherent. 2) The system and its components have 2 states: up (working) and down (failed). 3) The failure and repair rates of the components are constant. 4) All components are repairable. Repaired components are good-as-new. 5) The system is in steady-state. Example 1 demonstrates the evaluation of using SVI and GVI; the number of terms using SVI is more than in using GVI [3]. B. Example 1 [3] Let: , system [availability, unavailability] and are used in the paper; For convenience, both . however, note that is a key parameter in reliability and risk analysis; it can be evaluated from a or [5], and failure and repair rates of its components; usually is used, but sometimes is used in this context. Other performance measures of a system, such as mean up-time, mean down-time, and mean cycle-time can be determined easily from , , of the system [7]. The methods available [5], [6] to convert into a function of are limited to the case where is expressed as the sum
Manuscript received November 1, 1999. The author is with the Reliability Engineering Centre; Department of Industrial Engineering and Management; Indian Institute of Technology; Kharagpur 721302 India (e-mail: suprasada@pune.tcs.co.in). Responsible editor: S. Rai. Publisher Item Identifier S 0018-9529(00)06206-0. 00189529/00$10.00 2000 IEEE

, ,

be the minimal path sets of a system. 1) SVI:

(1) 2) GVI:

(2)

86

IEEE TRANSACTIONS ON RELIABILITY, VOL. 49, NO. 1, MARCH 2000

For (2), there is no direct method to get from . Following simple rules, (2) can be expressed as in either (1) or (3).

(3) is Ref. [5], [6] have the rules for calculating when or and . The rules are expressed in the mixed products of the labeled I and II. is expressed in terms of mixed products Case 1: and Rule I: To get from , multiply every term by ; and treat constants as if they were not present. (I) is expressed in terms of mixed products Case 2: and Rule II: To get from , multiply every term by ; and treat constants as if they were not present. (II) Hence, from (3), using rule I, for example 1:

Case 5: is expressed in terms of mixed products and Rule V: To get from , multiply every term by . (V) is expressed in terms of mixed products Case 6: and Rule VI: To get from , multiply every term by . (VI) Rule III (IV) applies when ( ) is obtained using path-sets, whereas rule V (VI) applies when ( ) is obtained using cutsets. In general (say, for fault-trees), or can be obtained as , , , , the sum of mixed products of and constant coefficients. If there exist -identical components/modules, then there can exist repeated sub-products within a product, i.e., there can exist higher powers of sub-products within a product (e.g., a seriesparallel system with -identical units in each subsystem). Therefore, the more generic rules are: is expressed in terms of mixed products Case 7: , , , , and a constant . Rule VII: To get from , multiply every term by . (VII) is expressed in terms of mixed products , , , , and a constant . from , multiply every term Rule VIII: To get Case 8: by . (VIII) In all cases, the constants should be treated as if they were not present. However, the constant coefficients remain in their corresponding positions as shown in rules VII and VIII. For example, consider:

(4) Both (3) and (4) have 7 terms. Similarly, these rules were applied to (1) to get ; there were 9 terms. Since, rules I and II can not deal with the products of inversion of several variables, contains more terms. Section III provides simple, easy-to-remember rules to find from , using GVI. III. THE GENERIC RULES To improve readability, , , , are used in the , , , , respectively. The proof rules for is based on mathematical induction, and is available from the author. Let:

the constant to be ignored is 1. Hence

i.e.,

For example 1,

can be found from (2), using rule III: (6)

i.e., (5) is expressed in terms of mixed products Case 3: and Rule III: To get from , multiply every term by . (III) is expressed in terms of mixed products Case 4: and Rule IV: To get from , multiply every term by . (IV)

After some mathematical manipulation, the expression in (6) can be expanded in the form of (4) [2], since they are identical. is given in (7). The fault-tree of the system is Example 2: described in [2]. (7) Using rule VIII

(8)

AMARI: GENERIC RULES TO EVALUATE SYSTEM-FAILURE FREQUENCY

87

The number of terms (products) in equals the number of terms in the availability (unavailability) expression. Therefore, it is straightforward to prove that rules IIIVIII will produce fewer terms, because these rules can take expressions of sum of products of mixed polynomials as input, and this kind of expression contains fewer terms. Since the computation time is proportional to the number of terms, these rules produces fewer terms and require less computation resources. ACKNOWLEDGMENT The author would like to thank the Dr. S. Rai, Editor; an anonymous referee; and Dr. W. Schneeweiss for their useful comments on earlier versions of this paper. REFERENCES
[1] J. A. Abraham, An improved algorithm for network reliability, IEEE Trans. Reliability, vol. R-28, pp. 5861, Apr. 1979. [2] S. V. Amari, Reliability, Risk and Fault-Tolerance of Complex Systems, Ph.D. dissertation, Indian Institute of Technology, Kharagpur, 1998.

[3] K. D. Heidtmann, Smaller sums of disjoint products by subproduct inversion, IEEE Trans. Reliability, vol. 38, pp. 305311, Aug. 1989. [4] S. Rai, M. Veeraraghavan, and K. S. Trivedi, A survey of efficient reliability computation using disjoint products approach, Networks, vol. 25, pp. 147163, 1995. [5] W. G. Schneeweiss, Addendum to: Computing failure frequency via mixed products of availabilities and unavailabilities, IEEE Trans. Reliability, vol. R-32, pp. 461462, Dec. 1983. [6] D. Shi, General formula for calculating the steady-state frequency of system failure, IEEE Trans. Reliability, vol. R-30, pp. 444447, 1981. [7] C. Singh and R. Billinton, A new method to determine the failure frequency of a complex system, IEEE Trans. Reliability, vol. R-23, pp. 231234, 1974. [8] S. Soh and S. Rai, CAREL: Computer Aided Reliability Evaluation for distributed computing networks, IEEE Trans. Parallel and Distributed Systems, pp. 199213, Apr. 1999.

Suprasad V. Amari For biography, see IEEE TRANS. RELIABILITY, vol. 46, Dec. 1997, p. 522.

You might also like