Active magnetic bearings (AMBs) are helping to make subsea gas compression possible. Joaquim Da Silva, François Carrere and Frédéric Ponson of SKF Magnetic Mechatronics, explain the engineering process involved to create the AMBs needed for this complex project.
 |
|
Collaborative design interdependencies (network). Illustrations courtesy of SKF. |
Subsea gas compression enables both improved energy efficiency and lower costs to help offshore operators to recover the remaining reserves in the existing gas fields.
To enhance efficiency in this application, it is now possible to operate and control the rotation of the subsea compression units using AMBs.
AMBs offer a number of advantages, including high performance, a small footprint, high reliability, and advanced monitoring capabilities suitable for unmanned installations.
However, the application of AMBs within compression units requires a long and complex development and validation methodology involving several multidisciplinary engineering teams, comprising the AMB supplier, the compressor original equipment manufacturer, engineering, procurement and construction companies, and the end user.
To simplify development, traditional “black box” specification should be replaced by a collaborative design approach.
Collaborative design dynamics can be modeled as a kind of distributed network with nodes, each representing the design stakeholders. Each participant is responsible for several design requirements, solutions and decisions, all of which are highly interdependent. The dynamic of this network is strongly affected by the relationships between design participants. Requirements and interdependencies must be well defined, particularly the sharing of technical communication/information between the different members of the project, especially by means of appropriate modeling/simulation tools that are able to loop on the requirements issued by the stakeholders.
The project development phase was deployed using M2S2 (Multiphysics/Multiscale System Simulation). This methodology provided all project stakeholders with behavioral models, each a compromise between the details of physical behavior and complexity.
The Ormen Lange pilot project
 |
|
Subsea Compressor Module Architecture - AMBs/Controller.Image courtesy of Aker Solutions and GE Oil & Gas. |
The subsea compression station at Ormen Lange comprises a “main module” equipped with two or four 12.5MW subsea compression units. The integrated vertical motor-compressor (max speed 11000 RPM) is equipped with AMBs (three radial bearings and a thrust bearing) linked to a variable speed drive, a separator module, a cooler, and a liquid pump.
The motor compressor is equipped with three main active magnetic bearing modules (cartridges) where one of the challenges was to use canned bearings for protecting the stator coils (bearings and sensors) from any contaminant present in the cooling gas. One of the issues to be addressed was the validation of predictive behavior of the thrust bearing and auxiliary bearing during the production life cycle of the machine; constraints and specification of axial load during rotor drop and external force perturbation were given by the compressor maker as well as the axial bearing load capacity of 45kN, whereas the radial bearings 1 and 2 had a 16kN load capacity and the radial bearing 3 had a 7kN load capacity.
The designers faced a wide range of challenging physic fields, including mechanical, electrical, magnetic, thermal and many others where the electrical time constants (< microseconds to milliseconds) are usually much smaller than mechanical ones (several milliseconds to seconds). Taking into account all these phenomena in numerical simulations requires using extremely detailed models, and different types of solutions resulting in a calculation process that is prohibitive in terms of both cost and calculation time. M2S2 provides an answer to this problem. It allows the superposition of the different physical and scale domains, providing both an overview of the system and the ability to zoom in on specific areas to predict system behavior.
Backup bearings can be considered as an “airbag” for machine integrity. During rotor drop, part of the “mechanical phenomena” that occurs is measured in the range of the system’s spatial scale, particularly in terms of rotor/stator interaction. Process and rotor/stator perturbations are applied in real-time to backup bearings, generating extremely complex phenomena (friction, deformation, heating) inside the bearing itself, all occurring within very different spatial scales that range from micrometers to centimeters.
An SKF in-house tool called BEAST (BEAring Simulation Tool) provides a fully geometrical 3D model incorporating tribology and contact mechanics (which also takes into account the effects of small-scale geometric variations, such as surface roughness). BEAST is based on multi-body techniques with special focus on local contact problems and allows for studies of dynamic behavior of all bearing components to be carried out under general loading conditions. Principal output data from BEAST relates to the movements of all bearing components, the contact forces between the components, and the forces with the environment. BEAST also provides detailed data from the contacts regarding smearing power, lubricant film thickness (of any type), pressure distribution, slip-speed distribution, transient temperature distribution, wear and many other factors.
 |
 |
|
Simulation results and drop test data comparison (position signal K1 bearing). |
Workflow of the Backup Bearing function design using Multiphysics/Multiscale Simulation. |
Tools validation
Tools validation was achieved by building a reduced scale compressor test-bench, carrying out a landing test campaign, and finally comparing the data collected with the simulation results. The shaft of this test-bench (200kg) was designed for supercritical operation and therefore needed to be flexible. In contrast to the full scale compressor, the test-bench rotor was configured horizontally, driven by a motor on the shaft end bearing. Bearings 1 and 2 were flexible mounted angular contact bearings that act in radial direction only, while bearing 3 was an angular contact bearing supporting radial and axial loads.
During the landing test campaign, the axial loads were generated by the magnetic thrust bearing, situated between bearings 1 and 2. This axial load was applied in direction of bearing 3 and immediately generated a forward whirl.
Validation was performed by comparing simulation results with experimental data collected during the landing test campaign. Comparisons were made on the following quantities; “orbital shapes” correlation, “spectral content of the position signal” and “whirl on set time” during the rotor deceleration phase. In both cases there is a peak at about 110 Hz, which is the frequency of the whirl trajectory seen in the orbit. The previous results showed an extremely good match between the simulation and experiment.
Once tools were validated, the designs were analyzed. The most relevant parameters were the pressure and the smearing power for the three backup bearings. The simulation results showed excessive pressure and smearing power specifically on the backup bearing 3 (mainly due to the high radial loads induced by shaft bending). This observation was particularly important, especially taking into account the extreme high peak values that are inaccessible for most multi-body or FEM commercial codes. This is only possible if the code has a special focus on local contact problems.
To validate and quantify the impact of different design parameters, the previous results showing excessive internal bearing pressure and smearing power were compared with experimental data obtained during a test campaign of cumulated 20 landings. Evidence of internal damages that could be attributed to the high values of pressure and smearing power predicted by this advanced simulation tool.
 |
|
Hertz pressure and smearing power simulation results (deceleration phase + variable axial load). |
Full scale compressor pilot
When the motor-compressor pilot unit was tested at full speed and full load, five landings were successfully performed and no damages to the rotor, seals or auxiliary bearings were recorded. The observed whirl direction and amplitude matched the simulation prediction 100% (<75% clearances). The post-analysis performed by the compressor maker on auxiliary bearings confirmed no external/internal damages. The information sets and knowledge gathered during this study helped SKF establish the first generation model for backup bearing life prediction, called Landing Life Model L2M, a tool for advanced bearing health assessment and a program of condition monitoring tool development that is now under progress.
Dynamics response of thrust bearing
 |
|
Multiphysics thrust bearing system simulation/optimization using “Reduced Order-Model”. |
Much has been written on the modeling of non-laminated (solid) axial bearings, including eddy currents induced in a solid magnetic frame; however, in most cases the proposed model is defined only in frequency domain and is only valid regarding small air gap variations. Even if it is very helpful to use these simple analytical models during the development phase of a new axial magnetic bearing, they do not allow the time response of the thrust bearing while facing a rapid transient and high amplitude force to be predicted. These types of disturbance are usually encountered when the compressor is operating near the surge line or in other abnormal process conditions. In case of high speed centrifugal compressors of several MW power, some extreme conditions can cause variations of several tens of kN in only a few milliseconds.
One answer to this problem is using a “co-simulation concept.” This method involves coupling finite element analysis software with a multi-domains dynamic system simulation software (such as Simulink, SIMPLORER). This method requires powerful computation and simulation time is usually very long.
The solution currently proposed and applied to the design of thrust magnetic bearing used in Ormen Lange motor-compressor is based on an “identified reduced order-model” taking into account following physical quantities (identified from FEM model);
The proposed method is based on advanced nonlinear fractional model identification:
Step 1 - Model structuration using physical system knowledge: This model uses geometry segmentation to generate an equivalent reluctances network. Elements to consider are; geometry type (E, C, hybrid shape), canned or un-canned bearings, impact of the magnetic environment (frame), magnetic materials characteristics, and large air gap.
Step 2 - Nonlinear fractional identification: The geometry segmentation is used to generate the associated reluctance matrix. All equivalent inductances are then calculated by inverting the reluctance matrix combined by windings properties. For each impedance element (i) the fractional model identification process has to be performed (eq. 1 left-side expression) where the fractional exponent is the analytical solution of the diffusion equation for a semi-infinite medium. This parameter is used as the optimization parameter during the identification process.
Step 3 - Synthesizing: Once the fractional transfer functions are identified, it is necessary to synthesize them as an integer order. Real (integer) systems typically have non-integer behavior over a given frequency band. The non-integer integrator is in fact limited to a defined frequency range (fmin-fmax). The synthesis of non-integer order integrator is then based on a recursive distribution of a finite number of poles and zeroes (eq.1 right-side expression). The nonlinearities such as large air gap variation and high current level inducing a large magnetic saturation are taken in account. These physical quantities have a significant impact on the representativeness and model accuracy.
Results obtained using identified nonlinear fractional model
The total magnetic flux response to a maximum step of current (with fixed rotor position 50% of the total air gap / canned bearing) was recorded, as was the position response to a profile of a specific external force perturbation. Data showed that the total magnetic flux converged with the same dynamics of its steady state value, while the position reacted properly to the external force perturbation.
These results showed that the method proposed in this paper predicts with great accuracy the dynamic and static response of the thrust magnetic bearing. Moreover, one of the most interesting benefits is the simulation time efficiency, which opens the way to implement this model in a Hardware-in-the-Loop process for real-time simulation purpose.
 |
| Left: FEM versus “Identified Equivalent Reduced Order-Model” simulation results. Right: Model identification workflow for a very simple axial bearing geometry. |
Conclusion
The collaborative design approach applied to oil and gas motor-compressors equipped with AMBs, in particular for subsea application is a powerful and efficient tool to collect and consolidate the specification of the project among the different project partners and key players. By limiting iterations in design as well as identifying clearly the technology readiness levels needed for each of the single components and subsystems, it offers the opportunity to define better design specifications, reduce the time for testing and qualification and ultimately reduce the cost of development of the project. An approach based on numerical models called M2S2 could become one of the cornerstones of the collaborative design process when applied to AMB solutions for the oil and gas industry.
Two examples have been presented to illustrate this approach. In both cases, the outcomes of M2S2 tools are “super models” that can be transferred to the partners to enable them to better assess the key working load parameters impacting on the robustness of the complex system and consequently provide a finer understanding of the design parameters and their tolerances. M2S2 achieves this by significantly improving the flow and share of information between partners.
Collaborative design offers a new perspective on managing projects for complex oil and gas projects. We also believe that it illustrates the nature of future relationships and new business models for oil and gas compressors equipped with AMBs.
Joaquim Da Silva serves as head of research and technology at SKF Magnetic Mechatronics, France. He received a master’s in electrical engineering from the Conservatoire National des Arts et Metiers (CNAM), France and a PhD in control and signal processing from the Univesité d’Orléans, France.
François Carrere is a technical leader for SKF Magnetic Systems, covering subsea projects like Ormen Lange Pilot project and the Asgard Compression Station. He has over 30 years’ of industry experience. He earned a doctorate in electronic and automation in 1982.
Frederic Ponson serves as engineering manager for SKF Magnetic Systems. He has a MsC in mechanical engineering and heat transfer. Since 2009, he has been in charge of product engineering for magnetic systems for industrial, and oil and gas applications.
Subscribe
