Junction Temperature of a SiC based DC/DC Converter from Vehicle Front Loading Test Bench

Health degradation issues in automotive power electronics converter systems (PECs) arise due to repetitive thermomechanical stress experienced during real-world vehicle operation. This stress, caused by heat generated during semiconductor operation within PECs, leads to the degradation of semiconductor's operating life. Estimating the power semiconductor junction temperature (Tj) is crucial for assessing semiconductor degradation in operation. Although physics-of-failure-based models can estimate Tj, they require substantial computational power.

To address this, a comprehensive and accurate training dataset is necessary to develop a reduced-order data-driven model or a real-time executable machine learning model that can predict the semiconductor junction temperature with lower computational complexity. Here, we provide a dataset for a Silicon Carbide (SiC) based automotive DC/DC converter. This dataset is derived from a simulation model, which has been validated for accuracy using a Hardware-in-Loop test bench. 4 driving cycle profiles are included in the dataset namely ExtraUrban, Urban, NEDC, and WLTP. The datasets includes the following data

• Inputs: Vbat(V): Battery Voltage, Iload(A): Load Current, Tcool(K): Coolant Temperature, Vdc(V): Output DC voltage

• Outputs: Tj2 : Tj of Low side SiC device , Tjd2: Tj of Low Side diode, Tj1: Tj of high side SiC device, Tjd1: Tj of high side diode

The data is sampled at intervals of 100 ms. This input and output data can be utilized to train machine learning models to create a virtual sensor for predicting Tj.