Fault detection of CPU chips using an interval estimation method

Accuracy of thermal measurements directly reflects the performance of CPU chips. This paper proposes a fault detection scheme based on the Markov jump heat equation for the CPU chips. Specifically, this scheme utilizes one-dimensional Markov jump heat equation to model the thermal behavior over the chip. And, a partial differential equation based interval filter is proposed to construct the thresholds for measured temperature. Moreover, the detection residuals are not only robust to uncertainties, but also sensitive to faults. Furthermore, a novel $L_{\infty }$ norm-based uncertainty attenuation performance is presented to describe the effect of uncertainties. Numerical example has demonstrated the effectiveness of the proposed fault detection scheme.