Software reliability growth model with bell-shaped fault detection rate function

Software reliability growth models (SRGMs) based on the non-homogeneous Poisson process (NHPP) are quite successful tools which have been proposed to assess the reliability of software. Among various SRGMs, two most important factors which affect the accuracy of reliability evaluation and prediction are the number of initial faults and the fault detection rate (FDR). Some SRGMs based on NHPP assume a constant FDR, and some others assume a increasing or decreasing FDR function as time passes. Those assumptions ignore either the learning process of the software testers or the fact that failures removed first have higher detected rate. A bell-shaped FDR function is proposed which integrates both learning phenomenon and inherent FDR. A NHPP SRGM called Bbell-SRGM is put forward which incorporates the proposed FDR function. Bbell-SRGM is evaluated using a set of software failure data. The results show that Bbell-SRGM fits the failure data better than G-O model and some other SRGMs.