Pumping It Up…or How We Built a Better ESP

Date: 07/05/2011

David EslingerThey may seem like simple devices, but the multiple mysteries of fluid mechanics make building an efficient Electric Submersible Pump (ESP) one of the more daunting challenges of our business. Schlumberger Engineering Advisor David Eslinger tells how Schlumberger-Reda has used the last decade to address these technical conundrums and surpass the competition by producing the best ESP in the industry.

Ten years ago, Schlumberger Artificial Lift’s market share was eroding because the performance of many of our Reda ESPs was declining versus the pumps of our competitors, Baker Hughes Centrilift and Wood Group ESP. To understand Reda’s decline and recent re-emergence as a leader in the Artificial Lift domain, one must follow the fascinating evolution of the design tools and testing techniques used to develop new pump stages—a single “stage” consisting of a rotating vaned impeller and a stationary vaned diffuser.

But before we get ahead of ourselves, I should explain that an ESP is a centrifugal pump with scores of stages driven by a downhole motor. That motor is powered by electricity through a cable from surface. Inside the ESP, rotating impellers impart angular momentum to fluid passing through the pump and the fluid angular momentum is converted to a rise in static pressure by fluid diffusion in the impellers and diffusers. The rise of static pressure in an ESP “lifts” under-pressured wellbore fluid to the surface. The flow of fluid through a pump stage is exceedingly complex because the stage commonly contains zones wherein the fluid may be recirculating or possibly stalled, rather than uniformly following the vaned passages. Exact analysis of this flow requires solving a mathematical equation for three-dimensional turbulent fluid flow, known as the Navier-Stokes equation—which is not something you can do with hand calculations, or with a hand calculator!

The Beginning
From the inception of the ESP business, undertaken by Russian inventor and Reda founder Armais Artunoff in the 1920s, centrifugal stage design was based on simple one-dimensional (1D) fluid momentum analysis which ignored the effects of fluid viscosity. The shapes of the impeller and diffuser vanes were established using 1D calculations along one to three passage streamlines assisted by various empirical “correction factors” and rules of thumb. These empirical factors were carefully guarded as trade secrets by pump companies and individual “hydraulics” engineers. (Decades ago, a retired Reda chief hydraulics engineer reportedly instructed his wife to burn his “black book” of stage design rules after his death.) The actual geometric layout of a vane involved multiple hand-drawn 2D projections to establish the actual 3D vane shape. The final layout was drawn on wood boards. Multiple sections were cut out, stacked and sanded to make the “master” vane pattern for part casting. Evaluation of the design consisted of a performance test, wherein water was pumped with the newly designed pump.

The trouble with the above development cycle is that it was so time consuming (up to two years) that the first design was either commercialized or scrapped. Multiple design iterations were simply not feasible. “Good” pump stages were replicated into larger and smaller pump diameters by scaling. Amazingly (and a tribute to the skill and “black books” of the hydraulics engineers of the period), several stages designed during this era are still in today’s Schlumberger Reda catalog.

Digital Design Arrives
The rapid expansion of numerical computing capability, starting in the 1960s, enabled a progression of game-changing turbo-machinery design tools. Batch and terminal computing enabled quasi-3D analysis which took into account the curvature of the stage flow passages. The widespread use of computer-aided design and manufacturing (CAD/CAM) in the 1980s enabled computer-based vane layout (still based on multiple 2D projections) and digitally controlled milling of precise metal tooling for stage casting—thus ending the era of wood patterns.

In the mid-1980s William Dawes at Cambridge University developed the first commercial computational fluid dynamics (CFD) software which numerically solved the 3D Navier-Stokes equation for turbo-machinery—and for the first time the pump designer had a reasonable chance of numerically predicting the performance of a stage without actually making tooling, casting stages, and testing a full pump. By the mid-1990s, pump stage design had made significant advances but development was still risky. Less than one quarter of all stages designed by Reda using CFD and tested as full pumps were commercialized. Probably this was due to two things. First, few design iterations were evaluated since the creation of new 3D vane geometry (even with CAD) was slow and cumbersome and the manual creation of a CFD mesh could take days. Second, the accuracy of CFD predictions was spotty due to the use of simple turbulence models, poor meshes, and limited computing power. Due to the lack of success of CFD-designed stages, Reda abandoned use of CFD by 2000.

Back to the Drawing Board
In early 2003 I was offered an opportunity to join my former Coiled Tubing Tools manager, Pat Bixenman, then engineering manager at Schlumberger Reda, to work with a team that was taking a fresh look at ESP stage development methodology. It was an exciting opportunity to make a significant impact at the Segment level—and to return to the field of pump technology after spending the first 10 years of my career designing positive displacement fracturing pumps.

A review of the state of the art in turbo-machinery design revealed that game-changing software suites were available, having first appeared in the late 1990s. These suites integrated 1D through CFD analysis tools with graphics interfaces that enabled rapid 3D vane creation and evaluation of stage design performance. Using these new tools, the hydraulics designer was freed from the tedium of creating vane geometry and generating CFD meshes, and could instead focus on the business of creatively managing the flow field to deliver improved stage performance.

However, even with rapid design and analysis there was still a significant weakness in the new stage development methodology: Feedback from test results still lagged design creation by six months because the new stages had to be cast before testing. The solution to this problem was to use commercial rapid prototype machines to print the stages in plastic and then to test them with air as the pumped fluid rather than water. Since the kinematic viscosity of air is 15 times that of water and viscous effects play an important role in stage performance, air testing was made “similar” to water testing by up-scaling the plastic test stages and increasing the test rotational speed. Fortunately, benchmark air tests on several legacy stages demonstrated that stage performance predicted by air testing closely matched actual water results. This was a real breakthrough, since it means that test feedback for a new design can now be available in two weeks rather than several months. Further, flow field measurements can be taken during air testing to calibrate CFD predictions and to locate stage performance deficiencies. The air test data for each new design is also used with an “inverse” 1D analysis to populate a database of Schlumberger-specific “correction factors” so that subsequent designs have a smart starting point based on previous successful stages.

Little Madness in This Method
Today the basic envelope of a new Schlumberger Reda stage design is established in two to three days by hundreds of 1D analysis iterations using a database of “correction factors” determined from air tests of past successful stages. Each iteration requires only a fraction of a second of computer runtime. The detailed 3D vane geometry is then designed over a few weeks by scores of CFD iterations, each automatically meshed and requiring 30 minutes to a few hours of runtime. Two or three stage iterations are printed in plastic and air tested before the geometry of the new stage is finalized and prototype castings are made.

A complete stage development cycle still requires close to a year because the manufacture of prototype castings is time intensive, but so far all seven stages developed using the new design methodology have been commercialized. Most importantly, the performance of each of the new stages has been a step-change when compared to that of the legacy Reda stage it replaced—and when compared to our competitors’ stages. For example, the Schlumberger Reda SN8500 stage, designed in 1995 using CFD, rendered a 15% pressure improvement and an 8%-point efficiency improvement versus the legacy Reda GN7000 stage designed using only 1D analysis. The Schlumberger Reda S8000N stage, designed in 2007 using the new design methodology, gave a 46% pressure improvement and a 2%-point efficiency improvement versus the Schlumberger Reda SN8500, which it replaced, and a 21% pressure and 4%-point efficiency advantage over the best comparable competitive stage.

The next step changes in stage performance and in development time reduction will likely come by integrating automated optimization into the design methodology. This will not result in “push button” pump stage design since the design space of a turbo-machine is much too large for computers of the foreseeable future to explore. However, just as designer-driven CFD design refines (not replaces) today’s initial 1D design, optimization-driven CFD will refine designer-driven CFD designs. Beyond just providing a single point “optimum,” these optimization-driven routines can generate a surface response to a range of design variables. These “response surfaces” will give the hydraulics engineer more precise insight into how to manipulate the flow field—much as 1D analysis “rules-of-thumb” helped hydraulics engineers of old. This will guide subsequent stage designs and shorten future development cycles.

Besides enabling a step-change in ESP performance, the new design paradigm has brought additional benefits. First, since much of the flow analysis overhead and manipulation of 3D vane geometry is computerized, today’s stage designer can spend a significant portion of time creatively managing the stage flow field and exploring innovative concepts. Second, today’s “hydraulics” engineer is a user of commonly shared analysis tools and a contributor to evolving and shared design databases and “best practices” rather than a Merlin guarding rarely shared information. This means that several designers can contribute at a high level and greater numbers of new stages can be commercialized.