An Evaluation of Three Reliability Growth Models.
NAVAL POSTGRADUATE SCHOOL MONTEREY CALIF
Pagination or Media Count:
This thesis presents an evaluation of three relatively simple reliability growth models for which accuracy, precision, and robustness performance were examined over a wide variety of true underlying reliability growth patterns. A continuous cumulative failure rate model, a continuous instantaneous failure rate model, and a discrete reliability model, each of which employ ordinary regression methods, were evaluated using standard computer Monte Carlo simulation techniques. Simple, straightforward statistical measures of performance are exhibited in graphical and tabular form. All the models displayed some degree of difficulty in tracking particular types or portions of anomalous reliability growth patterns. The cumulative model displayed this difficulty the least and exhibited good variability precision performance providing confidence in its use. The instantaneous model, while displaying generally good accuracy, exhibited poor variability performance. Except for a couple of anomalous situations, the discrete model showed good accuracy and variability performance. Forecasting performance of all the models proved to be worse than their capability to determine current reliability status. Author
- Statistics and Probability