Software for Fluid Power Technology

From Editor

The purpose of the Software Review section of the Journal is to present information to the reader about engineering software, including simulation programs, to highlight their specific features and their "fitness to purpose" in the unique field of fluid power and motion control. It is, of course, impossible to establish evaluation criteria matching the needs of all readers, therefore readers should not look for absolute ratings but more or less "fuzzy" opinions of the reviewer. A software program is like a wrench, just a tool to solve problems. It is good to solve some problems and not so good for others and this depends on both the nature of the problem and the users' attitude - and generally when we review software we do not know either. A software tool can be highly specialised and great for a some applications but not so well suited for others, on the other hand another software tool can be more flexible and generally applicable but without outstanding features. It is impossible, and even misleading, to say which one is better. What we hope to accomplish is to give the reader information necessary to take his/her own decision.

PumpLinx: A Modelling Tool for Pump Designers

1 Introduction

PumpLinx is the new contender in pump design software. It is a comprehensive CFD tool created specifically for pump analysis. Starting with a CAD-CAM file, PumpLinx enables the design engineer to create a virtual pump to be used as a numerical test-bed. The code predicts the transient internal fluid dynamics of the pump, including cavitation, loads, pressures, torques, and velocities. These three-dimensional results are output graphically and in spread sheet format to enable the engineer to use them as he or she would use empirical data to improve pump performance. SimuLinx, which developed PumpLinx, was founded in 2005 by an experienced group of software developers and pump engineers formerly from the NASA Marshall Space Flight Center and CFD Research Corporation in Huntsville, Alabama. According to its president, Sam Lowry, the first priority of this new company is to develop an engineering design tool for the pump industry and the unique design challenges it faces. More specifically, the first focus has been on positive displacement pumps.
Lowry states that a conscious decision was made to return to the drawing board in creating PumpLinx, in order to take full advantage of recent developments in software architecture. This fresh start and focused approach is believed to provide a strength and flexibility not possible when trying to incorporate new technology into an old framework.
This article provides SimuLinx’s perspective on the unique challenges and requirements of modelling positive displacement pumps and the features that Pump- Linx offers to address those needs.

2 Essential Pump Model Capabilities

Table 1 provides a list of key issues pertaining to modelling positive displacement pumps. These are not claimed to be the only modelling issues, nor are they necessarily unique to pumps, but they are seen as critical components of any effective pump design tool.
Table 1: Key issues for a Pump Design Tool

• Efficient and Accurate Treatment of Rotating/ Sliding Components • Cavitation and Bubbles • Robustness • Run Time • Ease of Use for the Designer • Output

2.1 Treatment of Rotating / Sliding Component

Historically, one of the biggest challenges for CFDer’s in modelling pumps has been the inconvenient fact that pumps contain rotating and/or sliding components. This means that the spider-web of numerical grid cells, inherent to most CFD methodologies, gets twisted and torn. The common solution is to create two sets of grids that slide along a shared interface. The challenge is then establishing accurate communication across this dynamic interface. Virtually all current commercial CFD codes use a spatially smeared average and/or an explicit (i.e. time lagged) transfer of data across these boundaries. This can lead to inaccuracies, or even convergence problems that prevent any solution.
In order to address this numerical challenge, Pump- Linx contains a proprietary grid coupling algorithm that automatically and precisely projects and mates two mis-matched moving grid sets across a shared boundary. When the model is first set-up, the user specifies which interfaces are shared, after which point the code takes over and automatically matches them as the pump moves. The solution across this interface is then solved implicitly (vs. explicitly) at every time-step. According to Yu Jiang, SimuLinx’s lead developer, this fully implicit interface matching makes it possible to handle
very complex pump geometries with significant improvements in accuracy, speed, and robustness.
Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 1:PumpLinx’s implicit matching (light grey patches) across a rotating interface in a gerotor pump

Rotating and sliding components also create a numerical challenge in that, in addition to the risk of being twisted and torn, the grid also gets compressed and stretched between moving parts. One common approach to handling such dramatic grid volume changes is dynamic re-meshing, in which an entirely new grid topology is created at each time steps. However, dynamic re-meshing comes at the expense that the old grid topology and values must be interpolated onto the new. This is very difficult to do and still conserve fluid mass and momentum, especially in tight clearances where the grid gets compacted to dimensions on the order of microns. PumpLinx, by virtual of the interface projection algorithms, does not need to re-mesh its internal grid topology and is able to ensure full conservation from one time-step to the next. This accuracy is seen as critical when treating the tight clearances inherent in positive displacement pumps.

2.2 Cavitation and Bubbles

The majority of positive displacement pumps incur cavitation at some point in their operating cycle. Even for those pumps where pressures are sufficiently high to prevent cavitation, minute quantities of noncondensable gases (e.g. entrained air) can still have significant effect on compressibility and bubble formation. As pump designers are well aware, cavitation and bubbles are a potential cause of noise, vibration, performance loss, and component damage. Figure 2 shows the predicted cavitation for a vane pump, indicating bubble formation at several locations.

Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 2:PumpLinx: Instantaneous gas volumes in a vane pump. The range is from 0 (blue) to 1 (magenta)

PumpLinx includes a cavitation/gas model to predict both expansion of non-condensable gases and vapor formation due to cavitation. The model has evolved from the work of Singhal et al (Singhal 2002) with the improvement that it can model the introduction, transport, and compressibility of non-condensable gases. The model predicts the mass/volume fractions of non-condensable gas and vapor throughout the pump. Given these numerical data, the design engineer can see when and where cavitation and voids are occurring and effectively understand and correct the problem. Figure 3 illustrates the importance of modelling cavitation for this class of pump. Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 3:PumpLinx: With and without the cavitation model including comparison with experimental data

2.3 Robustness

Robustness refers to the ability of a code to converge smoothly and rapidly to a solution for a given case. For pumps, the challenge of robustness can be exacerbated due to large density ratios (e.g. due to cavitation), sliding/rotating interfaces, and disparate length scales. With regard to the latter, pumps typically have length scales ranging from meters and centimeters in the pumping chamber down to millimeters and microns in small features and clearances. Tip clearances, for example, can be on the order of microns. In some cases the small features can be neglected. In other cases, however, they have significant influence on pump performance and must be included in the model.
As an example of a small feature that can not be neglected, Fig. 4 shows the resolution of small metering grooves in a gerotor pump. The geometry and dynamic communication between these grooves and the rotor chamber was revealed to have a significant influence on pressure lock and noise. Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 4:PumpLinx: Metering groove in a gerotor pump

PumpLinx’s approach to robustness is through its grid architecture and solution algorithms. For example, the implicit coupling across a dynamic interface is an important asset with regard to robustness. The cavitation model also has modifications to improve convergence and been shown to converge smoothly for cases with liquid to vapor density ratios up to 40000:1. Several other proprietary features are incorporated into PumpLinx to make it robust, but ultimately, to quote from Don Quixote, “The proof of the pudding is the eating,” and PumpLinx’s claim of robustness can only be determined by specific applications over time.

2.4 Run Time

The speed and turn-around time of a numerical tool is often a significant factor in its usefulness. A factor of five in turn-around time may not sound like much until it is translated into working hours. A code that can return a solution in half a working day is significantly more productive than one that takes several days, or longer, to run.
   Typically, 3-D transient pump problems are computationally intense. In order to resolve the features of a complex pump, a CFD model will often have 200,000 grid cells or more.
   PumpLinx addresses the speed issue by state-of-theart solvers and a binary tree grid structure that greatly reduces the amount of geometric data stored and the corresponding calculation time required per time step. It has been successfully run for cases of over 1 million grid cells.
   As an example of run time, 250 grid cells were used in the PumpLinx model for the industrial gerotor above (Fig. 4). The corresponding time required to model the pressures, velocities, and gas void fractions for three revolutions (the time for the initial cavitation to wash out and the pump attain periodic behavior), with 180 time steps per revolution, was approximately six hours on a Pentium 4 2.8 gigabytes desktop. Fewer cells would run faster, but at the expense of accuracy.

2.5 Ease-of use for the Design Engineer.

CFD packages typically fall into two categories: those appropriate for design engineers and those requiring a CFD expert to operate. In general, the latter category seems to be the rule rather than the exception.
SimuLinx’s commitment is to develop a software tool created first and foremost for the design engineer. Toward this end, the package strives to provide an intuitive path starting with incorporation of CAD-CAM files into the code and then proceeding seamlessly through model set-up, execution, and analysis. This is all within a single graphical users interface (Fig. 5).

Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 5:Integrated PumpLinx interface

As a code specifically developed for pump simulation, PumpLinx provides customized pump templates for specific pump types (axial, centrifugal, gerotor, piston and vane). These templates provide boundary conditions, operating conditions and properties relevant to the configuration of choice. Logical dependencies are implemented amongst these parameters. For example, RPM is specified at one location and is automatically propagated to the relevant components. The intent is to make the initial set-up of even a complex model a matter of minutes, not hours. Subsequent modifications or changes in operating conditions become a matter of seconds. If a pump template is not available for a given configuration, SimuLinx will create one and guarantees a successful first pump model.
The following case is provided as an example of the time required to set-up a virtual pump. The case started
with a NASTRAN grid of an experimental multi-ring axial piston provided to SimuLinx by Chongqing University. Starting with this complex geometry and using the standard PumpLinx Axial Piston Pump Template, a virtual model of the pump was up and running in less than twenty minutes. Figure 6 shows the predicted cavitation for this unique configuration.

Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 6:Predicted cavitation in a multi-ring piston pump (Pump courtesy of inventor Shiliu Li, College of Mechanical Engineering, Chongqing University )

The above cases were all set-up in the PumpLinx “Normal” mode, which uses defaults and logical dependencies to greatly reduce the number of inputs required by the user. The code also provides an “Expert” mode to those CFDer’s who may want the freedom to override dependencies and manipulate boundary conditions and other parameters outside the normal range for a given pump type.

2.6 Relevant Data

A partial list of the types of output data currently available in PumpLinx is provided in Table 2. Field data, such as pressure, velocities and gas fractions are available at every grid cell in the model domain and can be displayed graphically (e.g. as in this report). Table 2: 3-D data available in PumpLinx Field Data:

Integrated Data:

A virtually unlimited number of numerical point probes can be created and the data plotted (Fig. 7) or output to a spread sheet. Parameters such as flow rates, forces, torques, etc. can be integrated over any surface, (e.g. gear surfaces, walls, inlets, or outlets), and then plotted or stored. Full field data at each time step can be saved for future reference, analysis, or restart.

Double-click the images to enlarge them
and click once to make them thumbnail size again.

Fig. 6:PumpLinx: Predicted pressure versus time at monitoring points in a vane pump

3 Summary

SimuLinx, founded in 2005 with pump modelling as its primary focus, has recently released the pump design tool PumpLinx. PumpLinx is a transient 3-D CFD software package designed and created specifically to address the unique modelling challenges of fluid pumps, in particular positive displacement pumps. These challenges include complex geometries, rotating/ sliding components, small geometric tolerances, entrained gas and cavitation. PumpLinx is able to meet those challenges using state-of-the-art algorithms and numerics incorporated into a streamlined architecture that improves speed, robustness and accuracy. The framework of PumpLinx was created with the philosophy that it must be fast and intuitive for design engineer to learn and use. Toward this end, a single application and graphical window is used for grid generation, problem set-up, and post-processing. Templates for specific pump types (axial, centrifugal, gerotor, piston, vane, etc.) are provided and lead the engineer through the steps needed to create and analyze a virtual pump, starting with importation of CAD-CAM geometry and ending with display and output of relevant data.
The end result of PumpLinx is a numerical test bed that enables the designer to look inside the pump to view the dynamics of the pressure, flow, loads, and gas/vapor bubbles. PumpLinx provides the design engineer with the images and data that enable him or her to identify the causes of cavitation, reduce noise, improve efficiency, extend life, reduce the design cycle and, in general, build a better pump.

References

Singhal, A. K., Athavale, M. M., Li, H. and Jiang, Y. 1992. Mathematical Basis and Validation of the Full Cavitation Model, Journal of Fluid Engineering, Vol. 124, Issue 3, pp. 617-624

BACK