Unappreciated challenges in applying four quadrant pump data to waterhammer simulation
AFT Impulse™ Technical Paper
Authors: Trey Walters, P.E., Scott Lang and David Miller, P.E., Applied Flow Technology
Presented at the 13th International Conference on Pressure Surges 2018, Bordeaux, France, November 14-16, 2018
Part 1: Fundamentals
The transient analysis of reverse flow and rotation in pumps has evolved over the years into modern four quadrant pump waterhammer simulation. Exact characteristics for a given pump for reverse flow and/or reverse rotation are normally unavailable, and manufacturer curves are often mapped to previously published four quadrant data sets for similar pumps. Assumptions made in this mapping process can cause extreme differences in the simulation. If these assumptions are unaddressed, critically incorrect conclusions about the system’s transient behavior may be made, impacting both design and operation.
The available choices to the waterhammer analyst and the consequences of those choices are thoroughly detailed in Part 1 of this paper.
Part 2: Applications Examples
Transient analysis of reverse flow and rotation in pumps requires the use of four quadrant data. This data is normally unavailable for a given pump, and existing dimensionless four quadrant data is selected based on specific speed. There are different methods for dimensionalizing the four quadrant data, which can result in significant differences in transient predictions.
This study examines four examples with three exhibiting reverse flow. The two most convenient methods of dimensionalizing four quadrant data are used, and significant differences in transient predictions are demonstrated and discussed.
Part 1: RECOMMENDATIONS
Significant and potentially dangerous differences exist in waterhammer simulation results based on two different applications of four quadrant methods. These differences result from certain steady-state and transient assumptions about a pump’s behavior. Engineers are highly encouraged to expand the scope of their sensitivity studies to account for these differences.
Wan and Huang, 2011 (24) provide an alternative to the 4QBEP and 4QOP options discussed in Section 4. What they essentially do is keep the four-quadrant curve of interest but modify it using the manufacturers curve in the first quadrant zone of normal pump operation. This avoids the issues raised in this paper but ends up with a four-quadrant curve no longer consistent with the original test data. There also is a significant discontinuity at zero flow in their proposed method. It is not clear to the authors whether this offers any substantial improvement. Wan and Huang should be applauded for recognizing that an issue exists.
For the practicing engineer, the authors have no recourse but to recommend substantial sensitivity studies be performed in waterhammer simulations. Both 4QBEP and 4QOP should be evaluated. Comparisons should be made to MC transient results in the first quadrant (see Section 5). Multiple four quadrant data sets near the pump specific speed of interest should be included. Engineers should identify the initial and final steady states of the system and consider transient simulation results in light of agreement with the final steady-state results after all transients have died out.
Part 2: RECOMMENDATIONS
The examples in this paper reinforce the authors’ belief that substantial sensitivity studies should be completed in waterhammer simulations involving four-quadrant data. Both options should be evaluated, and multiple four-quadrant data sets near the pump-specific speed should be included. Engineers should identify the initial and final steady states of the system and consider transient simulation results with the steady-state values in mind. Comparisons should be made to MC transient results in the first quadrant (see Part 1, Section 5).
The issues raised in this paper are closely linked to off-BEP operation. However, they still exist and can impact all simulations regardless of BEP. In particular, the choice of four quadrant data set becomes more impactful if the reference point concern is settled. It is the authors’ opinion that additional caution should be exercised when analyzing reverse flow of pumps operating far from BEP.
Whenever possible, four quadrant data collected from actual testing of the pump in question should be used. Recognizing that this data is rarely available, it is critical that waterhammer engineers are aware of the impact assumptions embedded in common methods have on transient analyses.
Walters, Lang, and Miller, 2018 (10) provide all raw data files and more complete output such as pump speed decay and torque for all cases in this paper.
Below is an excerpt. Use the links above to view the full paper.
Part 1: Introduction
Some systems involve reverse flow through pumps during transient events. If sustained this can lead to reverse rotation. Predicting centrifugal (rotodynamic) pump hydraulic behavior and system response during transient events is critical to ensure safe design and operation. This impacts pipe pressure design, design of pipe structural supports, sizing and location of surge mitigation equipment, and guidelines provided to operators. Configurations susceptible to reverse flow through pumps fall into two basic categories:
- Parallel pump operation where pumps trip and at least one remains running
- Pumping to a higher elevation or pressure (e.g., a rising main).
The standard method to prevent reverse flow is the use of check valves usually at the pump discharge. Systems with check valves can have short-term, reverse pump flow before the check valve fully closes (e.g., see Lozano, Bosch and Walters, 2018 (1)). Some pumping systems do not or cannot use check valves, including pumping of slurries and large condenser cooling water systems. Moreover, some pump systems are purposely designed to have reverse flow through the pump such that the pump can run in turbine mode and be used to generate power (Binama et al., 2017, (2)). Systems without check valves often have power operated valves that may or may not close during a pump trip event, depending on the design.
Predicting pump behavior under reverse flow and potentially reverse rotation is a complicated task even with good data for a given pump. But good (or any) data is rarely available for reverse flow or rotation. Pump manufacturers perform rigorous testing of their pumps and publish performance data for head, power and efficiency in the zone of normal pump operation forward flow and positive rotation. Testing is rarely performed in the zones of reverse flow or rotation.
It was recognized as far back as the 1930’s that understanding reverse flow and rotation of pumps was going to be an important part of future engineering efforts especially in large water works projects under consideration at the time. Important first steps in this direction were made by Kittredge and Thoma, 1931 (3) and Knapp, 1937 (4). Progress in the ensuing decades culminated in the publications of Marschal, Flesch and Suter, 1965 (5) and Suter, 1966 (6). What emerged from these two publications is what we know today as the Suter Method of organizing four quadrant pump data into a dimensionless form convenient for digital simulation.
Part 2: Introduction
As described in detail in Part 1 (1), existing four-quadrant data is often used to analyze reverse flow and rotation in pumps when no test data for that pump is available. A four-quadrant data set that represents a suitably similar pump is selected based on specific speed, under the assumption that the similar pump will exhibit similar transient characteristics.
Creating the 4QDC requires a reference point to re-dimensionalize the four-quadrant data. There are two convenient and pragmatic choices as detailed in Part 1:
- The Best Efficiency Point (BEP), creating a 4QBEP curve
- The steady-state operating point (OP), creates a 4QOP curve
Because the two reference points are not likely to be the same, the choice of reference point has significant implications for both steady-state and transient simulations.
Additionally, it is not always clear what existing four-quadrant data set should be used. While using one of similar specific speed to the MC is advisable and most common, there are many from which to choose. Differences in individual pumps means that even a data set with matching specific speed may not be the best choice. The effects of varying four quadrant data sets are explored.
Due to this difference in applied characteristics, transient simulations can be dramatically impacted. The two 4QDC methods seem to diverge in their predictions the farther from BEP the pump is operating. It is very common to find pumps operating far away from their BEP as discussed in Part 1. This has many ramifications related to pump reliability and inefficient use of energy, but here we have developed and analyzed practical examples to demonstrate how it affects waterhammer simulation.