Ohio Supercomputer Brings Better, Stronger, Faster… Simulations!

154,000,000,000,000 Reasons

 

Back in July, we posted a Blog announcing Third Frontier Commission Invests in Ohio Businesses. Just recently, Douglas J. Guth in HiVelocity Magazine highlighted 154 Trillion reasons why. His article is a must read, especially for those Ohio businesses who have felt like they were part of the “missing middle.”

 

 

Read the full article… “Ohio’s supercomputer: 154 trillion calculations per second”    http://tinyurl.com/qhrdrpy

 

 

AltaSim Technologies is once again honored and humbled to appear in the article. We have worked with National Digital Engineering and Manufacturing Consortium (NDEMC) to show how supercomputer access can dramatically improve product design capabilities, in our case specifically in the area of design for cooling of electronic devices and circuits. While this work builds on our 20+ years of experience using HPC to analyze Multiphysics problems, the real win is for small and middle market companies to access applications that would otherwise be inaccessible. These applications will help them solve specific classes of problems using HPC capabilities at Ohio Supercomputer Center (OSC) without having to purchase or maintain computer power that most small and middle market companies cannot afford. In the near future, companies harnessing the power of the Internet can simulate real-world product testing quickly and accurately bringing new technologies to market at an accelerated pace.

 

 

We believe there are many companies with talented people who will be innovating faster than ever, harnessing tomorrow’s technology and capturing today’s markets is part of the equation to help them be successful.

MEMS Energy Harvester for Reusing Waste Heat

During our recent webinar with COMSOL on thermal-structure interaction modeling, we at AltaSim Technologies demonstrated modeling of a MEMS energy harvester that scavenges waste heat. Examples of sources for waste heat range from microprocessor chips, to internal combustion engines, to chemical processing plants. If the waste heat generated from these cases could be used to generate additional energy, then overall energy consumption could be reduced. Estimates from CANMET Energy Technology Centre indicate that worldwide, waste heat exceeds 1 TJ annually.

Estimating heat transfer coefficients

Previously we have discussed ways to complete a feasibility thermal management analysis. For a given thermal budget, we want to determine whether the junction temperature will meet specifications for reliability and thermal runaway.

 

In electronic systems that rely on air flow for cooling, convection is of paramount importance.

 

Power dissipation due to convection is given by q = hA(ΔT), where q is the power dissipated (W), h is the heat transfer coefficient in

W/(m2K), A is the surface area exposed to the flow (m2) and ΔT = Tsurface – Tambient (⁰C).  The largest degree of uncertainty lies with Tsurface and h, so in the feasibility analysis we can solve for temperatures given a value for h.

 

Obtaining a heat transfer coefficient a priori is difficult, we suggest using a range of values representative of different conditions. Ultimately if high values of h are selected it suggest that the thermal behavior may need to be redesigned.

 

The table below provides approximate heat transfer coefficients for different conditions.  [H.S. represents Heat Sink.]

Case

Electronics Environment

Heat transfer model

Thermal solution over component

Approximate Heat Transfer Coefficient (in W/m2K)

Approximate thermal resistance (⁰C/W)

A

Handheld device with up to 2 mm air gap, operates when horizontal

Conduction through air

(worst case)

None

Treat as conduction in gap

50 – 100

B

Small module, no fan, orientation controlled (always vertical). Open top & bottom, chimney style convection

Natural convection h also depends on ΔT & height

Optimized plate fin heat sink (vertical base & fins)

For plate fin heat sinks, correlations are readily available*

3 – 6

(based on H.S. area)

1 – 5

C

Laptop

Single small fan

Optimized heat sink

25 – 100

(based on H.S. area)

1 – 20

D

Desktop

Multiple fans

Optimized heat sink

50 – 150

(based on H.S. area)

0.25 – 5

* White, Frank M., Heat and Mass Transfer, Adison-Wesley © 1991, p. 408-9.

 

 

With heat sinks it is important to remember that simply increasing the number of fins does not automatically lead to increased heat dissipation, as the fins approach each other the resistance to flow also increases, causing h to drop.  This is why in case B above, the systems cannot be optimized further.

 

For cases C and D, further effort may be needed to select an appropriate heat sink.  Some suppliers can provide an estimate of the thermal resistance associated with the heat sink based on a representative airspeed in Linear Feet per Minute (LFM). So you may be able to complete your system-level feasibility analysis by simply using a network of thermal resistances, without having to explicitly assume a heat transfer coefficient value.

 

Learn more about electronics cooling. and AltaSim’s support for electronics cooling analysis.

 

COMSOL Tips & Tricks

Decrease Total RAM Required to Solve Sequentially Coupled Multiphysics problems.

Another COMSOL Tips & Tricks Deposit:

 

Are you solving a multiphysics problem?  If so, do you know if your equations are sequentially coupled or intimately coupled?  Well, if your equations are sequentially coupled, you can decrease the total RAM required to solve your systems by dividing your solution sequence into multiple steps.  A classic example of this is when solving Thermal-Structural problems.  Because displacements are usually small from thermal expansion, the heat transfer equations are not significantly affected by geometry changes.  For this reason, you can first solve the heat transfer equations on the undeformed geometry as a first step.  Then, with a second step, solve the solid mechanics equations with the temperature field solution from step one as an input.  This technique saves on computer memory requirements enabling you to solve larger problems using your existing computing power.  (One notable exception to this rule for Thermal-Structural problems is when thermal contact is enabled.  In this case, your equations become intimately coupled.)

 

Watch a 10 minute video on this: