Advanced Computational Fluid Dynamics Analysis in Control Valves

1. Acoustic Analysis

Valve acoustics prediction is well understood provided the valve geometry and valve trim conform to a small set of common geometries and designs. These prediction methods are well described in standards such as IEC 60534-8-3. Once the engineer ventures outside of these well-documented designs, though, he or she finds themselves in “terra incognita” with respect to prediction of sound pressure levels. Predicting noise from truly innovative designs is a very difficult endeavor, so it is necessary to go to a laboratory to run experiments to gauge the effectiveness of one design versus another.

Acoustic waves are compression waves travelling through a compressible medium at the speed of sound of the particular medium. If the full turbulent time varying pressure field can be understood, then it is possible to calculate the noise generated. The governing equations for acoustics are the same as the ones governing fluid flows. The downside of this approach is that it is very computationally expensive requiring a full RANS¹ + LES² model approach and a very fine mesh throughout the fluid domain in order to capture the pressure fluctuations.

Still with enough computer resources, it is now possible to calculate the sound pressure level produced by a valve trim. Two-dimensional studies show that reasonable results can be obtained for noise created by round jets, such as those found in low-noise drilled hole valve trim.

2. Thermal Shock Analysis

However, some high-temperature valves must also deal with the issue of thermal shock. For example, a bypass valve that is teed off the main steam line by some length of pipe in a steam system might remain closed most of the time and. During this “no flow” condition, the valve may cool down to a temperature significantly lower than that of the main line. When the valve needs to open (usually in a great hurry), steam at full temperature enters the valve and various components heat up faster than others, even if the expansion coefficients are the same between components and this can cause interference problems. The rate at which any one component heats up is dependent on the local heat transfer coefficient and on the local temperature difference between any valve component and the steam. These variables are driven by the local steam velocity across the component surface and the thermal boundary layer.

A common method to solve this is to use final element analysis (FEA) to conduct a transient thermal analysis. However, there is a major assumption in this analysis. Textbook estimates of heat transfer coefficients must be used, and these rely on a guess of fluid temperature and velocity, which means these estimates can be way off.

So, the only way to determine an accurate temperature profile in the valve is to perform a full time-varying fluid analysis. Fortunately, CFD can calculate the full temperature field inside the valve and can calculate the heat transfer coefficient with time to accurately determine the temperature of all the valve components. Software such as Ansys Fluent makes it possible to model the solid components as well as the fluid regime, which allows the program to calculate the conduction through the valve components and the thermal energy lost in the fluid. Ultimately the goal is to determine the time varying temperature of the valve components. These temperature results can be exported into a Finite Element package and the thermal strains can be computed in order to determine if enough clearance is maintained for good operation of the valve.

This issue is sometimes dealt with by installing a warming loop where a small secondary bypass line keeps enough steam flowing through the bypass line and valve to keep the valve close to the main line steam temperature. This effectively gets around the thermal shock problem but adds complexity and cost, and wastes energy. By getting a clear picture of the actual thermal displacements of the valve components, it is possible to create a reliable valve design without resorting to a warming loop.

Caveats: Steam valves that operate close to saturation, such as boiling water reactors, will immediately condense the steam as soon as it hits the colder valve components. This phase change from gaseous to liquid state extracts a large amount of energy from the steam (latent heat of vaporization). This manifests itself in heat transfer coefficients of two to three orders of magnitude greater than those of free convection. The author has not investigated whether CFD models exist that might deal with this phase change. It is important for the CFD engineer to be aware of situations like these that might cause errors in the results.

3. Multi-Phase and Cavitation Analysis

Valve applications with liquid flow sometimes experience multiphase flow primarily from out-gassing and/or operating too close to the local vapor pressure of the fluid, thereby creating cavitation and/or flashing. When a valve begins to cavitate, small areas inside the valve trim fall below the vapor pressure of the fluid, causing small vapor bubbles to form and then collapse again as the flow progresses to a region in the valve where the local pressure is above the local vapor pressure. This bubble collapse is cavitation and can be very destructive, especially if the bubble collapse occurs close to the body wall of the valve. If the cavitation is severe the body wall can be destroyed. Therefore, it is very important to know where cavitation occurs in the valve.

Another issue with multiphase and flashing/cavitating flows is that when the gas bubbles form, they occupy much more volume than the liquid phase, which causes a “closing off” of the flow passages inside the valve, thereby reducing the capacity. This reduction in capacity is very design-dependent so many valve manufacturers have their own internal calculating methods to account for this.

Modern CFD codes have a multiphase cavitation model that recognizes when the fluid drops below the local vapor pressure, and then implements the phase change as well as the momentum transfer between the phases. The result is an accurate picture inside the valve showing not only where the cavitation and flashing occurs but provides an accurate measure of the reduced capacity of the valve due to the application.

4. Valve Bonnet Cooling and Heating

With very hot applications (Figure 2) there is another issue that is harder to calculate. With the valve installed in the customary vertical position, thermal conduction and thermal drafts coming off the valve body and bonnet can create high temperatures around the yoke and actuator that might not be expected. Digital positioners and actuators typically have thermal limits lower than valve packing and might be adversely affected by the high temperatures.

This analysis is conducted in a similar manner to the previous “thermal shock” study except that free convection is the predominate method of heat transfer.

5. Pressure Wave Analysis

An interesting issue presented itself when a newly designed valve failed to operate as expected during laboratory testing. All sizing calculations were re-checked and design clearances verified and the conclusion was that the valve should perform.

The valve in question was cage guided, running in the flow-to-close (flow over the plug) direction. The cage had large rectangular ports to provide the maximum flow for this fast stroking valve. This was all very standard protocol. As soon as the valve began to open, however, an unknown force would violently lift the plug to almost full open (Figure 3). One of the engineers remembered a similar problem from a few decades ago and found a report detailing how the problem had been solved, but the reason (physics) this occurred was still unknown.

The engineers decided to undertake a transient, compressible CFD analysis of the full valve trim to see if the problem could be reproduced. This required a very fine mesh and very small time steps in order to capture the pressure wave as it travels at the speed of sound through the medium. The results showed that a combination of the high-pressure drop, the sudden opening of the ports and the fast stroking speed produced a shock wave that travelled from the ports towards the centerline of the valve beneath the plug. As the shock waves from all the ports converged, a short-duration, high-pressure bubble formed under the plug with more than enough magnitude and duration to accelerate the plug upwards. Understanding the physics of the phenomenon allowed for modification of the design and then re-testing using CFD was performed to see if the pressure bubble effect had been reduced.

6. Detached Eddy Simulation

CFD can measure that pressure fluctuation, as shown in Figure 4.

7. Transient Moving Mesh

Time history of forces and moments acting on the valve spindle can be predicted with transient moving mesh CFD analysis. Such capability helps the designer optimize the valve geometry and ring adjustments to meet the design specifications.

Summary

Many of these CFD models were developed for specific applications and have been incorporated in commercially available codes, creating very powerful additions to the analyses. It is up to the valve engineer to be aware of the current state of the art and read up on the additions to new releases and see where new CFD technology can be used.

Editor’s Note: Dr. Homayoon Feiz presented this paper on behalf of Asher Glaun P.E. at the 2015 Valve Manufacturers Association Technical Seminar & Exhibits. They are both engineers with GE Oil and Gas. The presentation was re-purposed for VALVE Magazine by Kate Kunkel, senior editor. You may reach Kate at kkunkel@vma.org.

¹ RANS: Reynolds average navier-stokes turbulent model

² LES: Large eddy simulation model