Off-Normal Event Detection and Response
Off-normal detection watches for the early signs of a failing discharge and switches from performance to recovery or safe termination.
What off-normal means
A discharge is off-normal when it departs from its planned trajectory in a way that threatens performance or the machine. Examples include a growing locked mode, an impurity influx, loss of density control, or an actuator failure. Detecting these early is the difference between a controlled recovery and a disruption.
Detection methods
Detection combines threshold monitors on individual signals with model-based checks that compare the observed state to the expected state. A signal crossing a limit is simple to catch. Subtler faults are caught by watching residuals: when the plasma stops responding to actuators as the model predicts, something is wrong even if no single signal is out of range.
A tiered response
- Continue: minor deviation, feedback corrects it
- Adapt: back off performance, change scenario phase
- Recover: invoke a dedicated controller to stabilize the fault
- Terminate: ramp down safely before the fault escalates
- Mitigate: if termination cannot be safe, minimize the damage
The decision layer
An off-normal supervisor sits above the feedback controllers and selects the response tier. Its logic is often encoded as a state machine or an event matrix that maps detected conditions to actions. This layer is deliberately conservative and analyzable, because it must be trusted to act correctly in rare situations.
In the Kronos program
For the Hyperion breeder, off-normal detection focuses on the failure modes that lead to disruptions, especially locked modes and vertical control loss. The response ladder prefers avoidance and safe termination over mitigation. These detectors and responses are exercised against fault scenarios in the flight simulator ahead of hardware operation.