PMI presented at the Energy & Process CFD Symposium in Houston on May 18, 2017. The meeting featured presentations from industry experts discussing the use of simulation for process engineering, energy, and oil and gas applications.
At the symposium, Anthony DeFilippo delivered a presentation, “The Use of Complex Physics Models within the Refining Industry.” Three case studies were presented, highlighting ways that Porter McGuffie combines complex physics models when solving problems. Specifically, the cases look at multi-phase interactions modeled using Eulerian Two-Phase (E2P), thermo-acoustics modeled with Detached Eddy Simulation (DES), and structural response modeled with finite element analysis (FEA.) The focus of these examples was not on how they were carried out, but rather, how CFD-derived information was used to support engineering decisions.
Download Featured PowerPoint
The Sour Oil & Gas Advanced Technology (SOGAT) conference was held in Abu Dhabi on March 20-24, 2016. This gathering, which is focused on the gas conditioning issues that must be addressed to deliver sweetened gas for industrial development and infrastructure needs, has become the most prestigious meeting on this topic in the Middle East.
This conference can be considered the counterpart to the Laurence Reid Gas Conditioning (LRGC) conference held each year at the University of Oklahoma in the United States.
At SOGAT, Michael Porter, P.E. delivered a paper introducing a new analysis framework developed by PMI for the analysis and design of thermal reactors for sulfur recovery units. This new framework represents an entirely new paradigm in the use of CFD to analyze combined thermo/acoustic phenomena. Prior to PMI’s work in this area, these types of problems were only considered in research or specialized computational environments. In his presentation and the accompanying paper (linked with this post), Mr. Porter demonstrated that PMI’s techniques can reduce the computing time necessary to obtain valid solutions by several orders of magnitude over techniques currently used to study these phenomena. With this reduction in the resources required to perform these analyses, this category of coupled thermo/acoustic problems can now be undertaken for refineries and other industrial applications.
Link To Paper
Links to videos:
High-Rate Vibrating Flame
High-Rate Flame in Can
High-Rate Without Choke Ring
Turndown Rate With Choke Ring
On February 23, 2016, at the prestigious Laurance Reid Gas Conditioning Conference, PMI provided a “sneak peek” at the new framework that they have developed for the analysis of Sulfur Recovery Unit Thermal Reactors (TRs). Although the framework has been developed specifically for TRs, it is actually applicable for any combustion-fired unit. For the first time, the use of CFD is now available to model flow, combustion and acoustics on an industrial scale. PMI has uploaded two videos to our YouTube channel (channel link) showing the flame shapes for a condition that caused problematic vibrations in the field. The high rate with flame in can (link to video), in particular, allows the viewer to see the effect of an acoustic standing wave on the flame.
PMI will be presenting a paper on this topic in late March at the 2016 SOGAT Conference in Abu Dhabi, UAE. Stay tuned for additional information on this analysis technique following the conference.
High Rate Vibrating Flame Laurance Reid 2016
High Rate Flame in Can – Laurance Reid 2016
PMI presented at the Laurance Reid Gas Treating Conference this year.
Here is the Powerpoint
Link to References
By expanding an undersized baghouse, the Lafarge Roberta plant in Calera, Alabama was able to reduce daily cleaning cycles by a factor of ten and eliminate frequent bag changes due to excessive abrasion.
In 2002, the Lafarge Roberta plant built a new cement line that included a ten-compartment pulse-jet raw mill/kiln baghouse. The new baghouse turned out to be undersized. Lafarge consulted with GE Energy to correct the situation and improve the line’s efficiency.
In production, air volume was 22% over design capacity. The increased air volume pushed the air-to-cloth ratio to the limit, causing drastic increase in filter bag cleaning cycles and leading to bag failure due to flex fatigue. The increased volume also caused high velocity in the ductwork and in the hoppers, creating abrasion around the top of the bags. In the first three years of operation, two sets of filter bags (12 880 total) were used.
Analysis showed that the baghouse needed to be expanded. An additional 8% increase in air volume was also needed. Among the factors to be considered were how to redesign all the ductwork to maintain proper velocity and proper air and dust distribution. In addition, changes to the inlet baffling in the baghouse hoppers were required.
Paper Published by WorldCement
What is a Baghouse? (Wikipedia)
Viewing and manipulating large models requires significant compute resources. For post-processing FE and CFD results three areas control the amount of wall time required to generate each image: number of CPU’s, available system RAM and Video Cards. This post focuses on video cards with information provided from benchmarks performed by PMI. It is shown that the latest generation of video cards provide significant improvements in performance over video cards that are a few years old.
Based on initial tests that we performed for benchmarking our machines we found large differences in machine performance depending on the hardware that was available. Additionally, with the release of Star CCM 10.02, CD-Adapco has started supporting GPUs. With measurable performance differences between machines with different hardware and the new support of GPUs we decided to perform additional benchmarks to determine a good machine configuration, primarily for pre and post-processing of CFD models.
Description of Cards Used for Benchmarks:
- Nvidia FX3800 (workstation card) (http://www.nvidia.com/object/product_quadro_fx_3800_us.html). This is the primary video card used by PMI for the last couple years. The cards have provided acceptable performance, but we were starting to get persistent OpenGL errors, primarily with SolidWorks, Hypermesh and Star-CCM+. Based on internet research, we believe this is because the card only supported OpenGL 3.1. 4gb Of memory.
- Nvidia GeForce GTX 660 (desktop card) (http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-660). This is a great midrange video card with 2gb of memory.
- PNY Nvidia Quadro K4200 (workstation card) (https://www.pny.com/nvidia-quadro-k4200). This is a beast of a card with 4gb of memory.
The test model was about 34 million cells and had a file size of 15 Gb. The test started by opening a Geometry scene followed by a velocity scene. A mesh scene was then opened and the model was rotated through the same increments using standard view controls in CCM+.
Below is a table of the recorded times with each video card.
|| Version 10.04
||660 Windows Driver
||660 Nvidia Driver
|(Vel already open) Metal Temp Scene
|MT, rotate, +x+y+z to -x+y-z
|MT, rotate, -x+y-z to +x+y
CCM allows you to specify Nvidia Graphics or Windows.
The GTX660 with Nvidia driver would crash when trying to load the second scene. We could never figure out exactly why it wouldn’t use some of the pc’s resources for the left over memory needed.
You can see the K4200 is an amazing card. The initial load times are about the same but when it comes to rotation there’s nothing even close and yes that is 2 seconds. By the time I started the timer and looked up the rotation was complete. This is huge when exporting images to create a video. The code and sample video can be found HERE. Workstation cards are designed for specific types of work. In this case the workstation cards are a must.
When installing multiple cards make sure you are removing all of the old drivers. When installing the new drivers there is an option for clean install. In the test computer I used a software called Display Driver Uninstaller. Use this program at your own risk.
By, James Greeson
Apparently I still need to learn how these blog posts work, I think they’re not supposed to have abstracts…
This post is a medium length (14 pages) exploration of the differences in results that occur when a lifting lug finite element model is analyzed using linear and nonlinear techniques. From the results presented it is shown:
- The ASME BPVC is unclear on which stress limits should be applied to a non-pressure retaining component using linear analysis techniques, and
- For the geometry considered, the rated load for the lug is significantly increased when nonlinear techniques are applied.
The post is written at a mild technical level. There are no equations, but an understanding of stress analysis is helpful. Detailed modeling procedures, such as the selection of mesh density, are outside of the scope of a “short” post on nonlinear analysis and are not provided. Sufficient graphical information and discussion is provided to allow interpretation of the results. The post ends with a discussion of the pros and cons of linear and nonlinear analysis techniques, framed in terms of the lug analysis and PMI’s past experience.
By Sean McGuffie
PMI is frequently asked which tube arrangement is better for shell-and-tube heat exchangers with two-phase flow on the shell-side: square or triangular pitch? The short answer is: that’s an interesting question. To shed some light on the shell-side flow patterns within the tube bundle, let’s take a look at two models.
Geometrically, the models have the same percentage of flow space blocked by the tubes ~30%.
- The test sections are designed to model the flow characteristics, with no heat transfer, near the center of a typical bundle.
- The models are analyzed using two-phase flow physics with gravity.
- The two species considered were: water (r ~ 900 kg/m3, m ~ 0.2 cP) and steam (r ~ 3 kg/m3, m ~ 0.015 cP).
- The approach (superficial) velocity is 0.5 m/s for both phases, with a bulk light phase volume fraction of 37.5%. This bulk volume fraction is typical near the center of the bundle.
The pictures below show the heavy phase velocities from each model.
By the defined flow paths, channeling is apparent in both bundle types. A higher level of channeling occurs in the square bundle, as there is a direct path for the flow to travel from bottom to top.
The pictures below show the light volume fractions from each model.
As can be seen in the images, the volume fraction distribution in the triangular bundle is very regular, with blanketing of the light fluid on the downstream sides of the tubes indicated by the high volume fraction in the vortex downstream from the tube. In the square bundle the distributions are not as regular, but there seems to exist blanketing of the light fluid on the top and bottom of the tubes. The flow for this case cannot be characterized as steady, due to the irregularity of the volume fraction distributions. Click here to see a short-animation of the quasi-steady convection of fluid through the square bundle.
While in this case it looks like the triangular bundle would exhibit less blanketing and the problems associated with that flow condition, several additional factors must be considered:
- No heat transfer was modeled
- A single flow condition was analyzed, and
- The model was a simplified test section
Based on the small set of results presented here, it becomes obvious that a correct answer would need to address the following questions:
- What are the volume fraction and velocity distributions within the bundle, because:
- The flow is 3-dimensional
- Flow between sections along the exchanger must be considered, and
- The make-up water inflow into the bundle is highly dependent on bundle interaction with the shell
- Phase-change must be considered
- Where do 2-phase flow transitions (i.e., plug, slug, wispy, dispersed, etc.) occur, and
- How does the shell-side flow regime affect the local heat transfer coefficients
As can be seen above, another answer to the initial question could be – it’s complicated. The best answer to the question can be achieved through conducting CFD analyses on the whole exchanger. PMI has conducted more than 20 such analyses, primarily on heat exchanger equipment used in sulfur recover processes. For further information on these types of analyses please click the “Contact” button above.
By Sean McGuffie
In engineering, the damping ratio (ζ) is a dimensionless measure describing how oscillations in a system decay after a disturbance. One way to evaluate the damping ratio is to excite the system and then use what is called the log decrement method to evaluate the damping. Figure 1 illustrates the log decrement method.
Unfortunately, real systems don’t always behave in a way that allows clear use of the log decrement. Figure 2 illustrates the motion of a tall (~80 m) exhaust stack that is excited and then allowed to vibrate freely.
The vibration signal has been filtered to remove all but the natural frequency (~0.54 Hz) of the stack. If the stack was viscously damped, we would expect to see an exponential decay in the motion and the log decrement method could be used to evaluate the damping ratio. If coulomb (friction) damping controlled, we would expect to see a linear decay in the motion with time. Looking at Figure 2, neither type of decay is evident.
In order to get a clearer measure of the damping, it is often useful to use a technique that is often used in the acoustical world. If the vibration amplitude is plotted on a logarithmic scale, it can be shown that:
ζ = 0.183/n Equation 1
Where: n = number of cycles required for the amplitude to drop by 10 dB (a factor of approximately 3.16).
The motion data in Figure 2 has been potted with a logarithmic scale in Figure 3.
There are two distinct regions of decay in this portion of the signal. A straight line passing through the first decay segment indicates that the signal would take approximately 12 cycles to decay by 10 dB. A line passing through the second decay segment indicates that the signal would take approximately 20 cycles to decay 10 dB. Using the Equation 1, the respective damping values would be 0.015 and 0.009. The decreasing damping ratio as a function of amplitude is a strong indication that the damping in the stack is coulomb or frictional rather than viscous.
By: Mike Porter