Assessing the impact of novel polymers and thermal management in a power electronics module using machine learning approaches

Acharya, Palash V.; Lokanathan, Manojkumar; Ouroua, A.; Hebner, Robert E.; Strank, Shannon; Bahadur, Vaibhav

doi:10.1109/itherm51669.2021.9503255

Cited by 2 publications

(8 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…9). The range of SHAP values are relatively higher in the positive direction because the average [28] temperature value for the steady-state databank is very low when compared to the overall range, as is evident in the histogram (tile 1 Â 1 of the pair plot in Fig. 9).…”

Section: Resultsmentioning

confidence: 91%

“…SII available in the Supplemental Materials on the ASME Digital Collection) applied to the package are as follows: (i) natural convection (5 W/m 2 K) at 25 C applied at the external surfaces of the package, (ii) adiabatic condition at the Fig. 5 Geometry of the half-bridge SiC module analyzed in this study and rendered using SOLIDWORKS [28]. The module is covered with a 4 mm thick encapsulant (not shown in the figure to display the details that have been captured for wire bonds, MOSFETs and diodes).…”

Section: Description Of Simulation Methodologymentioning

confidence: 99%

“…Fig. 6 High-resolution meshing of the model, consisting of 2.31 million elements [28] side faces of the package, and (iii) forced convection HTC applied at T sink at the internal surfaces of the heat sink. A steady total power (P D ) of 1500 W (corresponding to a heat flux of $400 W/ cm 2 ) and a square power waveform with a maximum value of 1500 W and a 50% duty cycle was dissipated equally in all 14 MOSFETs for the steady-state and transient simulations, respectively.…”

Section: Description Of Simulation Methodologymentioning

confidence: 99%

“…To tune the hyperparameters, an exhaustive grid search crossvalidation with a two-step approach was used to carry out optimization. The first step consists of evaluating a suitable range of hyperparameter values by sampling the combinations at random, and the second step consists of carrying out a concentrated search in the narrowed-down parameter space to arrive at the combination yielding the best results [28]. For neural network architectures, the weights and biases were optimized using adaptive moment estimation (Adam) optimization algorithm, which utilizes momentum and adaptive learning rates to converge faster.…”

Section: Development Of Machine Learning Modelsmentioning

confidence: 99%

“…The obtained results were used to assess ongoing material development efforts and their implications on the thermal performance of power electronic (PE) packages. It is noted that a part of the present study (specifically, steady-state analysis) was discussed in another recent study [28]. Presently, we build up on the methodology developed in our past work [28] to provide a holistic assessment of the influence of various parameters on the steady-state as well as the transient performance of the power electronics package.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Machine Learning-Based Predictions of Benefits of High Thermal Conductivity Encapsulation Materials for Power Electronics Packaging

Acharya

Lokanathan

Ouroua

et al. 2021

Journal of Electronic Packaging

Self Cite

View full text Add to dashboard Cite

Machine learning (ML)-based predictive techniques are used in conjunction with a game-theoretic approach to predict the thermal behavior of a power electronics package and study the relative influence of encapsulation material properties and thermal management in influencing hotspot temperatures. Parametric steady-state and transient thermal simulations are conducted for a commercially available 1.2 kV/444 A SiC half-bridge module. An extensive databank of 2592 (steady-state) and 1200 (transient) data points generated via numerical simulations is used to train and evaluate the performance of three ML algorithms (random forest, support vector machine and neural network) in modeling the thermal behavior. The parameter space includes the thermal conductivities of the encapsulant, baseplate, heat sink and cooling conditions deployed at the sink; the parametric space covers a variety of materials and cooling scenarios. Excellent prediction accuracies with R2 values > 99.5% are obtained for the algorithms. SHAP (Shapley Additive exPlanations) dependence plots are used to quantify the relative impact of device and heat sink parameters on junction temperatures. We observe that while heatsink cooling conditions significantly influence the steady-state junction temperature, their contribution in determining the junction temperature in dynamic mode is diminished. Using ML-SHAP models, we quantify the impact of emerging polymeric nanocomposites (with high conductivities and diffusivities) on hotspot temperature reduction, with the device operating in static and dynamic modes. Overall, this study highlights the attractiveness of ML-based approaches for thermal design, and provides a framework for setting targets for future encapsulation materials.

show abstract

Section: Resultsmentioning

confidence: 91%

Section: Description Of Simulation Methodologymentioning

confidence: 99%

Section: Description Of Simulation Methodologymentioning

confidence: 99%

Section: Development Of Machine Learning Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations