As human-beings, our differences are what makes us unique (if I may quote the Seek Discomfort crew – “What makes you different is what makes you beautiful”). For turbomachines, this sentiment also rings true.  We design different turbomachines because we have varied roles, needs and constraints for them. To that effect, there is no universally best turbine, compressor, or pump. Therefore, figuring out which set of “skills” a turbomachine should have is the key role of a design engineer so that they may effectively capture and estimate performances of the machine they will work on early on while having the certitude this is the best that can be done.

Generative design is one of these recent buzzwords that characterizes an approach to the design of components (or systems) that has been around for quite some time already. Rather than producing one geometry for one value of each input (such as boundary conditions, flow coefficients, number of stages, etc.), generative design allows you to create thousands of designs within minutes that you can review, compare, and filter to select the one that best suits your needs. Let’s look at an example of an axial turbine design process comparing traditional preliminary design vs. generative design.

Approach 1 or what most companies call Traditional Preliminary Design,  is to look in textbooks and previous examples of what a given turbine for that application “should” look like. It may involve things like using Ns-Ds diagrams, load-to-flow diagrams, blade speed ratio vs. isentropic velocity ratio correlations, scaling/trimming existing designs, etc. These have served their purpose well enough, but they have their limitations which make them fairly challenging to really innovate. Such limitations include previous experience/data being restricted to a given fluid, relative clearance size, given configuration, lack of secondary flows, etc. A summary of a traditional preliminary design workflow (familiar to too many engineers) is presented below.

Now, we know that changing (ahem, improving) your workflow is not always easy. But growth happens through discomfort and switching to a generative design approach does NOT mean rebuilding everything your team has done in the past. What it effectively gives you is the confidence that the input parameters you finalized will provide not only the desired performance but the best ones that can be achieved (and it saves time too…a lot of time). From there, you can use these inputs in your current design software or you can continue the design process in our design platform, AxSTREAM® (meaning you can add generative design capabilities upstream of your existing workflow or replace parts/all of that workflow depending on what makes the most sense for you). You can pay your engineers to do engineering work, instead of visiting online libraries and guessing input parameters in hope they will find the needle in the haystack. Or, with generative design, you kind of look for haystacks and shake them until the needle falls off.

So, how does this work in AxSTREAM, you may ask? Very well, I may reply :D.

Alright in all seriousness (my editor is probably rolling their eyes right now), let’s dive into a workflow using generative design. In AxSTREAM, you can provide input values as either a fixed number or a range of numbers.  Fixed numbers are for things you know for sure like the inlet temperature of your gas turbine because that comes out of your cycle code (Did you know we are redesigning our cycle tool to integrate with our thermal-fluid network and include new capabilities like scripting, transients, etc.?). The range of values is used when you don’t know the numbers and that is where the magic happens. In AxSTREAM, you let the software know how many different geometries it should create from scratch, and it outputs some pretty cool stuff, see figure below.

The ranges can be as reasonably wide as you would like so you can initially cover and create a broad design space at first, with the goal to understand how each combination of inputs influences your outputs. From there, it’s easy to set filters to discard flow paths you are not interested in and eventually narrow down on which inputs work best for your application, design philosophy and constraints so you can run (if desired) a new generation (I did say it was called Generative Design, right?) of geometries until you know that what you see on the screen will be the very best there is at that 1D level. You can then efficiently navigate the design space (where each point is a different design that meets your targets) and select the best machine for your case, see figure below.

Additionally, machine selection can also be made considering the performance of different flow paths at off-design conditions to ensure that your time is well-spent and that your design is efficient throughout the entire speedline.

In summary, traditional preliminary design allows you to answer the question of “Does this design meet my requirements?” while generative design provides the answer to “What is the design that best meets my requirements?” Through AxSTREAM’s 1D generative design capabilities, users can study the effect of parameters on size, performance, number of stages, etc. to enable easy and early critical design decisions (nothing is worse than getting 90 percent of the way through a project only to realize the design is no good).

If you are interested in learning more about generative design an
d attending Turbo Expo this year, don’t forget to ask about our free guest passes,  swing by booth #424 to say hello,  and join us for our stage presentation on generative design for cutting-edge turbomachinery development taking place Tuesday, June 14th at 1:00 PM in Hall 3 Theater. Learn more and contact us here: https://www.softinway.com/news-events/upcoming-events/asme-turbo-expo-2022/