Rainer Kurz, Solar Turbines Inc., Sand Diego, California, USA;

and Said Mokhatab, Consultant, Tehran, Iran

The compressor design process has undergone dramatic changes within the past decade, which has led to a substantial reduction in the overall design cycle time and a fundamental improvement in compressor stage performance.

These changes were primarily brought about by enhanced compressor design tools, which was the result of a number of important developments that occurred in the turbomachinery and computer industries. These developments did not occur overnight. Rather there has been a gradual but steady evolution as computers have become faster, less expensive, and easier to use. Additionally, tools for aerodynamic, structural, rotordynamic and manufacturing analysis can be integrated.

Despite the great success of computational tools, the design of compressors still relies heavily on experimental data derived from designing and testing actual compressors under realistic conditions. This means that, although computational tools and suites of tools are available for purchase, all of these tools still require a significant amount of background data (that is usually proprietary to the compressor designers) to accomplish a successful compressor design.

Modern computational fluid dynamics (CFD) codes not only predict the viscous, three-dimensional flow through impellers, but they can also be used to predict the flow through the entire stage (i.e. through a geometry consisting of rotating and stationary parts) of a compressor (Figure 1). They not only predict the performance of a compressor at its design point, but also under off-design conditions. However, while these calculations reveal unsurpassed details of the flow structures, they are still not very accurate with regards to such bulk values as overall efficiency, volumetric flow, or head, unless they are sufficiently calibrated by test data from reasonably similar geometries.

History

Fluid flow, energy and heat transfer follow fundamental laws of physics. The basis of these laws is that certain quantities are “conserved.” Conserved quantities are, for example, mass, momentum and energy. The fundamental relationship of these properties relating to fluid flow, the Navier-Stokes equations, has been known since the early 19th century. However, since these equations are non-linear, non-homogenous and coupled partial differential equations, solutions to practical problems can only be found using numerical methods and high speed computers.

The numerical solution requires one to divide the physical space into a large number of small volumes. Then, the conservation laws can be applied and solved for the small control volumes. The appropriate definition of the size (and therefore number) of these control volumes is important, because a large number of volumes requires more computational time, whereas a smaller number of volumes may yield inaccurate results.

A particular challenging problem for turbomachines is the effect that turbulence has on friction and heat transfer. Turbulent fluid motion is an irregular condition of flow, in which various quantities show a seemingly random variation with time and space. Early approaches model the macroscopic effects of turbulence by lumping them into an additional viscosity of the fluid. Even today’s CFD tools use turbulence models that derive the effect of turbulence from properties of the mean flow, based on empirical or semi-empirical relationships. Therefore, the use of appropriate turbulence models for a given problem, together with appropriate models describing the transition of flow from a laminar state to a turbulent state, require a solid understanding of these models and the problem at hand.

Interfaces

The art of designing a compressor requires the combination of aerodynamic desires with the rotordynamic, structural, manufacturing and probably commercial reality. Important boundaries are imposed by the requirement to achieve stable rotordynamical behavior and structural integrity (Figure 2).

Aerodynamic forces affect the rotor and its stability. The rotor hub diameter is a result of the requirement for a sufficiently stiff shaft. Blade thickness and blade geometry have to be sufficient to reduce stress levels to an allowable limit. The capabilities of the manufacturing process regarding the capability to generate certain geometries within certain tolerances have to be taken into account. Cast, milled, welded or brazed components impose different constraints on the geometries that can be achieved. And, last but not least, the end product has to be commercially attractive to the end-user.

During the early years of computer use in turbomachinery design, calculations that were previously performed manually were now automated and run on large mainframe computers. The use of these mainframes led to a reduction in design cycle time, since these digital computers were much faster than the manual calculations they replaced, and an increase in modeling accuracy, since large-scale numerical techniques could now be applied to finite element analysis (FEA) and CFD analysis. However, a number of fundamental bottlenecks to accelerating the design cycle and improving the design itself still remained. The design cycle was slowed by the large number of users queuing up to run jobs on each mainframe.

The difficulties in data entry and interpretation of results were caused by programs that were run in batch mode with a poor user interface (limited or non-existent graphics), and an engineering department structure which assigned very limited design and analysis responsibilities to separate groups. The engineering department structure slowed the design process due to the time involved in passing the results of one group’s work on to the next group in line. For instance, the aerodynamic design group would complete a flowpath design and pass its design to the aerodynamic analysis group, who would evaluate this design and find some flaws. It would send the design back to the design group with suggestions for improvement, and the design group would rework the flowpath based on the analysis group’s suggestions. Eventually, the design would make it out of the aerodynamics area, but would then have to run the gauntlet of mechanical design, structural analysis, system integration, and prototype development testing. Clearly, this department structure slows the design process, separates the designer from the final product, and can easily lead to designs that are not fully optimized due to errors in communication and misinterpretation of the design goals.

The turbomachinery design tools that have evolved from these early mainframe programs have encouraged a fundamental transition in the compressor design process and engineering group structure. Today, a single engineer can have an entire compressor design system running on a modern high-speed desktop computer. This design system can include any combination of the following:

  • A meanline compressor design program for one-dimensional design, analysis, and data reduction.
  • A three-dimensional flowpath and blading design program to create the full aerodynamic flowpath.
  • A viscous three-dimensional CFD analysis program with integrated meshing tools for coupled (impeller-diffuser return channel/volute) viscous flow analysis.
  • A three-dimensional FEA program with integrated meshing tools for rapid stress and natural frequency analysis.
  • A suite of rotordynamic analysis and bearing design tools for rapid determination of rotor critical speeds and stability.
  • A computer aided design (CAD) program to generate manufacturing drawings of the final hardware.

Compressor design

The tools that form this design system have two distinct advantages over their counterparts from even a few years ago:

  • The liberal use of graphical user interfaces (GUI) has led to computational tools which are easier to learn, more intuitive to use, and allow more rapid assessment of individual designs because of the immediate visual feedback you get when design changes are made.
  • The use of object linking and embedding (OLE) to connect individual programs allows design changes made in one program to be transmitted to all of the other linked programs. As a result, for instance, a single skilled designer with a desktop computer can assess the effect of a change in impeller blade thickness on meanline aerodynamic performance, three-dimensional aerodynamic performance (inviscid and viscous), and impeller stress and natural frequency, with the push of a single button.

An experienced designer, with these tools at his disposal, can markedly accelerate the design process because there is no waiting for a time slot to open up in other groups and little is lost in translation between the aerodynamic, structural, and rotordynamic design phases. This acceleration of the design process results in reduced design cycle time and lower design costs. In addition, the designer will now have a stronger sense of ownership of their design, since they will be the pivotal people at each phase of the design process.

One of the most exciting compressor design system enhancements, now in development, is the linking of numerical optimization techniques to the design codes. The implementation of automatic design optimization should significantly reduce the design cycle time, because the computer can analyze thousands of potential designs in the same amount of time it takes a person to look at one or two alternatives. These optimization techniques should eventually lead to higher performing designs as well, because the computer can find optimized solutions that a designer may never have considered. Take for instance, the fully linked comprehensive design system discussed above, where the meanline design code, 3D flowpath design code, and 3D structural analysis code are all linked together through OLE.

To implement an optimization technique within this design system, a comprehensive set of design “rules” such as optimum impeller diffusion characteristics and optimum diffuser area schedules would be added. These rules will likely come from the collective experience of a group of seasoned designers or by carefully examining previous successful (and unsuccessful) designs. These optimization routines would be linked to the entire design system, and would then iterate through a large number of potential design combinations until one or more optimal geometries was identified. The designer would also be able to input his goals in rank order, for instance “sacrifice up to one point in peak efficiency to keep the factor of safety on the peak blade stress above 2.0,” or “of primary importance is a flow range greater than 25%, followed by an efficiency of at least 85%, followed by....” Clearly, adding optimization routines to this linked design system will lead to a very powerful design tool, one that will change the entire process of compressor design, as well as the role of the design engineer.

Conclusion

Enhanced tools for compressor design and analysis can, in the hands of experienced engineers, and backed by a relevant body of experience and test data, yield compressors with ever higher performance. This will significantly increase the value of turbo compressors to the end user.