The in-service reliability of power electronics modules during their normal operation in their work environment is a major concern for application developers. Failure mechanisms act on power modules limiting their lifetime and leading to unpredictable interruptions of power converter operation. That reduces the availability of power converters and can have large financial and safety implications in applications such as in wind turbines and railway traction. Therefore, many attempts are made to use Physics-of-Failure models to estimate the lifetime of power modules while in service utilizing the rainflow counting algorithm. However, large uncertainty in the lifetime estimate given by Physics-of-Failure methods limits the usefulness of that estimate and cannot help improving the availability of power converters. Condition Monitoring on the other hand provides information about the current health state of power modules based on online measurements of failure indicators that can be obtained from the power modules. This information can be used to inform the prognostics stage to provide an estimate of lifetime based on PoF models and online measurements in a Fusion-based approach such that uncertainty in the resulting lifetime estimate can be reduced. In this thesis, the main emphasis is to use online measurement data of failure indicators that can be obtained during the normal operation of power modules to infer the health status of the power module. Failure indicators such as the on-state voltage and junction temperature are estimated or measured online from the power converter. They are indicative of the two dominant failure mechanisms of power modules which are wire-bond lift-off and solder fatigue. Therefore, different simultaneous failure mechanisms can be discriminated. However, in order to infer the health information from the online measurement and discriminate between failure mechanisms, the measurement noise and the effects of operating conditions should be removed from the measurement. The approach proposed in this thesis is based on combining online measurements with pre-determined models of the power module in its original state. Comparing the online measurements with the models reveals the deviation of the power module from its original state. To achieve this, Kalman filter is used to estimate junction temperature based on a noisy estimate from a thermo-sensitive electrical parameter. In addition, measurement circuits are developed to realize the online measurements during normal operation of power modules. The health information inferred from the online measurement of failure indicators can be used in the future to estimate the remaining useful lifetime of the power modules and to inform the Physics-of-Failure models in a fusion framework in order to reduce the uncertainty in the lifetime estimates.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:689846 |
| Date | January 2016 |
| Creators | Eleffendi, Mohd Amir |
| Publisher | University of Nottingham |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://eprints.nottingham.ac.uk/32913/ |
Page generated in 0.0196 seconds