The Pham Nordmann Zhang (PNZ) Software Reliability Model Revisited

Khalid Edris and Omar Shatnawi


Software testing, software reliability, fault generation, imperfect debugging, software engineering


The Pham Nordmann Zhang (PNZ) software reliability model [12] is revisited and some research directions are further discussed. The PNZ model assumed that on a failure, the fault causing the failure is removed with certainty. In reality this may always not be true. In this paper, a newly developed continuous SRGM with two types of imperfect debugging and learning process of the testing team as testing progresses is proposed. The first type, known as fault generation, describes the situation when each fault removal attempt increases the fault content of the software. The second type, less damaging, is the case of imperfect debugging where all detected faults are not removed completely. Here the numbers of removal attempts are more than actual fault content but imperfect debugging does not change the content of faults in the software. The concept of learning has been incorporated in the fault removal rate to show the gain in experience and improvement in the testing efficiency of the team as the testing grows. To model learning, fault removal rate has been taken as logistic function. The proposed model has been validated and compared with well-established existing NHPP models by applying them on two real fault removal datasets. The results are encouraging in terms of goodness of fit criteria, predictive validity criterion, and software reliability evaluation measures for software reliability data due to its applicability and flexibility. A discrete version of the proposed model has also been presented.

Important Links:

Go Back