Endurance test for high-voltage motors with magnetic bearings

Paul Boughton

High-speed synchronous motors with with magnetic bearings are proving useful in driving compressors used in a Dutch natural gasfield.

At speeds of over 3600 rpm – for instance involving large turbo compressor applications – high-voltage motors with conventional sleeve bearings frequently reach their limits. In these cases, as an alternative, active magnetic bearings can be used. In addition to their extremely wide speed control range, this technology is also in compliance with the toughest of environmental legislation, as oil is not used - and of course oil collection pans and oil supply systems are eliminated. Although this bearing technology is still in its early stage for high-voltage motor applications, perfect operation in plants and systems can be guaranteed using simulation models and confirmed using a so-called ‘landing test’. Such a test for a high-speed synchronous motor with magnetic bearings is illustrated in the following example. The synchronous motor involved is in the multi-megawatt class and drives a compressor in a natural gas field in The Netherlands.

The largest natural gas field in The Netherlands is located south of the city of Groningen. It is one of the largest natural gas fields in the world. It is operated by the Dutch Nederlandse Aardolie Maatschappij BV company, or briefly NAM.

For this particular field, sand, water and other foreign bodies must be filtered from the gas before it is pumped to users through a pipeline. This process results in a drop in the pressure, which must be compensated for in the event that the natural pressure reaches a specific lower limit.

Turbo compressors from Siemens Power Generation Industrial Applications are used to boost the pressure. These turbo compressors are driven by Siemens high-voltage motors manufactured in the Dynamowerk Berlin. When designing the motors it was seen that the optimum solution was to use Active Magnetic Bearings (AMB) due to the wide speed control range of 600 to 6300rpm required for the pumping process.

The motor is a high-speed, two-pole synchronous motor with a power rating of 23MW. The motor is fed from an LCI converter, has a horizontal type of construction (IM7211) and is water-cooled. The motor has degree of protection EExp. Pedestal bearings and stator are mounted onto the baseframe. The motor has three bearings and the brushless excitation generator is located outboard of the NDE main bearing. The rotor weighs 9.2t and the complete motor 66t.

The characteristic feature of this motor – the Active Magnetic Bearings (AMB) – is implemented as follows: The laminated AMB rotors together with the precision machined surfaces for the electronic sensors of the AMB are located on the shaft at the bearing positions. The AMB stators with sensors are arranged around these.

Back-up bearings are provided directly next to each of the three magnetic bearings just in case the AMB closed-loop control was to unexpectedly fail.

These back-up bearings have an air gap of about. 0.5mm while the air gap of the magnetic bearings and the sensors is about 2mm. Further, the back-up bearings are dry-lubricated; the sleeve bearing shells have a coating of a special material and are split. This means that if required, they can be replaced without having to disassemble the complete stator/rotor of the motor.

The corresponding sleeves of the back-up bearings on the shaft are provided with a galvanic anti-adhesion coating. This coating prevents that too much material being transferred from the sleeve bearing shells to the sleeves when the back-up bearings are being used.

The dynamic rotor characteristics of the motor had to be designed to take into account both normal operation – ie the behaviour when using the magnetic bearings – as well – although most unlikely – the behaviour if the magnetic bearings were to fail. This behaviour was verified in the design phase using simulation models. The complete simulation system included the rotor model with the three-coupled magnetic bearings and the back-up bearings. Especially high demands were placed when modelling the back-up bearings as this model is used to evaluate the safety-related issues and is used to predict what would happen if a fault were to develop. The model parameters must be defined in close cooperation with the manufacturer of the magnetic bearings. All of the most important dynamic rotor data can be simulated and predicted using this model. This data includes the natural frequencies in the magnetic and back-up bearings, stability analyses, steady-state behaviour for defined imbalance conditions as well as the simulation of transient failure scenarios for the magnetic bearings. This allows the system to be further optimised.

The predominant objective of the landing tests was to check the simulation results. In order to fine-tune the model parameters, it was mandatory to carry-out a limited number of landing tests. This also included determining the stator braking torque because only then was it possible to simulate the tests. The run-down time, the shape of the bearing orbit as well as the shaft acceleration levels in the area of the rotor end caps were important measured data that could be directly compared.

With a landing test, the magnetic bearings are deliberately de-activated while the rotor is still rotating. The rotating rotor then ‘falls’ into the back-up bearings.

A component of the kinetic energy of the rotor is absorbed by the back-up bearings and the other – higher component – as electrical braking energy when regenerating into the line supply through the stator. In the test set-up this regenerative energy was dissipated in a resistor.

In order to take into account the high operating speed, the worst-case conditions were checked – the failure of the magnetic bearings at maximum speed – on a test motor in a concrete bunker. The concrete bunker is a separate part of the building whose concrete walls are approximately 1m thick. As it was expected the bunker remained undamaged. Further, the standard motor cover was replaced by a wooden cover. This still guaranteed the internal cooling and also made it far simpler to access the back-up bearings.

The factory cannot simulate all of the conditions encountered in the actual plant itself for the electrical/mechanical test set-up. This is the reason that the motor was accelerated up to the appropriate test speed using a drive through a gearbox. However, the gearbox and the drive motor were not powerful enough to generate an adequate braking torque.

The motor was not excited while being accelerated up to the test speed. For the landing test, the bearings were de-activated, and, at the same time, the test object was excited.

The resulting stator current was converted into heat in a water-type resistor therefore generating an appropriate braking torque. This meant that the rotor, depending on the initial speed, came to a standstill between 11 and 19s after the AMB were de-activated.
In the real plant itself, the braking time is assumed to be about 5 to 7s from an initial speed of 6300rpm. In this case, if a fault develops, the energy is regenerated into the line supply via the drive converter. This generates an even higher braking torque than was possible in the trial. Further, a higher flow resistance is manifested at the compressor blades that additionally brake the compressor – motor train.The higher the electrical braking torque, then the lower the mechanical braking torque of the back-up bearings.

This results in a shorter braking time and lower wear of the sleeve bearings.

The test motor was equipped with additional sensors. Acceleration transducers, strain gauges and PT100 resistance thermometers were incorporated in the rotor and the signals transferred using a telemetric system. The fixed parts of the bearings were equipped with sensors to measure the rate of acceleration, temperature, speed and trigger signal. During the landing test, all data was digitised with a sampling rate of 56kHz, recorded and then evaluated.

The basis data –system damping and rotor natural frequency – was first determined at standstill from the levitated condition. In order to determine the accuracy of the simulation program, three pre-trials were carried-out at 1500rpm.

In order to more precisely determine the friction coefficient of the back-up bearings, in an additional trial, at the same speed, the rotor was only electrically braked after 15s.This trial resulted in the most accurate value of the friction coefficient.

The measuring results were aligned with the simulation parameters. Although the pure theoretical data was already quite a good match, several parameters were adapted due to the accuracy required to predict the next trial.

Using the partially new adapted parameters, the landing test was simulated at 4400rpm. The result of the simulation and the comparison with the subsequent landing test resulted in a good match.

It was not possible to create a backward whirl effect – neither theoretically in the simulation nor practically in the landing test. Evtl als Kastentext: “Backward whirl means that if the rotating rotor falls into the back-up bearings, then it would roll with almost no slip in these. The shaft centre point then moves in the opposite direction to the rotor direction of rotation. Theoretically, backward whirl is more likely to take place the higher the coefficient of friction between the rotor and the back-up bearing shells. When backward whirl occurs, this can result in extremely high forces that, depending on the motor type, are extremely difficult to handle.

“Contrary to this, there is the so-called ‘forward whirl’, where the shaft centre point rotates in the direction of rotation of the rotor in the back-up bearings. This situation is non-critical from a mechanical perspective and does not result in any bearing damage.”

The predicted braking time of about 15s was a good match to the data from the result of the trial. The temperature rise of the back-up bearing shells was lower than expected. Although an accurate measurement was not possible, the temperature rise was determined to be about ?T = 10K.

The level of wear of the back-up bearing shells was extremely low so that these could be used for the test at 6300RPM without having to be reconditioned.

After the simulation parameters were again slightly corrected, the test at 6300rpm was predicted to be non-critical. The trial confirmed this. The braking time was about 19s.

From the video recording it was possible to detect the appropriate vapourisation at the back-up bearings. The temperature increase ?T at the back-up bearings was determined to be about 20 to 30 K. The motor was completely functional after the test and no parts of the motor were flung-off so the environment was not in danger. o

Bernd Burmester is with Automation and Drives, Large Drives, Siemens AG, Berlin, Germany. www.siemens.com

"

Recent Issues