chronos-mesh

ADR-0004: Multi-Echelon Recursive BOM Explosion & Joint Demand/Lead-Time Safety Stock

Context

In automotive and heavy-machinery manufacturing (e.g., SAP S/4HANA PP/MM multi-plant networks), Tier-1 assembly disruptions rarely originate at Tier-1. Instead, shared Tier-2/Tier-3 upstream components (such as 7nm automotive wafers feeding both powertrain MCUs and hydraulic valve drivers) create diamond-graph Bill-of-Materials (BOM) dependencies. Furthermore, textbook safety-stock formulas that assume constant lead times systematically underestimate buffer requirements during maritime corridor blockades, where transit variance explodes.

Decision

  1. Diamond-Aware Recursive BOM Explosion (SupplyChainGraph.explode_bom):
    • Traverses the multi-tier component DAG (Tier-3 Raw Silicon -> Tier-2 Foundry -> Tier-1 Subassembly -> Final Assembly) and accumulates gross component requirements across shared upstream nodes rather than tree-assuming single paths.
    • For example, 100 units of POWERTRAIN-ECU-X9 require 200 MCU-MODULE-A1 (3x wafer dies each) and 400 VALVE-DRIVER-B2 (1.5x wafer dies each), converging on 1,200 WAFER-DIE-7NM units upstream.
  2. Joint Stochastic Demand & Stochastic Lead-Time Safety Stock (compute_safety_stock):
    • Models independent daily demand variance $\sigma_D^2$ and replenishment lead-time variance $\sigma_L^2$ simultaneously: \(\text{SS}(z_\alpha) = z_\alpha \sqrt{\mu_L \sigma_D^2 + \mu_D^2 \sigma_L^2}\)
    • Captures the quadratic sensitivity to mean daily burn rate $\mu_D^2$ multiplied by transit variance $\sigma_L^2$, feeding directly into node Time-to-Survive (TTS) computation.

Consequences