An automated method for creating compact dynamic thermal models for in-situ prognostics of power electronics and power LED packages

In their typical application domains, power electronics components (power MOSFETs, IGBTs) as well as high power LEDs are subjects to cyclic operation. The subsequent on-off cycles result in cyclic changes of the junction temperature, resulting in thermo-mechanical stresses at critical thermal interfaces such as the die-attach layer. Ultimately, in long-term, these stresses lead to the degradation of such interfaces, causing a continuous increase of the corresponding partial thermal resistances of the overall heat-flow path of such packaged devices. Further thermal interfaces on system level may also be subject of degradation as a result of ageing during the product lifespan.Over the last two decades, structure function analysis has proven to be a powerful tool in laboratory testing to detect the resulting failures such as die attach voiding or delamination. Combining power-cycling tests with thermal transient measurements resulted in commercial test equipment widely used nowadays in reliability testing of power electronics components, but as of today, no embedded, in-situ solution was published that is aimed at health-monitoring and prognostics purposes of such components during field operation. The aim of this paper is to present a method that could automatically generate time series of element values of a Cauer-type compact thermal model of the heat-flow path power electronics components and power LEDs during their operation. The series of partial thermal resistance values obtained this way allows a quasi-real-time analysis of the mission profile dependent degradation of the heat-flow path, allowing prognostics purposes, such as estimation of the remaining useful lifetime (RUL) from the perspective of the thermal properties.