Demystifying “Pushbutton” Approaches for CFD & FEA Design, Analysis, Redesign, & Optimization of Turbomachines
Although there is not just one way to design a turbomachine there sure is one way not to do it; blindly.
A misconception that I commonly see when teaching engineers about fundamentals of turbomachines, as well as when leading design workshops, is that some engineers (mostly the younger generations) envision themselves plugging numbers, pushing buttons and getting results immediately without any real brain power behind their actions.
Nowadays, software packages are an integral part of an engineer’s toolkit, but in the same way that a mechanic would not (or should not) use a screwdriver as a hammer, each software has its own applications and ways to use it.
Working with turbomachines (as well as many other systems) requires an engineer to have some fundamental knowledge of the physics that govern the flow as well as some understanding of how equations are used behind the graphical user interface. Validation of a code requires thorough comparisons with available data for a wide range of applications to ensure that the results will be reliable. However, a small R245fa axial turbine for waste heat recovery will not be designed in the same exact manner as a powerful nuclear steam turbine or an aerospace gas turbine. The fundamental equations will of course hold, whatever the application, but the way to approach the system will differ and computers can only make so many choices “on their own”.
“Should I add an extra stage to gain 0.5% efficiency despite an increase in mass and manufacturing?” is one question only an engineer can answer. A “Pushbutton” conceptual design approach would take inputs that are close to prototype data while the task at hand is to figure out what can be done. Using a range of values for each input parameter in a preliminary design tool like AxSTREAM’s allows for a significantly increased flexibility while allowing for thousands of possible designs to be reviewed, leveraged, filtered, and finally selected. Although personalized scripts could help answering some of the questions, the final decisions have to be made by the designer who can review the integral parameters/data of the machines before proceeding with the detailed design phase.
Moving forward with the design process the 1D/2D (meanline/streamline) analysis of a machine is more straight-forward. In its simplest form is a pushbutton type of task, although tweaking of loss model settings, alterations of the flow path for optimization, off-design predictions and more, need some user-defined actions. With any flow path there are a lot of different parameters (angles, dimensions, twist, fluid composition, etc.) that can be optimized, and looking at all of them at the same time is neither realistic (especially in terms of computing time) nor really sound (the more parameters are varied the further from the original design we are really to obtain). The software is here as a tool but cannot stand alone.
In the same way, the profiling of blades can be seriously made easier thanks to algorithms and different criteria but in the end it is again the engineer that controls and inspects the geometry. Blades will be different-looking if we are dealing with a cooled or an un-cooled gas turbine or between any of these and a steam turbine although they may all be axial machines.
A pushbutton FEA or pushbutton CFD application is, again, barely realistic. Due to the seamless data transfer between the AxSTREAM modules, for example, the process of running FEA or CFD is made as convenient as possible by allowing geometry, boundary conditions, loads, etc. to be automatically inherited, But even then some human actions are required. These include creating a mesh (even though AxSTRESS & AxCFD possess a turbomachinery-optimized mesh generator the mesh can be refined, restructured, etc.), selecting the convergence controls, and so on. The “calculation start” button is always only one click away but with most software (unlike AxSTREAM) the geometry, boundary conditions, fluids, etc. need to be imported between the different tools used and the calculation requires some manual settings and checking, especially before running time-consuming tasks like CFD and FEA.
Trusting software to let it do the “complete” work annihilates any real freedom in the design process of a turbomachine. One cannot simply enter initial data, hit the “design my machine” button, go get a coffee and expect a final, magical, efficient machine. In-house codes of large OEMs are a good example since they are only applicable to their range of machines and usually work fine as long as they are used within this very limited range. However, when new applications or designs arise the codes do not usually hold trustworthy results and the engineer only blindly sees a final or intermediate result that is just not how it should be although it might still do the job. If the task had been performed with a more flexible and user-controlled tool the machine would most likely have been more efficient but instead the software made choices invisible to the engineer that were not necessarily appropriate for the application.