Per-channel dangerous failure rate from mean time to dangerous failure:
λd = 1 / (MTTFd_years × 8760) [failures / hour]
The probability of dangerous failure per hour (PFHd) depends on the architecture category (a simplified, educational model in the spirit of ISO 13849-1 Annex K — not a substitute for a real certification calculation). Verified numerically: PFHd decreases monotonically as category rises B→4, as DC rises, as MTTFd rises, and as the CCF score rises, for every combination sampled — matching the standard's intent that better architectures/diagnostics/independence always certify easier, never the reverse.
Cat B, 1: PFHd = λd (no diagnostics)
Cat 2: PFHd = λd·(1−DC) + 0.01·λd (test channel misses (1−DC))
Cat 3, 4: β = 1 − CCF/100 (common-cause coupling)
PFHd = λd·(1−DC)·β + λd²·8760 (both channels must fail undetected)
Cat 4 additionally scales the residual term ×0.1 (near-complete diagnostics)
PFHd is then banded into a Performance Level a–e (lower PFHd = higher PL, roughly a decade per band from 10⁻⁴ down to 10⁻⁸ /h) — shown here as a log-scale gauge instead of a lookup table, so you can see the achieved value sweep across the band boundaries as sliders move.
The required PLr comes from the classic ISO 13849-1 risk graph: Severity (S1/S2) × Frequency (F1/F2) × avoidance Possibility (P1/P2) walk a branching path to a letter a–e — drawn here as an explicit tree so the whole decision, not just its outcome, is visible. The architecture only counts as "safe" when the achieved PL is at or above the required PLr — the channel-flow diagram's pulses turn green when a simulated fault is caught and red when it would slip through, at a rate that scales with PFHd.
Note: Category 2/3/4 need CCF ≥ 65 to be considered valid architectures in the real standard; below that the diagram flags the design red regardless of PFHd.