Definition
CFC measures how often propagating faults are stopped before they cross a predefined system boundary, weighted by event severity.
The Measurement Gap
Every reliability metric in common use (MTBF, MTTR, OEE, PM Compliance) is node-local. None can express that a pump failure propagates through a heat exchanger into a reactor into a column. FMEA was supposed to capture cross-boundary effects but in practice stops at the equipment boundary. Modern plants are deeply interconnected scale-free networks, so the most expensive failures are not those with the worst MTBF but those whose blast radius exceeds the containment. Failures propagate through thermal, hydraulic, electrical, and control dependencies that span system boundaries, and FMEA (the tool designed to catch these interactions) in practice models effects within the equipment boundary and misses cross-system cascades. Existing dashboards measure assets as independent nodes while the plant fails as a network.
Decision Supported:
Where to invest in redundancy, isolation, protection systems, or operator procedures to harden cross-system boundaries. Prioritized by the severity of failures that have actually breached them.
Formula
Plain language: For each fault with propagation potential, did it get stopped before crossing the next system boundary? Weight by severity, then take the weighted hit rate.
Notation:
CFC = Σ (s_e × I(contained_before_boundary_e)) / Σ s_e
Variables:
| Variable | Definition | Unit |
|---|---|---|
| s_e | Severity weight for event e (economic loss, safety class, or criticality from the asset registry) | Variable |
| I(·) | Indicator: 1 if fault contained before crossing next system boundary, 0 if it propagated across | Binary |
| System boundary | Defined by functional location topology (e.g., utilities ↔ reactor train) | Categorical |
| Denominator | Sum of severity weights for all propagation-capable events in the window | Variable |
Only events with a viable propagation path are counted. A more advanced version uses physics-informed graph neural networks to predict propagation probability prospectively.
Proposed formula, pending validation pilots.