Advanced control algorithms
The Einstein Telescope (ET) is a very complex instrument. Thousands of control signals will be needed to keep the six interferometers in their optimal operating point. Many of the control signals are nonlinear, are influenced by the control signals of other control systems, or are influenced by noise. Read more about the technology domain Advanced Control Algorithms here.
The control is needed at multiple levels. At each of the corners of the ET, local controls are needed for the different vacuum towers and mirrors. There is also a need for control systems that can guarantee the stability of the mirrors that are located in the vacuum towers at 10 km. This stability can be compromised by, for example, seismic noise or so-called Newtonian noise. The control requires complex control loops at multiple levels.
At low frequencies, current interferometers can also be affected by noise generated by the control systems. New algorithms, partly based on Artificial Intelligence (AI), will be needed to prevent such disturbances in the Einstein Telescope .
It is important that all systems are stable (in lock ) at the same time. Especially with a triangular shape , it is not enough that one of the interferometers works well. The three interferometers of both the high-frequency and the low-frequency part must run stably together. Preferably, all six interferometers are stabilized and operational at the same time. The expectation is that this is only the case 25% of the time, because even a minimal earthquake can cause a mirror to have to be brought back into lock . Research is in full swing into how such a situation can be handled and the mirrors can be stabilized more quickly.
The control precision of the existing, second generation interferometers is not sufficient for use in the much more sensitive Einstein Telescope . For the new, third generation Gravity Wave Detectors (GWD), it is therefore necessary to significantly improve the performance of the control systems.
Advanced control algorithms are used to both monitor and control the six interferometers. The noise caused by the control signals must be reduced at low frequencies. The control circuits must also be able to provide qualitative feedback on the stability of the interferometers during measurements.
The observation is that existing second generation interferometers are disturbed by occasional events, the glitches . Some glitches can be attributed to mini-earthquakes, but there are other glitches whose source is still unknown. As a result, the interferometers lose their operating point. It takes considerable time to stabilize the mirrors and bring the GWD back to normal operation. One of the possibilities is to identify the stable parameters in the Einstein Telescope . These parameters can then be used to bring the ET back to a normal operating range more quickly. As soon as a glitch is detected, a proactive response can be made to the disturbance, and downtime can be minimized.
The research for the Einstein Telescope focuses on the following areas of expertise.
- Improved handling of nonlinear conditions, and better control signals for nonlinear control.
- Better ability to handle exceptional events ( glitches , mini-earthquakes). The use of a second control loop should allow a faster return to normal operating parameters when the operating point is lost.
- Massive real-time processing of sensor data for seismic reconstruction and proactive control. The complexity of the data is expected to be very high, increasing the complexity of the required control.
- To enable fast processing of measurement data in real time, the implementation of AI methods in FPGA (Field Programmable Gate Array) systems is an absolute necessity. The FPGA ICs must be able to control the control signals at an early stage, while ( exascale ) supercomputers further process the measurements and warn other telescopes in time that an event is about to occur.
- AI solutions and adaptive methods are needed due to the large number of signals and control possibilities, but also to identify and resolve correlated control signals.
- Extraction of a quantitative reference from the control signals (with estimation of the performance for the best response to noise). The best control of the system and the performance of the ET must be identifiable from the control signals.
- It is of utmost importance that noise induced by the controllers and sensors is minimized. In the first place, sensors should be developed with a minimum noise level ( noise figure ). Only then is it possible to improve the controllers as well.
The new control algorithms are also suitable for use in other highly complex devices or instruments. The use of a second control level based on safe set points can be transferred to other applications. For example, industrial installations in the event of voltage fluctuations in the electricity grid.
In an analogous way, the principle of the analysis of the control quality can also be used in other industrial systems. When an anomaly is identified, it is then possible to return more quickly to a setting of the control parameters to a situation where the installation did have a nominal control.
By implementing AI hardware via FPGA systems, the systems can also be used for other control systems. The concepts of wave field reconstruction are in principle transferable to other areas that deal with waves of different types, even sound. Think for example of the control of acoustic systems (event acoustics, concerts/music events).