The traditional NHPP (non-homogeneous Poisson process) models are proved to be a success in a practical test. However, the model performance always suffers in the realistic software testing environment due to the ideal assumption which derived the traditional NHPP models, such as constant fault detection rate and smooth or regular changes. In this paper, an NHPP-based software reliability growth model is proposed considering an irregular fluctuation of a fault detection rate, which is more in line with the actual software testing process. The fitting and predictive power of the proposed model is validated using the related experiments. The experimental results show the proposed model has a better fitting and predicting performance than the traditional NHPP-based models using the real-world fault data. Meanwhile, the confidence interval is given for the confidence analyses of the proposed model.