Abstract: Let a system A be composed of K independent subsystems B₁, B₂,…,B_K in parallel (or in series), and the i th subsystem be composed of m_i-independent success failure units C _i1, C_i2,…,C_imi  in series (or in parallel). There exist hierarchical test data as follows:①for unit C_ij , the numbers of tests, successes and failures are N_ij, S_ij  and F_ij respectively (i=1,2,…,K; j=1,2,…,m_i);②for subsystem B_i, the numbers of tests, successes and failures are N_i, S_i, and F_i respectively (i=1,2,…,K);③for system A, the numbers of tests, successes and failures are N, S, and F respectively. Using the hierarchical success failure test data, the authors present an information theory method for calculating the approximate confidence limit of system reliability in this paper. The equivalent system success failure data (N_*, S_*, F_*) are obtained from the hierarchical test data by means of the principle of equal information quantity. Then, the approximate confidence limit of the system reliability is obtained from the equivalent system success failure data (N_*, S_*, F_*) according to the exact classical method of the individual success failure unit. Finally, the authors analyze the application of the formulae proposed in the paper by way of examples.

Key words: success- failur e unit, par allel-ser ies system, ser ies-par allel system, information quantit y, r eliabilityconfidence lower limit