S2M Makes the Transition to HIL
S2M uses Opal-RT Hardware-in-the-Loop technology to perfect magnetic bearing-based rotor technology used in deep sea natural gas extraction
Originally published in the May 2009 issue of Mesures Magazine (France)
Integrated with a turbomachine, the purpose of S2M’s magnetic-bearing-based rotor is to accelerate natural gas extraction from a deep-water Norwegian gas field. The size of the S2M rotor is stunning: 4740 lbs with a length of less than 20 feet, but S2M have been masters of this technology for a long time. The new challenge for control system development teams at S2M is to design a controller while the actual system is not physically available. To address this challenge, S2M has resorted to HIL (Hardware In the Loop) validation and testing for the first time.
This is not quite 20 000 Leagues Under the Sea, but the technical difficulties of this project still pose a number of challenges. The Norwegian Ormen Lange gas field is located in deep waters, with the actual extraction facilities installed at seabed, 3300 feet below sea level. Underwater gas field exploitation without traditional offshore platforms is said to be more cost-effective, but in this high-risk environment, explosions are a constant threat and it is not possible to use ball-bearing compression systems. Furthermore, the limited rotational speed of 10 000 rpm for ball bearings does not meet the specific requirements of this application. Therefore the development of S2M’s magnetic-bearing and high-speed motor technology was delayed.
Founded in 1967, Vernon, France-baased S2M is a world leader in designing and manufacturing of magnetic bearings and permanent-magnet high-speed synchronous motors. The company was acquired by Swedish bearing and mechatronics giant SKF in 2007. One of S2M’s particular areas of expertise is the substitution of ball bearings with magnetic bearings, thereby keeping the motor shaft in a magnetic field. In practical terms, the torque generated by the motor is applied to the rotation of the shaft. On either side, radial and axial magnetic bearings generate a force that corrects the shaft position. Friction-free, the motor can then reach speeds of 12 000 to 300 000 rpm, while the rotational speed needed by turbocompressor shafts does not exceed 70 000 rpm.
S2M are experts when it comes to magnetic bearing technology, but now face new challenges. In the case of the company’s Ormen Lange project, the size of the turbocompressor rotor used is remarkably large: 4740 lbs for a length of less than 20 feet, and will not be physically available during control system development. Indeed, as it is often the case, mechanical components require a longer development time than their electronic and software counterparts. Hence, S2M was forced to carry out the early stages of development as well as all control system refinement long before the rotor was ready and available for a final test. A preliminary test on a 1/10th-scale model is scheduled as part of the validation process, which will end with a full-scale test prior to on-site installation in 2015. Once installed, the system will not require maintenance for four years. This implies redundancy of the controller, as well as consequent testing ensuring electronic and software reliability. The magnetic bearings are guaranteed against rust and will not require maintenance for thirteen years after installation.
First HIL Experience
“Until now, we have used machine prototypes to test magnetic bearing control systems to ensure that control loops work properly. For this project, we did not have the material means to complete the validation process. It was the first time that we tested the controller first on a virtual machine, then on a small-scale model”, explains Lakdar Sadi-Haddad, R&D engineer at S2M.
A solution that met the three following objectives had to be found. The first was to find a homogeneous development environment that would allow for modeling and simulation of control laws, power amplifiers and the physical active-magnetic-bearing based rotor (MIL, Model-in-the-Loop) to complete functional validation of the system. The second was to have a rapid-control prototyping solution for testing new control algorithms on an existing machine to complete functional validation of the code before integration with the final controller (SIL, Software–in-the-Loop). The third objective was to be able to substitute the real machine with a virtual one that could reproduce its behavior exactly. This would allow for functional verification of the final controller, which is in turn connected to a simulated version of the physical system under control. All of this suggested mathematical modeling of the whole system and use of HIL (Hardware–in-the-Loop) simulation for the validation of a complex control system.
The Ormen Lange underwater gas field
Discovered in 1997, Ormen Lange is located in the Norwegian Sea, about 120 km from the Norwegian coast. This 40 km long, 10 km wide gas field is the second largest known reservoir on the Norwegian continental shelf. Production started in October of 2007 and gas produced is transported to the United Kingdom via a 1170 km subsea export pipeline. Production should increase gradually, to a maximum exportation volume of 70 million cubic meters of natural gas per day. Because the field is located in deep waters, very advanced extraction techniques were required. Offshore platforms are not used since subsea wellheads are connected to an onshore infrastructure. Turbomachines equipped with magnetic-bearing-based rotors speed up gas extraction.
The HIL validation experience was new to S2M and special tools specifically suited to this application needed to be chosen. In the case of modeling environments, MATLAB/Simulink from The MathWorks Inc. was already in wide use across the company and was therefore an obvious choice. Indeed, MATLAB/Simulink is considered the de facto standard modeling environment in a large number of industries and by engineers and researchers worldwide. “I was indeed familiar with MATLAB/Simulink”, confirms Lakdar Sadi-Haddad (S2M). This software enables a variety of matrix calculation types to be carried out, such as fixed point arithmetic and complex system validation by flexible modes, vibrational control in rotor coordinates and response to perturbations during increases in speed.
The rigid rotor model was relatively easy to find, and took only a month to generate. However, constructing the flexible model that would take speed-induced strains into consideration was far more complicated. S2M’s Computation Office took the lead in addressing this and successfully provided a dynamic model with speed-adaptive vibrational modes. S2M’s objective was to have a homogeneous development process, where mechanical and electrical development departments would not be required to rework components at every stage of development. “In this regard, MATLAB offers the ideal interconnection between different models”, added Lakdar Sadi-Haddad (S2M).
The active magnetic bearings (AMB) keep the motor shaft in a magnetic field. One of S2M’s missions was to develop the control system of this mechanical set-up.
Real-Time Simulation Needs
The project faced a number of important time constraints. The modeling and simulation HIL platform needed to support a 10 µs time-step. During this extremely small execution cycle, the virtual machine replicating the behavior of the rotor and four magnetic bearings must compile 14 analog acquisitions (analog/digital conversion of the control inputs), execute the model, then produce 7 analog outputs (each digital/analog conversion corresponding to the position of the rotor in each of its 7 axes).
During the same time-step, the controller must manage 7 control loops (corresponding to the system’s 7 degrees of freedom), 7 analog/digital conversions (position measurements), and 14 power commands (commands of the power amplifiers).
To execute the loops at such high speed, considering the amount of data that must be transferred, S2M required a testing and simulation solution that would meet the high-performance requirements of their project. In the end, S2M’s choice was the solution provided by Paris-based Viveris Technologies. This industrial computing company specializes in embedded systems testing & validation solutions, and is the exclusive distributor of Opal-RT Technologies’ products in France and Spain. Based in Montreal, Quebec, Canada, Opal-RT develops open, PC-based Real-Time Simulators and HIL testing systems for electrical, power electronic, and mechanical systems.
“Opal-RT provides a wide variety of solutions to assist modeling and simulation throughout the design process. Opal-RT simulators enable the engineer to detect errors faster, reduce reiterations of the development cycle and optimize tests. In addition to distributing Opal-RT simulators and associated software, Viveris offers project support and turnkey solutions. The idea is to understand the needs of the clients, right from the beginning, and to provide them with the tools and support to enable them to overcome the challenges they face in their development projects. Our research departments are also involved in aspects of technology development relating to Opal-RT simulators and the company’s RT-LAB simulation software platform”, reports Thierry Caldairou, Project Manager at Viveris Technologies.
Opal-RT develops a complete line of Real-Time Simulators based on RT-LAB, the company’s real-time simulation software platform. In addition to being fully integrated with MATLAB/Simulink, Opal-RT simulators have an open architecture that takes advantage of widely available commercial-off-the-shelf technologies, including INTEL multi-core processors and FPGA-based input/output devices. RT-LAB enables integration of models coded in a variety of environments including MATLAB/Simulink, AMESim, Modelica/Dymola, ADA, and C. In particular, MATLAB/Simulink users are able to generate code that can be executed in the RT-LAB environment. The Canadian company also offers drivers to interface the simulator with a wide range of Input/Output cards. “Many commercial input cards can be used on the RT-LAB platform, but none provided the performance we required. As a result we chose Opal-RT’s, which are based on FPGA components”, clarifies Lakdar Sadi-Haddad (S2M). Analog input and output cards were integrated with Opal-RT’s HILBox industrial PC, which acted as the target on which MATLAB/Simulink models were executed. The HILBox is responsible for the real-time simulation of the power converters and the behavior of the rotor and bearings.. After modeling is completed, the controller communicates with an S2M-developed card, where the appropriate algorithm is implemented. The control card was developed in parallel with simulations carried out on MATLAB/Simulink.
The controller designed by S2M was validated using Opal-RT’s HILBox industrial PC, responsible for real-time simulation of MATLAB/Simulink models of power converters, rotor and magnetic bearings. The controller will also be tested on a 1/10th scale model.
Youssef Belgnaoui
Translation: Rémy Bélanger & Daniel Coyle
