Time-domain Parameter Estimation for Fault Identification in a Current-source Push-pull Parallel Resonant Inverter
Open Access
- Author:
- Lovell, Terrance Daniel
- Graduate Program:
- Electrical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 20, 2012
- Committee Members:
- Jeffrey Scott Mayer, Thesis Advisor/Co-Advisor
- Keywords:
- Parameter estimation
Inverter
Current-Source
Push-Pull - Abstract:
- The Current-Source Push-Pull Parallel Resonant Inverter (CSPPPRI) is a dc-to-ac power converter that produces a nearly pure sine wave ac voltage, making it well suited for avionics applications. In such applications, the reliability of the converter is paramount. Thus, a method of detecting incipient failures would be of significant value. In this thesis, a method for detecting incipient failures in the CSPPPRI is developed. This method uses least-squares parameter estimation (LSE) to detect changes in circuit parameter values by analyzing time series data for converter input and output variables collected by a data acquisition system. The time series data is used in conjunction with circuit models for each of the major sub-circuits of the CSPPPRI to estimate the internal circuit parameter values, which may change due to aging or stress. As part of the parameter estimation process, a virtualized ac waveform is created by determining the internal switch states. A commercial CSPPPRI was subjected to three test conditions: nominal output capacitance and load, reduced output capacitance (to simulate degradation of circuit components) and nominal load, and nominal output capacitance and increased load (to test for load dependence of the algorithm). For each test condition, key circuit parameters were estimated. The estimated parameters indicated changes in the underlying component values, but unfortunately the indicated changes were dependent on the loading of the inverter. Alternative susceptance (originally thought to be reactance) calculations made using the switching frequency were added to identify whether changes of estimated circuit parameters were due to seeded faults (reduced output capacitance) or loading effects (increased load). Changes due to loads affect the calculated susceptances equally, while changes due to circuit parameter variation would not. Thus the divergence of the susceptance values can be used to determine whether a change in estimated circuit parameter was to the result of a fault or loading.