Pergamon Microelectron. Reliab., Vol. 35, No. 4, pp. 637-656, 1995 Copyright © 1995 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0026-2714/95 $9.50+.00 ENVIRONMENTAL STRESS SCREENING STRATEGIES FOR COMPLEX SYSTEMS: A 3-LEVEL MIXED DISTRIBUTION MODEL EDWARD A. POHL 1 and DUANE L. DIETRICH 2 1Air Force Institute of Technology, Dayton, OH 45433 and 2University of Arizona, Tucson, AZ 85721 (Received for publication 24 June 1994) ABSTRACT- Environmental Stress Screening (ESS) is employed to reduce, if not eliminate, the occurrence of early field failures. In this paper, a three level ESS model is presented for a complex electronic system. Screening is performed at the component, board and system level. Components are screened for a specified duration before being assembled into printed circuit boards (PCBs). Defects introduced during the assembly of the PCBs are screened at the board level, while defects introduced during final assembly are screened at the system level. Components and connections are assumed to come from good and substandard populations and their times-to-failure distributions are modeled by mixed distributions. Mixed exponential distributions are used to model component times- to-failure and mixed Weibull distributions are used to model the times-to failure for board and system level connections. The mixed Weibuli distributions are used to model wear- out characteristics at the board and system level. Optimal screen durations in the presence of wear-out are obtained by minimizing the systems life-cycle cost. ESS is shown to be a cost effective strategy when properly implemented. The optimal screening strategies are shown to be relatively robust to the system warranty period. 1. INTRODUCTION Today's competitivemarket forcescompanies to spend considerableeffortand time improving the qualityand reliability of theirproducts. M u c h of the effortisdevoted to identifying design risks, failure mechanisms, and evaluatingdesign alternatives.Despite the extensiveup frontefforts put forthby companies, many products fall short of their "designed-in"reliability goals. Sincemajor businessdecisionsimpacting corporate profits and performance are made on the basisof the failure propertiesof a product itis imperativethatthe product achieveits"designed-in" value. 637