There is a question often heard when training customers in the company’s turbomachinery design software. Is this a good fan? The answer they get is always this. It could be. Unsurprisingly, they often find the answer unsatisfying. The point of the indeterminate response, however, is not to pass judgment on their work. Instead, the response is an attempt to explain that most fans or blowers are part of a larger flow system and that performance is always a function of what conditions the fan sees. Obviously, a blower designed for one flow rate and head rise is going to perform differently at another condition. The point being stressed is that instead of trying to design the perfect fan or blower, one must each design a fan to meet its own specific requirements.

Complicating the matter of fan/blower design is the fact that a fan is often tightly coupled to the overall system. One of the biggest mistakes designers (or more accurately, managers) often make is to compartmentalize the design without regard to its place in the overall system. A poorly designed inlet, or a sharp turn just upstream, can be devastating to fan or blower performance. No matter how one measures performance, be it head rise, flow rate, noise level, or efficiency, the measurement will be affected by the surrounding environment.

In the past, designing a fan or blower was a trial-and-error affair. Fortunately for today’s designers, a host of design and analysis tools are available to aid in the process of designing the fan, tools that help remove some, but not all, of the guesswork.

Fig. 2. Streamline curvature results of a radial blower from AxCent.

Analysis methods A practical design of a new fan requires an integrated approach that allows for a hierarchy of modeling methods. Starting straight off with a full three-dimensional CFD analysis can be a very time-consuming process. The trick to effective design is to narrow down the enormous range of possibilities with more rapid modeling solutions and systematically move to more complex and accurate methods as one approaches the final design iterations.

The plot in Fig. 1 shows blower results from a meanline modeling package called FANPAL. The performance characteristics can be generated in only seconds of computational time. Using this approach, hundreds of design possibilities can be explored in a single day.

Meanline programs rely heavily on empirically based models. These results can be greatly improved when the modeling parameters have been calibrated to specific test data.

Streamline curvature techniques offer the ability to analyze flow fields in three dimensions. These methods run in just a small fraction of the time of CFD analysis, but they do have limits. Most important, loss, deviation, and blockage must be prescribed to the streamline curvature solver (generally via meanline analysis). Fig. 2 shows the pressure distribution in a radial blower using a multiple streamline curvature technique.

While streamline curvature techniques do not quantify performance in the classic sense, they are extremely useful in controlling pressure and velocity distributions in three-dimensional space.

Fig. 3. Quasi-three-dimensional CFD solution of a rotor/stator fan design.

CFD analysis is the ultimate analytical approach to modeling fluid flow. Within CFD itself is a hierarchy of solution levels that range from minutes to literally months of computational effort. Quasi-three-dimensional methods are probably the most cost effective. These techniques can account for some, but not all, of the three-dimensional phenomena that occur in a flow field. Fig. 3 shows a blade-to-blade CFD calculation. The method solves for a single-stream plane and the gross effect of area change in the third dimension is accounted for in the solution. However, the more complex twisting of the streamlines in three-dimensional space is not considered.

Full, three-dimensional CFD where the flow is assumed to be unchanging with respect to time is the most advanced method in common use today for industrial applications. More advanced methods, such as time accurate analysis, large-eddy simulation, and direct numerical simulation, are computationally intensive methods that have not yet trickled down from research applications.

Acoustics

Fig. 4. Results of acoustic modeling at the meanline level.

Modeling acoustic performance is a prime concern for fan designers. The fan industry is quite unique in the world of turbomachinery in that many manufacturers do not know the aerodynamic efficiency of their designs, but are very aware of the noise levels they produce. Obviously, this is driven by customer priorities. Purchasers of new computers won’t have any idea about the efficiency level of the computers’ fans, but they can readily recognize that they are quieter than their previous computers. This aside, there is an undeniable link between aerodynamic efficiency and noise level, and many low-level models explicitly factor in the efficiency into the noise calculation.

Predicting noise levels is one of the most challenging frontiers in thermo-fluids science. Empirically based models offer some help. These models are generally based on basic parameters of the system and can be run at the meanline level, such as the output shown in Fig. 4 from FANPAL. Simple methods such as these are useful in figuring approximate sound levels and frequency distributions, but they cannot account for more subtle effects that come from the detailed three-dimensional shaping.

CFD-derived techniques, where the acoustic energy is linked to the levels of turbulent energy, have generally had mixed success. More involved CFD methods using unsteady calculations have had more success, but only at very significant costs. Unfortunately for engineers, there seem to be preciously few options available between the extreme ends of the spectrum of simple empirical based models and advanced CFD based approaches. Some progress using unsteady panel type methods has had some success and holds some hope for the future.

Optimization

Automated methods of optimization have come on very strong in the last few years. Originally developed for the aircraft propulsion industry, these techniques have experienced an explosion of different applications in many different industries. The idea is simple. The optimizer explores a specified range of design inputs until it finds the best combination of parameters to achieve the optimum design. The optimum can be defined anyway the user specifies. Examples include optimization based on efficiency, pressure rise, manufacturing cost, noise level, or any number of other factors or combination of factors.

The optimization technique can be used with any modeling level of analysis and is only limited by the time and budget of the engineer. Fig. 5 shows the results of an axial fan optimized with a quasi-three-dimensional CFD technique. Note the very significant difference in the design depending on how the optimum is defined.

Fig. 5. Geometries for a baseline and two optimized axial fans.

Conclusion

It should be remembered that these methods and tools have not removed the need for genuine design experience and insight. Computer-aided design tools are just that. They are analogous to the hammer and saw, not the carpenter. They still require a skilled engineer to achieve the proper result. But for the qualified designer, computer-aided design tools have greatly improved the design process of turbomachinery over the last several years to the point where the question “Is this a good fan?” can be answered confidently, “It will be.”