On nonlinear reliability growth

Test-analyze-and-fix is the most commonly recognized method of improving equipment reliability. The author considers the evaluation method used to measure the effectiveness of design fixes implemented during test. A nonmathematical, practical approach is offered to evaluate nonlinear growth as might be observed during an actual reliability growth test. Graphical illustrations show how the AMSAA (Army Systems Analysis Activity) and Duane least squares growth analysis techniques deal with nonlinear growth data. It is pointed out that both analyses are insensitive to discontinuities and/or sudden changes in growth slopes and rely heavily on early test data. A piecewise regression method is offered to overcome the shortcomings of the Duane regression and AMSAA methods. A case-history approach is used to analyze both hypothetical and actual data representing a wide variety of growth curve shapes