Total system reliability: Integrated modelfor growth and test termination

Reliability demonstration testing is not the most efficient method of assuring product reliability prior to shipment. It is costly, time consuming and has inherent technical and social limitations. The dilemma facing the reliability and quality engineer is whether to continue demonstration testing and risk shipping a product late or ship the product and risk warranty and field service returns. Either option can cause the company to lose significant market share and profit. This paper sets out to solve this dilemma by meeting both the time to market constraints and the product reliability goals. The weaknesses of existing reliability demonstration techniques are explored and a comprehensive methodology is introduced involving controlled development processes, stress testing, root cause determination and process change feedback mechanisms. All prototype products are manufactured on the final volume process line resulting in the early identification and correction of process-related problems. Testing commences on the first available prototypes with system stress/robust testing being employed to stimulate failures, determine their root cause and correct them. Reliability growth modelling assesses the ongoing improvements occurring in reliability during the development cycle, while a statistical stopping rule is used to determine the optimal product release time without risking product warranty. The approach is applicable to systems incorporating both hardware and software elements. The methodology has been validated on three development projects of telecommunication systems comprising hardware and software. In addition to enhancing team behaviour and performance, the development times have been reduced by 14% and the ramp-up time to full worldwide product shipments has been reduced by 50%. Copyright 2005 John Wiley & Sons, Ltd.