Sensor Validation in Computer Systems
IC-AI '04 & MLMTA'04 , VOL 1 AND 2, PROCEEDINGS(2004)
摘要
A method is described for real time temperature-sensor operability validation to enhance the reliability, availability and serviceability (RAS) of high-end servers. The method can be used in any situation in which system's components are cooled using natural convection, forced airflow or liquid cooling. The method is based on indirect measurements with an approximately known measurement equation obtained using a data-driven inferential sensing technique. A thermal model of a component is derived using the thermal RC networks technique and transformed into a regression problem. Available measurements on the component already mounted and operating in the system are collected over a time period and used to estimate the regression model parameters. The regression problem in our case is ill-conditioned because it is a discretized version of the system identification problem which is inherently ill-posed. The analytical solution shape for the thermal model is used to regularize the solution of the regression problem. The obtained regularized measurement equation is used for the inferential measurement of the component's internal temperature and for sensor operability validation. Experimental results are presented for operability validation of temperature sensors in the components of high-end servers.
更多查看译文
关键词
reliability,availability and serviceability,high-end servers,regularization,indirect measurements,sensor validation,ill-posed problems
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络