Symbolic Dynamics and Analysis of Time Series Data for Diagnostics of a dc-dc Forward Converter

Abstract

This paper presents a novel approach to diagnosis of dc-dc converters with application to prognosis. The methodology is based on Symbolic Dynamics and Diagnostics. The data derived method builds a statistical baseline of the converter that is used to compare future statistical models of the converter as it degrades. Methods to determine the partitioning and number of partitions for the Symbolic Dynamics algorithm are discussed. In addition, a failure analysis is performed on a dc-dc forward converter to identify components with a high probability of failure. These components are then chosen to be monitored during accelerated testing of the dc-dc forward converter. The methodology is experimentally validated with data recorded from two dc-dc converters under accelerated life testing.