Integrating reliability into the design of fault-tolerant power electronics systems

This paper presents a methodology for integrating reliability considerations into the performance analysis carried out during the design of fault-tolerant power converters. The methodology relies on using a state-space representation of the power converter, based on averaging, similar to the ones used when analyzing linear time-invariant systems, and assumes an unknown-but-bounded uncertainty model for the converter uncontrolled inputs, such as load or variations in input voltage. The converter must be designed such that, for any uncontrolled input, the state variables remain within a region of the state space defined by performance requirements, e.g., output voltage tolerance or switch ratings. In the presence of component faults, and depending on the uncontrolled inputs, the converter may or may not meet performance requirements. Given the uncertain nature of these uncontrolled inputs, and for each particular fault, we introduce an analytical method to compute the probability that the performance requirements are met, which will define the reliability of the converter for each particular fault. By including these probabilities in a Markov reliability model, it is possible to obtain the overall converter reliability. The application of the methodology is illustrated with a case study of a fault-tolerant interleaved buck converter.