Approximating the Steady-State Temperature of 3D Electronic Systems with Convolutional Neural Networks

Thermal simulations are an important part of the design process in many engineering disciplines. In simulation-based design approaches, a considerable amount of time is spent by repeated simulations. An alternative, fast simulation tool would be a welcome addition to any automatized and simulation-based optimisation workflow. In this work, we present a proof-of-concept study of the application of convolutional neural networks to accelerate thermal simulations. We focus on the thermal aspect of electronic systems. The goal of such a tool is to provide accurate approximations of a full solution, in order to quickly select promising designs for more detailed investigations. Based on a training set of randomly generated circuits with corresponding finite element solutions, the full 3D steady-state temperature field is estimated using a fully convolutional neural network. A custom network architecture is proposed which captures the long-range correlations present in heat conduction problems. We test the network on a separate dataset and find that the mean relative error is around 2% and the typical evaluation time is 35 ms per sample ( 2 ms for evaluation, 33 ms for data transfer). The benefit of this neural-network-based approach is that, once training is completed, the network can be applied to any system within the design space spanned by the randomized training dataset (which includes different components, material properties, different positioning of components on a PCB, etc.).

Approximating the Steady-state Temperature of 3D Electronic Systems with Convolutional Neural Networks

Thermal simulations are an important part in the design of electronic systems, especially as systems with high power density become common. In simulation-based design approaches, a considerable amount of time is spent by repeated simulations. In this work, we present a proof-of-concept study of the application of convolutional neural networks to accelerate those thermal simulations. The goal is not to replace standard simulation tools but to provide a method to quickly select promising samples for more detailed investigations. Based on a training set of randomly generated circuits with corresponding Finite Element solutions, the full 3D steady-state temperature field is estimated using a fully convolutional neural network. A custom network architecture is proposed which captures the long-range correlations present in heat conduction problems. We test the network on a separate dataset and find that the mean relative error is around 2 % and the typical evaluation time is 35 ms per sample ( 2 ms for evaluation, 33 ms for data transfer). The benefit of this neural-network-based approach is that, once training is completed, the network can be applied to any system within the design space spanned by the randomised training dataset (which includes different components, material properties, different positioning of components on a PCB, etc.).

Exploring the impact of balconies on cooling energy demand in an arid climate zone

This paper explores the impact of balconies on the energy demand required for cooling in the arid climate zone of the city of Adrar, in Algeria. For the purpose to assess several situations of the balconies, we have chosen a parametric method based on a campaign of thermal simulations. The open and eliminated balcony type were selected and characterized by four parameters: the balcony to room ratio, the orientation, the window type, and the balcony position. A set of 100 simulations was selected randomly based on the Monte-Carlo probability technique. The final sample was corrected based on Cook’s distance which gave 85 simulations as a final sample size. A generalized regression model was performed to identify the impact of each parameter. The accuracy of the model is above 97% and the sensitivity analysis shows that the most important factor is the balcony to room ratio which could reduce the energy demand up to 26% followed by the window type (24%), the orientation (8%) and the balcony position (5%). This conclusion stresses the idea of considering the balcony as a passive solution to reduce the cooling energy demand.

Predicting energy demand of residential buildings: A linear regression-based approach for a small sample size

The key design strategies that reduce the energy demand of buildings are not present in most thermal codes in many countries. Therefore, modeling techniques offer an alternative to combine the architects' modus operandi with the energy efficiency in the early stages of architectural design and with higher speed and precision. However, a review of the scientific literature using modeling techniques shows that most researchers use a relatively large sample of thermal simulations. This paper proposes a simplified method based on the linear regression modeling technique and considers a relatively smaller sample of thermal simulations. A total of 6 key building design strategies were identified, related to the urban context, building envelope, and shape factor. A simulation protocol containing 60 possible combinations was designed by random selection. In the present study, the Pleiades software was used to estimate the annual energy demand for heating and cooling for a typical dwelling in a humid climate zone. A parametric study and sensitivity analysis to identify the most efficient parameters was performed in SPSS 21. The resulting model predicts the annual energy demand with an accuracy of 93.7%, a root mean square error (RMSE) of 5.88, and a scatter index (SI) of 8.59%. The models performed could efficiently and quickly assist architects while designing the buildings in the architectural practice.

Coupled Electrochemical-Thermal Simulations and Validation of Minichannel Cold-Plate Water-Cooled Prismatic 20 Ah LiFePO4 Battery

This paper discusses the quantitative validation carried out on a prismatic 20 Ah LiFePO4 battery sandwiched between two minichannel cold-plates with distributed flow having a single U-turn. A two-way coupled electrochemical-thermal simulations are performed at different discharge rates (1–4 C) and coolant inlet temperatures (15–35 °C). The predicted battery voltage response at room temperature (22 °C) and the performance of the Battery Thermal Management System (BTMS) in terms of the battery surface temperatures (maximum temperature, Tmax and temperature difference, ΔT) have been analyzed. Additionally, temperature variation at ten different locations on the battery surface is studied during the discharge process. The predicted temperatures are compared with the measured data and found to be in close agreement. Differences between the predicted and measured temperatures are attributed to the assumption of uniform heat generation by the Li-ion model (P2D), the accuracy of electrochemical property input data, and the accuracy of the measuring tools used. Overall, it is suggested that the Li-ion model can be used to design the efficient BTMS at the cell level.

Adaptive Thermal Comfort of an Office for Energy Consumption-Famagusta Case

The aim of this study was to determine how much thermal comfort can be obtained through heat/energy transfers between the office/external air and the transparent/opaque surfaces of an office by combining different transparent and opaque wall surface ratios with different window opening percentages using dynamic thermal simulations. It found that the optimum window-to-wall ratio (WWR) for energy conservation is 40%, with a 20% window opening ratio. The 80% and 90% thermal comfort ranges of the adaptive thermal comfort methodology are found in May, October, September, and the yearly average, while June and August are only in the range of 80% acceptability. The office constantly loses heat through air flow with any glass size on its external facade and any window opening ratio. Moreover, all sizes of opaque and transparent internal surfaces transferred heat from outside by conduction, while the opaque wall similarly always transferred energy to heat up the office air internally and outside air externally through convection. The external glass also heats the office air by convection, except in the months of January, November, and December.

Hygrothermal simulation and risk evaluation - The impact of discretization on numerical results and performance

Regulations for modelling when deducting thermal simulations are represented in the standards [1]. However, the level of model detail regarding discretization in hygrothermal simulations and especially for evaluating the mould risk on surfaces of organic vapour barriers is almost never discussed. The presented approach shows that the chosen discretization of the simulation model is one of the most influencing factors for the risk analysis of surfaces of very fine layers, such as paper vapour barriers, in walls with interior insulation via hygrothermal simulations. To reduce the computational performance issues caused by very fine finite volume meshes [2], the hygrothermal properties of the connecting surfaces of the finite volumes can be calculated instead. For the risk analysis the VTT-Model was implemented in the hygrothermal simulation program HAM4D_VIE, followed by a comparison of the effect of discretization on the results of the surfaces of the vapour barrier. The results of the comparison are discussed with regard to numerical results and their qualitative impact on computational performance. The presented numerical model will be proposed as an alternative for risk analysis on surfaces of vapour barriers, where mould growth would either stay undetected or the necessary discretization with elements comes at the cost of computational performance.

Optimizing Closed-Loop Liquid Cooling Solution for Extreme High Power Multi-Packages

Current technological developmental efforts for high performing liquid assisted air-cooling solution (LAAC) is presented in this research. The roadmap needs of high-power packages such as client, datacenter and graphics product segments are addressed. Systematic evaluation and qualification process for developing cost effective LAAC solutions to provide cooling capabilities for multiple socketed packages are also discussed. This effort includes qualifying high performing heat exchanger (finned radiator) and cold plate with different fin designs. Thermal simulations were conducted and correlated with experimental data. The concept of power density approach is introduced for predicting the cooling capability of newly developed LAAC solution for different lidded package form factors. The paper presents supporting data of performance evaluation from graphics segment that extends the solution capability to data center segment. The technology developed here will address the increasing cooling capability needed for future products with high power dissipation and multiple socket cooling configuration.

Three-Dimensional Thermal Simulations of 18650 Lithium-Ion Batteries Cooled by Different Schemes under High Rate Discharging and External Shorting Conditions

In this work, three-dimensional thermal simulations of single 18650 lithium-ion battery cell and 75 V lithium-ion battery pack composed of 21 18650 battery cells are performed based on a multi-scale multi-domain (MSMD) battery modeling approach. Different cooling approaches’ effects on 18650 lithium-ion battery and battery pack thermal management under fast discharging and external shorting conditions are investigated and compared. It is found that for the natural convection, forced air cooling, and/or mini-channel liquid cooling approaches, the temperature of battery cell easily exceeds 40 °C under 3C rate discharging condition. While under external shorting condition, the temperature of cell rises sharply and reaches the 80 °C in a short period of time, which can trigger thermal runaway and may even lead to catastrophic battery fire. On the other hand, when the cooling method is single-phase direct cooling with FC-72 as coolant or two-phase immersed cooling by HFE-7000, the cell temperature is effectively limited to a tolerable level under both high C rate discharging and external shorting conditions. In addition, two-phase immersed cooling scheme is found to lead to better temperature uniformity according to the 75 V battery pack simulations.