KAGRA MIF Meeting 2012/10/23

2012/10/23 13:30 - 14:30 Participants: I don't remember

Fast shutter


Guardian-Watchdog protection for KAGRA’s controls.

Minutes of a discussion with Paolo Ruggi

Following the discussion in the Main Interferometer meeting of Oct 23rd, I interviewed Paolo Ruggi on the guardian protections implemented at Virgo. Virgo has an independent DSP-based watchdog system called guardian. The guardian works as follows, a signal is chosen (or more), and continuously filtered within a chosen bandwidth to generate an r.m.s. signal. If the signal exceeds a set threshold the controls are “opened” and can be re-closed only after manual operator reset. No automatic controls reset is allowed, unless an operator analyzes the event and green-lights a re-start. The thresholds are kept quite wide, so that they are triggered only if the controlled go “wild” and they are never triggered by earthquakes, with the notable exception of the Sendai earthquake. The guardian also triggers when accelerometers saturate at high frequency during maintenance access, and generate false feedback signals. This turned out to be quite useful. The guardian was implemented in ~2003, when a wrong feedback sent a mirror into increasingly large oscillations, until it was shaken off the wires and dropped to the floor. After that the guardian was never changed.

About earthquake resilience strategy, the problem arises only with large and remote earthquakes sending low frequency perturbations. These earthquakes tend to excite the attenuation chain modes until the mirror controls are overwhelmed and lock lost. For a while an earthquake mode was studied to put more viscosity in the system during earthquakes. Basically the strategy was the same used also during storms. During storms at Virgo the high winds shake the building foundations (which is on a small hill). The most damaging effects come from tilts that affect the accelerometers and inject false signals in the feedback. The strategy during wind and other commotion times, is simply to shift the blending point between LVDTs and accelerometers at higher frequency to limit the amount of tilt noise that the accelerometers would otherwise inject in the chain. The shifting blending point strategy was found to be good enough both during wind and earthquake times. Consequently there are no special feedback settings for earthquakes anymore. Of course all details and parameters used in Virgo are available, for suitable adaptation to KAGRA's different situation.

My Comments and Notes to stimulate further discussion: : In KAGRA the situation is different, and I believe better than in Virgo. The chain-damping is made by the Eddy current damper acting on the top filter, no accelerometers are foreseen (at least initially) on the IPs. (note: we have foreseen space for accelerometers on the IPs, which may be implemented if needed, with a suitable budget). Therefore KAGRA is expected to be immune to seismic tilt noise re-injected by the accelerometers (that do not exist) and there will no merging of position and accelerometer signals. Because we do not expect to have wind problems in KAGRA, we may need to do nothing like the shifting merging frequency like in Virgo. I believe that while a separate guardian processor is possible and recommendable, no hardwired analog guardian circuit is easily feasible, because the mechanical resonances of the attenuation chains and suspensions are low frequency, and take time to build up. An analog system capable to distinguish normal from pathological behavior at low frequency may be quite difficult to devise.

The combination of signals from the LVDTs and the shadowmeter at the bottom of the well may be used as tiltmeter against microseismic and seismic excitation. Some of this strategy was attempted by Ruggi at Virgo, but never had time to finalize.

Next Meeting

KAGRA/subgroup/ifo/MIF/Minutes20121023 (last edited 2012-11-02 16:23:16 by YoichiAso)