Figure 1. Pump No 1 characteristics (Experimental vs PumpLinx predictions).
HMS Group is one of the leading pump manufacturers in Russia and CIS countries for oil & gas, water & utilities, thermal & nuclear energy, steel & mining, chemical processing and ship building.
Highly qualified and experienced engineers of HMS Group with extensive application of advanced computer technologies develop reliable, energy-effective and efficient pumping solutions for the most stringent requirements of Russian and international customers under tight schedule. HMS Group
Figure 2. Simulation model for calculation (cut away)
Cost Cutting with Pump Performance Prediction
Courtesy of WORLD PUMPS July/August 2013
Prediction of prototype performance is often a critical requirement for success. New designs have typically required expensive and time consuming hardware testing. However, virtual testing with CFD has the potential to reduce or even replace hardware testing, but only if it can predict performance with sufficient accuracy, ease-of-use and speed. This article considers three case studies where CFD testing has been proved effective.
Over the last two decades, a number of commercial CFD codes have been developed to simulate pumps and compressors. A relatively recent entry into this virtual test world is the simulation tool PumpLinx. In order to evaluate the effectiveness of PumpLinx in terms of accuracy, ease-of-use, and speed, the R&D department of HMS Group conducted tests using test data for several pump designs. Some of the results and conclusions are provided in the text below:
‘Speed and Hardware Requirements: With regard to hardware requirements, the computational resources required to solve similar problems in PumpLinx are several times less than other codes that have been evaluated. This allows the use of PCs for the simulation, thus eliminating the requirement for special clusters. The memory requirement was found to be 0.9 Gigabyte of memory for a model with one million elements; with a typical component model ranging from 0.5 to 3 million cells. Using a standard laptop, the calculation of accurate results requires much less time than the competition, being from 10 to 25 times faster, with the accuracy of the predicted head rise, power, and efficiency to within 5% over the operating range of the several pumps tested. Further details of the specific tests are available via the World Pumps Web-site.’
‘Evaluation Results: A number of numerical experiments were conducted and compared with experimental data. In the article, a few examples of the calculations for typical flow parts are shown. In general, it should be noted that mesh creation can be done with the maximum time having been two hours for the pumps in the current study. The GUI enables the convenient and quick setting of parameter for the computational model, and its strengths include: 1) being able to set parameters for groups of entities, e.g. for the walls with the same properties it’s possible to set parameters together; 2) the ability to set the properties of the working fluid, in the data base; and 3) templates for a range of pumps. The results presented below were run using a steady state ‘frozen blade’ multi-reference frame approach which solves for the flow through the impellers in a rotating reference frame. Steady state simulations provide results that are averaged over time. In comparison, transient simulations solve for a sequence of time-steps, rotating the impeller at each time-step. Transient simulations require significantly more time to run, but with the advantage that the results are as accurate, or more accurate than, steady-state simulations, since they can include the effects of periodic phenomena such as pulsations or bubble formation and collapse. PumpLinx can efficiently run transient simulation, but for the first two cases presented, good accuracy was obtained using the steady-state approach, with the advantage that the simulation time was on average only twelve minutes per data point. The third case presented demonstrates the need for transient simulation for improved accuracy. A transient simulation in PumpLinx typically requires approximately 2-3 hours per point. This capability to run a transient simulation, in a reasonable time, is critical for problems where periodic phenomena like cavitation are significant, as when trying to predict the suction performance of a pump.’
‘In addition to quantitative evaluation of the software, the engineers at HMS Group evaluated the ease-of- use and efficiency of the software in creating the mesh and setting up boundary conditions, based on a scale of 1 to 10, with 10 being optimal. For this pump, PumpLinx received marks of 9 out of 10 in these categories.’
The HMS Group performed an independent evaluation of the CFD simulation tool PumpLinx for performance prediction in terms of speed, accuracy, and ease-of-use, by using test data for several pump types. The tests concluded that PumpLinx is an excellent engineering tool for virtual testing of pumps; predicting power, head and efficiency within 5%, with an ease-of-use and speed that enables performance tests within two days at most, enabling HMS Group to provide superior products at lower cost, in less time.’