IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 32, NO. 3, SEPTEMBER 2009 667 Baseline Performance of Notebook Computers Under Various Environmental and Usage Conditions for Prognostics Sachin Kumar, Student Member, IEEE, and Michael Pecht, Fellow, IEEE Abstract —This paper presents an approach for electronic product characterization. A study was conducted to formally characterize notebook computer performance under various en- vironmental and usage conditions. An experiment was conducted on a set of ten notebook computers to evaluate the variations, trends, and correlations in performance parameters, as well as the extreme value these parameters can attain in various environmental and usage conditions. Software was developed to periodically retrieve information on the performance parame- ters [i.e., central processing unit (CPU) temperature, videocard temperature, motherboard temperature, %CPU usage, %CPU throttle, CPU states C1, C2, and C3, and fan operation] from each computer’s basic input–output system. An automated software script was written to simulate user activity. The variability in the performance parameters was identified and the empirical relationships among parameters were established. Empirical equations can be used to calculate the expected value of the performance parameters. A system’s deviation from normal behavior can be identified by performing a comparison between performance parameters’ expected and an observed value. The effect of environmental factors, including different power sources, ambient temperatures, humidity, and use conditions on the percentage CPU usage was studied. Index Terms—Electronic prognostics, health monitoring, note- book computer, reliability, usage monitoring. I. Introduction Prognostics involves the assessment of a system’s actual health condition, followed by modeling fault progression, predicting its performance, and determining remaining useful life. The results of diagnostic methods often provide the ground for prognostics. Diagnostic techniques for a system are based on observational data from that system and its operating Manuscript received February 14, 2008; revised June 4, 2009. Current version published August 26, 2009. This work was supported by the Prog- nostic and Health Management Consortium, Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park, MD, and the Prognostic and Health Management Center at City University of Hong Kong, Hong Kong, China. This work was recommended for publication by Associate Editor B. Sammakia upon evaluation of the reviewers comments. S. Kumar is with the Center for Advanced Life Cycle Engineering, University of Maryland, College Park, MD 20742 USA (e-mail: skumar@calce.umd.edu). M. Pecht is with the Center for Advanced Life Cycle Engineering, Univer- sity of Maryland, College Park, MD 20742 USA, and also with the Prognostics and Health Management Center, City University of Hong Kong, Hong Kong, China (e-mail: pecht@calce.umd.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCAPT.2009.2026890 environment, whereas prognostic techniques require historical performance data, system knowledge, a profile of future usage, and an understanding of the operating environment conditions. Traditional reliability measures provide confidence that a product is going to serve its intended purpose for a certain period under specified operating limits [1]. However, they do not take into account the unforeseen changes in operating envi- ronment conditions or operating load [2]. The accuracy of any reliability prediction depends on the model defined for the sys- tem under specific environmental and usage conditions [1], [3]. Prognostics models consider the operating environmental and usage conditions in defining the prognostic distance for a product or system before it encounters a fault. This early warning may provide enough time for management to take necessary action before failure. Models assist in identifying fault progression, diagnosing faults, and predicting failures. The output of prognostic models helps the system operating team to make intelligent, informed, and appropriate decisions about logistical actions based on available resources and operational demand [3], [4]. For years, diagnostic and prognostic concepts and methods have been used for mechanical systems, but prognostics termi- nology is relatively new for electronic products and systems, and interest has been growing in predicting failures and pro- viding warnings in advance of catastrophes, especially since the release of U.S. DoD Policy 5000.2 [5] in 2004. The new policy states that the top design priorities for the development or acquisition of new weapons systems and end items are embedded diagnostics and prognostics and equipment/system health management capability. Electronic products can be monitored by assessing their performance indices [6], [7]. These indices from a pristine system can define the baseline performance of that system, and the baseline can be used later for identifying degradation or failures. Early detection of a problem based on baseline performance will allow preventative action to be taken in order to avoid problems. The definition of reliability warrants that a product must perform its intended functionality to be considered reliable under stated conditions. The functionality of a product can be assessed by monitoring its performance parameters. Build- ing knowledge of the performance parameters’ variability is essential in order to make informed reliability decisions [8]. To perform prognostics for an electronic product it is 1521-3331/$26.00 c  2009 IEEE Authorized licensed use limited to: University of Maryland College Park. Downloaded on September 10, 2009 at 13:02 from IEEE Xplore. Restrictions apply.