Improving Analog Functional Safety Using Data-Driven Anomaly Detection

Fei Su, Prashant Goteti

Abstract: Safety is a critical objective for automotive developments. Functional Safety of automotive analog and mixed-signal circuits faces several challenges; on the other hand, analog behavior provides an opportunity for early anomaly alert, thus improving functional safety. In this paper we propose a machine learning based methodology using data-driven anomaly detection for analog automotive circuits. The contribution of this work is to provide a framework of mining the dynamic in-field time series data in the context of system operation to detect anomalous events from analog functional safety perspective, with minimal hardware overhead. We present a realistic example to illustrate and analyze the proposed method. It presents an approach for improving functional safety of analog circuits in automotive applications.