SSMT – Disaster-Prevention Playbook (9.5–9.8)

Risk:
Thermal damage in stored or transported SKUs (partial thaw / unintended re-freeze)

Inputs:
Per-SKU temperature probe.
Declare T_m_tag for each SKU (for example: "frozen dairy", "fresh produce", etc.).

Guard (symbol space):
a_phase_sku := tanh(
    c_m * ( (T_item_K - T_m_sku) / DeltaT_m_sku )
)

Action (enter):
|a_phase_sku| > Phi_sku
for >= T_sku_min minutes
→ flag SKU at risk

Clear (exit):
|a_phase_sku| <= Phi_clear
for >= T_clear minutes
→ return SKU to normal flow

Notes:
• The same numeric band works for different SKUs because `a_phase_sku` is bounded (-1,+1).
• Use different `Phi_sku` per SKU class if needed, but keep the math identical.
• Separate warm-risk SKUs (watch `a_phase_sku` → +1) from re-freeze-risk SKUs (watch `a_phase_sku` → -1).

Risk:
Thermal runaway in racks / enclosures

Inputs:
Rack inlet temperature(s) on the hottest intake path.
Short-window rolling volatility V_T.

Guards (symbol space):
e_T_inlet
V_T := sqrt( Var_{window}( e_T ) )

Action (enter):
e_T_inlet >= +E_hot
OR
V_T >= V_star
for >= T_hot_min minutes
→ initiate protective action (shed load / escalate cooling tier)

Clear (exit):
e_T_inlet <= (+E_hot - E_hyst)
for >= T_clear minutes
→ stand down

Notes:
• `E_hot`, `E_hyst`, `V_star`, `T_hot_min`, and `T_clear` MUST be declared in policy.
• Monitor the worst inlet, not the average.
• Volatility (`V_T`) catches unstable thermal oscillation even if absolute `e_T_inlet` is not yet extreme.

Risk:
Unsafe cryogenic state in propellants or lines

Inputs:
Line/tank temperature.
Pivot tag T_m_tag = propellant.
Choose lens appropriate to regime (beta for deep cold, qlog near 0 K).

Guards (symbol space):
a_phase_prop := tanh(
    c_m * ( (T_prop_K - T_m_prop) / DeltaT_m_prop )
)

e_T_beta := ( T_ref / T_prop_K ) - 1
# cold-emphasis contrast

e_T_qlog := ln(
    ( T_prop_K / T_ref + alpha ) / ( 1 + alpha )
)
# near-zero-stable contrast; publish alpha > 0

Action (enter):
a_phase_prop <= -Phi_freeze
OR
e_T_beta >= E_cold*
for >= T_min minutes
→ inhibit fill / pause operation

Clear (exit):
a_phase_prop >= +Phi_clear
for >= T_clear minutes
→ resume

Notes:
• You MUST publish which lens is used (`beta`, `qlog`, etc.) and its parameters (`alpha`, T_ref).
• Sensors should reflect the coldest actual fluid segment, not just ambient.
• Symbol logic isolates safety from any local unit scale or sensor vendor.

Risk:
Fuel gelling and flow blockage

Inputs:
Fuel line temperature near intake.
Pivot tag T_m_tag = cloud_point for that fuel blend.

Guard (symbol space):
a_phase_fuel := tanh(
    c_m * ( (T_fuel_K - T_m_cloud) / DeltaT_m_cloud )
)

Action (enter):
a_phase_fuel <= -0.10
for >= 10 minutes
→ pre-heat / switch blend / protective action

Clear (exit):
a_phase_fuel >= +0.10
for >= 10 minutes
→ return to normal

Notes:
• `T_m_cloud` MUST reflect the actual phase-change behavior of the current fuel formulation.
• Because `a_phase_fuel` is bounded (-1,+1), the same trigger and clear values can be applied anywhere once `T_m_cloud` is declared in the manifest.
• This prevents guesswork like “is the line at −15?” and replaces it with “is the symbolic dial in the danger band long enough to act?”