Disclaimer
This might be a very broad question. I ask it out of curiosity, because the topic came up in a discussion lately. Hence, I do not have any specific applications or even data that I could show.
Question
In many technical applications, emergency shutdown procedures are essential for safety, e.g. in medical devices. Consider for example a Computed Tomography (CT) scanner, or radiotherapy devices, where the amount of ionizing radiation has to be kept at a save level. Or an magnetic resonance imaging (MRI) machine where the deposed radio-frequency energy and the switching speed of the imaging gradients must be monitored and kept below certain thresholds.
All these quantities can be measured redundantly and independently. In case of a technical failure, the measurement system will detect the fault and initiate an emergency shutdown, BUT at the time the measurement system detects the problem, the patient already was exposed to too high/risky amounts of energy/radiation/etc.
How is this situation handled in those systems? Is the reasoning as simple as
- Our system needs a time $\Delta t$ to detect the faulty state and to switch off
- If $\Delta t$ is short enough, no serious harm is to be expected to happen
or is there a threshold below the real threshold, i.e.
- The emergency shutdown has to be initiated when the intensity (i.e. radiation/energy/magnetic field change per time) $I_\textrm{max}$ is reached
- To fulfill the safety conditions, the device switches off slightly below, e.g. at $0.98 \times I_\textrm{max}$
In the first case, the patient may be shortly in danger, in the second case the system performance may be reduced.
Or does it work in a completely different way?
EDIT:
It is certainly also possible to simulate the system (e.g. applied RF fields, x-ray intensity, ...), measure the current state and predict the upcoming values. If the currently predicted value is in agreement with the actually measured one, it is a reasonable assumption to assume that the predicted value for the next time-step is also correct. Hence you could base the shutdown trigger on the predicted value: When it is too high, shutdown. Also stick to the rule to switch off the system if the currently simulated value and the measured value do not agree.