Defining Upfront CFD According To Blue Ridge Numerics
How do you define “upfront CFD”?
The marketplace has defined upfront CFD as flow and thermal simulation integrated within an MCAD environment and used during early stage product development, normally by multi-tasking engineers before products enter physical testing.
Isn’t it just “CFD Lite”?
That misconception is one of the bigger challenges we face. There are indeed a number of “CFD Lite” packages on the market but they are very different to upfront CFD. Multi-tasking mechanical engineers face the exact same product development questions as dedicated analysts, so their analysis requirements are no different. They don’t have time to deal with tools that give them half of what they need. True upfront CFD deploys the full spectrum of fluid flow and heat transfer analysis capabilities but in a very different way. The solvers are fundamentally the same: it’s the method of delivery that is so different. The Lite packages are really just teasers developed by the developers of traditional analyst-centric CFD packages. The Lite packages are out there hoping design engineers will sample them and then pursue an upgrade path that requires them to purchase a seat of traditional CFD. The Lite products aren’t meant to be a solution – and that’s why they aren’t successful.
Will upfront CFD require a change of culture? Where do you see the resistance to this?
There is process change and then there is culture change. Most of the time we see a need for slight process change. This is usually in companies where CFD is not used and there has been absolute reliance on physical testing. Cultural change is something we encounter more at companies where traditional CFD is in use. The dedicated analysts have a lot of fears and concerns about mechanical engineers using CFD. While upfront CFD can be very complementary to the use of traditional CFD – both can add value in most organisations – there are often cultural, political and even philosophical barriers to bridge. This resistance has been nurtured by some of the software suppliers. Only a few years ago an executive from a major traditional CFD developer equated upfront CFD to giving guns to children.
For companies implementing upfront CFD as their first simulation tool, the process change starts with a view that simulation is better for design trade-off studies than physical testing. Innovation and optimisation can be achieved with greater efficiency in a digital environment while ultimate qualification or validation is still best done in the physical realm on a flow bench or test rig. Once an engineer grabs on to these concepts they are off to the races.
The traditional process is to develop a detailed model in MCAD and then proceed to physical prototyping and testing to assess performance and plan design changes. Using this process some trade-off studies will be done: some based on lab tests, others based on intuition or past experience. In order for upfront CFD to really add value the customer needs to build a simple, functional model in MCAD and conduct a first-pass simulation. Based on what is learned, more realistic geometric detail, material properties and operating conditions are added so a set of candidate design variations can be submitted for simulation. Design reviews are conducted with 3D, interactive simulation results instead of just bench test data. Ideas and compromises are made and eventually a production ready MCAD model will be submitted for final validation at the lab. The process is a bit different and may seem a little awkward at first but the newfound efficiencies in time, money and communication usually convert even the most sceptical engineer after a project or two.
How does your approach differ from other CFD vendors?
Blue Ridge Numerics was the first and possibly still the only CFD company founded for the sole purpose of meeting the needs of mechanical design engineers. Starting in 1992 at a time when Unix was king, we developed CFdesign software exclusively for the PC Windows workstation and we employed industry standard FEA techniques to make CFD more tolerant of real-world 3D solid model geometry while automating the meshing process.
Our user profile is a person who knows a lot about the products they engineer, they are comfortable driving MCAD, but solving fluid flow and heat transfer problems is maybe one of ten areas of responsibility. They typically have well defined, focused product performance questions and only a few days to get reliable answers. Using CFD software is something they only do when a project requires
which may be a few days every six to ten weeks. They simply don’t have time to be experts in fluid and thermal dynamics and they certainly don’t have time to be expert on subjects like finite volume or finite elements, model meshing, solver selection, etc. This is our typical customer and we’ve built an organisation that really understands the needs, wants, challenges and goals of this customer.
By contrast, all the major CFD companies were founded by technologists using a finite volume approach to develop software for the dedicated analysts. The analytical power of these codes is beyond question but they are inherently weak when it comes to handling complex 3D MCAD assemblies. Most of their users do not use MCAD. As these vendors matured they built organisations with core competencies essential to satisfying the ongoing requirements of full-time users. Anyone familiar with Clayton Christensen’s book The Innovator’s Dilemma will understands this scenario.
How do you handle CAD integration?
From our perspective, native, associative CAD integration is the first tenet of upfront CFD. All geometry changes must be made in the MCAD system and the CFD tool must automatically and instantaneously recognise and adapt to design changes without loss of product data. CFdesign maintains this level of integration with SolidWorks, Autodesk Inventor, Pro/ENGINEER, CATIA, UGS NX and Solid Edge, with plans to add CoCreate.
Our approach eliminates the need for importing or translating a product design into the analysis environment, a time consuming process that induces error and strips the geometry of all intelligence (part IDs, material properties, etc.) assigned in the MCAD environment and disconnects the assembly from essential product data such as the BOM, assembly constraints, tool paths and drawings. I find it ironic that most CAE companies are pushing simulation driven design’ without realising the only way to deliver on the promise is to allow the MCAD model to drive the simulation process. When MCAD drives CAE then you can have simulation driven design.
Why not form partnerships with CAD vendors to deliver embedded CFD?
We do maintain very close partnerships. Our approach to integration requires great collaboration with MCAD developers. As to why we don’t fully embed CFdesign into the MCAD environment there are several reasons. First, we view it as inefficient use of resources. The time and manpower needed to maintain embedded applications from release to release for eight MCAD systems is huge. This approach would diminish our ability to deliver new analysis capabilities essential to the mechanical design market. Second, embedded applications tie up the MCAD seat during the meshing and analysis processes. This is a big penalty to the customer. Finally, if the MCAD integration is done correctly and the CFD interface makes it simple enough to set up and launch an analysis, the user really doesn’t care if it’s embedded.
What of the traditional CFD specialist?
The CFD specialist isn’t going anywhere. They have tremendous job security because they are uniquely qualified to stay on the bleeding edge of CFD technology and put it to use in ways that can uncover the product development breakthroughs of tomorrow. That said, there are thousands of companies that will never have a CFD specialist on staff and for the companies that do, they all report not being able to address all the requests for assistance they receive. Most products still go to market without the benefit of CFD.
This simply illustrates that traditional CFD is extremely valuable but it is not a one-size-fits-all solution. That’s why there is a buzz in the market around upfront CFD. It is satisfying needs that have been unmet for decades.
CFD is computationally intensive. What type of workstation is required (i.e. processors, memory) and when are dedicated compute servers required?
More is always better and I guess it always will be but you can do upfront CFD analyses with standard engineering workstations. We recommend a minimum of 1 GB of RAM and a 20 GB hard drive. Most of our customers use single processor machines but CFdesign is built to optimise multi processor systems. Very few of our customers have dedicated servers. I suppose this would be nice but it really isn’t realistic for our market. We’ve recently added 64-bit computing support and this definitely helps. In the near future we will introduce a new approach to distributed computing that will allow mechanical design groups to maximise the computing power available on a network. Our own engineers use laptop PCs so hardware doesn’t have to be elaborate to do upfront CFD.
There’s been a lot of talk about the eventual move towards integrated FEA/CFD, where true structural/fluid interaction can be simulated. How far away from this are we?
The era of structural/fluid interaction is here. Right now, though, this means different things to different people. In 2003 we introduced a software module that allowed solid components to move through fluids and of course show the interaction. We also added the ability to pass loads directly into Nastran, Ansys, Abaqus, Cosmos and Mechanica so engineers could evaluate flow-induced and thermal-induced stress. Both of these capabilities are extremely valuable for upfront simulation and decision making. I think what most dedicated analysts would like to do though is to conduct optimisation studies where the CFD application passes loads to the FEA package and then the FEA package passes deflection of deformation information back to the CFD system for a closed, multi-loop simulation. This is very cool stuff but in the short-term the compute times and software application expertise requirements make this an area for R&D not product development.
With compute power increasing all the time, is real time analysis a genuine possibility in the near future?
If real time means instantaneous results on a 3D model then I think we are a long way away. If, however, it means instantaneous feedback with stable results in ten to thirty minutes then I think we’re there. When CFdesign is used early in the process to assess preliminary designs it is entirely possible to get reliable answers in less than thirty minutes. As the assemblies get larger and more complex and the analysis requirements are scaled up to create a real-world operating environment, the analysis time usually turns to hours. This is still much less than the time to build and test a physical prototype and at the end there will be views into product performance that simply cannot be obtained from a test lab.
Read the complete story ...
Be the first to review this article