|INSIGHT: Unsteady Clouds on the Horizon
3D steady analysis is no longer sufficient to provide the flow insights needed to advance modern turbine and compressor design. Design processes must now address the fourth dimension--time--to tackle the unsteady blade row interaction effects that adversely impact performance and durability. Last month we sat down with Bob Ni to discuss the merits of unsteady analysis during design; in this month's issue of The Flow, we sit down with Bob and Michael Ni to discuss how turbomachinery designers large and small can capitalize on cloud computing to make unsteady analysis a reality. Bob is the Chairman and CTO of ADS. Michael is the Manager of Commercial Products at ADS and leading the company's cloud computing initiatives.
FLOW: Mike, what is cloud computing?
MIKE: We touched upon this when we first broached the subject of cloud computing in this newsletter last year. There are lots of definitions, I view it simply as a web service that gives organizations instant access to computing capacity on a pay as you go basis.
FLOW: What does cloud computing have to do with unsteady analysis?
MIKE: Simply put, cloud computing makes it practical for organizations to put unsteady analysis to use. Though the benefits of unsteady analysis are pretty well understood in turbomachinery design circles, most organizations have not had the luxury of time and computing resource to take advantage of it. Cloud computing removes these boundaries and allows you to rent unsteady analysis time on a pay as you go basis. As such, we think it's a tremendous enabler, especially for the smaller and mid-sized shops that don't have the luxury of large internal clusters.
FLOW: Is it cost effective?
BOB: For unsteady analysis, absolutely. Imagine that you're a mid sized turbomachinery designer with a small internal cluster in house. You're designing a 1.5 stage turbine and you've run steady calculations on your internal cluster to arrive at a couple of design candidates. Now you want to conduct unsteady simulation to anticipate and mitigate adverse flow phenomena impacting performance and durability. As always, your blade counts for each row are different and there are more passages to analyze than your cluster can support. You have four choices:
- Use a technique like harmonic balance to eliminate the need for capacity.
- Scale the case configuration down to allow it to run on your cluster.
- Convince your boss to add extra CFD capacity to your cluster.
- Run large scale unsteady on the cloud.
Options 1 and 2 are reasonable ways to get the simulation to run within the constraints of your cluster, but you'll trade off simulation accuracy at a time where it counts most. Option 3 sounds good but requires capital investment and long lead times for deployment. This can be challenging to justify, especially if the capacity you're adding is only used sporadically during a design cycle.
That leaves Option 4. What's great about this option is that you're paying as you go by the CPU hour. So you can provision as many processors as you need for the simulation and know that you're paying only for what you consume.
As you'll see from the case study in the next section, we were able to conduct large scale unsteady analysis on a 1.5 stage turbine and get the results back in 3 days with a computing infrastructure cost well under $1,000. More importantly, the unsteady predictions revealed flow behaviors that resulted in a 2% reduction in predicted efficiency. Is this insight important and worth a couple of days and a couple of hundred dollars to find out? You bet it is.
FLOW: Is it secure?
MIKE: We believe cloud computing from vendors like Amazon Web Services to be as secure as most large organizations, and probably more secure than most small and midsized businesses. You can really start to think of the cloud as an elastic resource that sits behind your firewall, with the requisite fine-grained access controls, network and hardware isolation and operational certifications to go along with it. Amazon's recent announcement of GovCloud even layers on ITAR and FIPS 140-2 endpoint compliance if desired.
FLOW: If cloud-based unsteady analysis is as cost-effective as you suggest, why would you settle for simpler approximation techniques?
BOB: Execellent question! As always, it depends. Simpler (and faster) approximation techniques make sense early in the design phase as your team dials in on some potential design candidates. Once they've been narrowed down, we believe that high fidelity insight is critical because you are trying to undertand important time-varying flow behaviors impacting performance and durability. At this stage of the game, we think it makes good sense to conduct cloud-based unsteady analysis given the cost.
FLOW: Are you suggesting it's not necessary to have an internal cluster?
MIKE: Absolutely not. We believe it's always prudent to have an in-house cluster that will address your "steady state" simulation needs. We do, however, advocate the use of cloud-based CFD for applications like large scale unsteady analysis that are critical but episodic during a design cycle. Why invest in a cluster if the capacity isn't needed year round?
FLOW: Any final thoughts for our readers?
BOB: Three parting thoughts. First, if you are looking to push the envelope on durability and performance, unsteady analysis will be a requisite for success. Second, cloud computing makes it practical to carry unsteady analysis by removing the cost and computing resource constraints. Third, the cloud model makes it easy to "try before you buy" since it's pay as you go, so dip your toe in the water and see for yourself. We think you'll be pleasantly surprised.
FLOW: Thanks Mike and Bob.
MIKE, BOB: Our pleasure.
CASE STUDY: Conducting Large Scale Unsteady Simulation in Design Time with Amazon EC2 To speed the development of more fuel efficient and durable jet engines, the U.S. Air Force Research Laboratory and ADS partnered with Amazon Web Services to devise an effective design simulation solution. With Amazon EC2, ADS demonstrated that large scale unsteady simulation can be dialed up on-demand and performed affordably and within the time constraints of commercial design. <more>
TECHTIPS: Generating Time Resolved Plot Data for Animations in Code Leo
It is often much easier to visualize the results of unsteady analysis through animation. Here's how to configure Code Leo to generate time resolved plot data that can be used with Tecplot to produce animations. <more>
Welcome to The Flow
Welcome to The Flow, a newsletter for monthly insights on turbomachinery CFD published by AeroDynamic Solutions, Inc.
Each month we'll spotlight a topic of interest, discuss a case study and/or provide useful pointers about how to get the most out of the ADS CFD system.
You are receiving this email because you or someone else you know thought you would be interested. To unsubscribe please click here. We value your privacy.