Pumps & Systems, February 2009
Computer simulations of pumps and compressors can now serve the same function as hardware testing. These simulations can be done in less time with less cost while providing engineering data of similar quality. Furthermore, computer modeling can be performed directly by the engineer doing the hardware design, thus providing a tight link between analysis and design optimization.
Computer Pump/Compressor Simulations
We will start by defining pump simulation. Modeling and simulation can take many forms, but in this article, pump simulation refers specifically to 3-D computational fluid dynamics (CFD), an example of which is shown in Figure 1.
Figure 1. CFD model of an external gear oil pump (surface pressures and x-y data plots)
In this context, Computational Fluid Dynamics applies to liquid or gas, compressors or pumps, and can be used to model fluid motors and a wide range of other fluid components. Such CFD codes usually start with a CAD model of the geometry and then create a 3-D numerical mesh representing the flow path through the device. This mesh is subsequently used to model the dynamics of the flow based on fundamental laws for conservations of mass and momentum. The output of these models includes plots and three-dimensional maps of flowrates, loads, head-rise, power, pressure ripples, velocities and torques, depending on whether it is for a pump, compressor or motor.
For liquid applications, the more advanced codes include aeration and cavitation. Aeration refers to the presence of non-condensable gases, such as air; and cavitation refers to the formation of vapor from the liquid. Both can have a significant effect on pump performance and life.
If temperature influences the properties and performance, such as in compressors, the code must also solve conservation of energy equation (Figure 2).
Figure 2. CFD predicted air temperature (on a cutting plane) in a lobe compressor
To be effective, CFD codes should provide the same data as a hardware test. The question is often raised, however, whether these simulation tools are sufficiently accurate, easy-to-use and fast to be used reliably as virtual hardware tests.
The Advantages of Simulation
The general advantages of computer simulations have been well established. For the majority of pump/compressor developers, computational stress simulations have been an integral part of their design process for decades. This is true for axial, centrifugal and positive displacement (PD) pumps. In contrast, the use of CFD simulation has lagged behind other computer simulation tools because of difficulty-of-use, slow turnaround time, and inaccuracies or deficiencies in the results.
Within the last few years, however, state-of-the-art pump modeling has evolved to where setting up a detailed model of even a complex pump takes less than an hour. This reduction in set-up time has been accomplished, in part, by new meshing algorithms that start with a CAD file and automatically generate the numerical grid needed for CFD. With the newer software, running a simulation on a standard desktop now takes between one hour to overnight, depending on the type of pump and the extent of the system. These statistics include both axial/centrifugal pumps and positive displacement (PD) pumps, the latter of which are more challenging to model from a numerical perspective.
Using state-of-the-art results from a good CFD code should be within 10 percent accuracy or better compared with experimental data. Accurate and quick simulation can reduce the time and cost associated with hardware testing by providing the same type of data that would be collected during testing. With simulation, it is easy to add "measurement" probes in locations that would be extremely difficult to access in a physical experiment. Perhaps most importantly, numerical simulation offers flow visualization that allows the engineer to "look" inside the pump during operation and provides invaluable insight as to how the pump operates and how it might be improved.
Validation and Application
The engineering data of interest for the modeler are the same as those of
