FMEA Methodology

How Kronos enumerates failure modes, scores them, and turns the ranking into detection, mitigation, and maintenance requirements.

Failure Mode and Effects Analysis

FMEA is the disciplined enumeration of how each component can fail, what effect each failure has, how likely it is, and how detectable it is. The output is a ranked list of risks that drives where sensors, redundancy, and maintenance effort go. Kronos runs FMEA per subsystem and rolls the results up to the plant level for both machines.

Scoring

Each failure mode gets three 1-10 scores: severity S (effect if it occurs), occurrence O (likelihood), and detection D (how hard it is to catch before it matters - higher is worse). The Risk Priority Number is RPN = S x O x D. Modes are addressed in RPN order until residual risk is acceptable.

from dataclasses import dataclass
@dataclass
class Mode:
    name: str; S: int; O: int; D: int
    @property
    def rpn(self): return self.S * self.O * self.D

modes = [
  Mode('magnet quench', 10, 3, 4),
  Mode('diagnostic dropout', 6, 5, 3),
  Mode('plug coil overstress', 10, 8, 5),   # burner, gate-driven O and D
]
for m in sorted(modes, key=lambda x: -x.rpn):
    print(m.name, m.rpn)

From RPN to requirements

• High S: add protective interlock and safe-state path
• High O: add redundancy or reduce duty on the component
• High D: add or fuse diagnostics so the mode is observable early
• High RPN overall: all three, plus a predictive-maintenance model

Detection score D is where the AI stack earns its place: sensor fusion and anomaly detection lower D by making latent faults observable. The subsystem FMEAs - magnets, plug coil, tritium plant, and diagnostics - apply this method with honest scores, including the burner gates.