Manufacturing

MPS (Master Production Schedule): High-level plan of what finished products to produce in what quantity over a planning horizon (typically 3–12 months). Driven by sales forecasts, customer orders, and inventory targets. Focuses on end products only. MRP (Material Requirements Planning): Explodes the MPS through the Bill of Materials (BOM) to calculate what components and raw materials are needed, in what quantity, and by when. Creates planned orders (production orders or purchase requisitions) to satisfy the MPS. Key difference: MPS = 'how many cars do we build this month'; MRP = 'how many engines, tyres, gearboxes do we need, and when must we order them'. Both are part of the broader S&OP (Sales & Operations Planning) process. Modern systems use advanced APS (Advanced Planning & Scheduling) like SAP APO / IBP that incorporate capacity constraints into MRP, unlike traditional unconstrained MRP.

OEE (Overall Equipment Effectiveness) is the gold standard KPI for manufacturing productivity. OEE = Availability × Performance × Quality. Availability = (Planned Production Time - Downtime) / Planned Production Time. Measures unplanned stops, breakdowns, changeovers. Performance = (Ideal Cycle Time × Total Count) / Run Time. Measures speed losses — running slower than rated speed. Quality = Good Count / Total Count. Measures defects and rework. World-class OEE = 85%+ (Availability 90%, Performance 95%, Quality 99.9%). A factory running at 50% OEE has enormous improvement headroom. MES systems calculate OEE in real time from machine data (via SCADA/OPC-UA), production counts, and quality results. Improvement levers: TPM (Total Productive Maintenance) for availability, SMED (Single Minute Exchange of Die) for changeover time, SPC for quality.

A BOM defines the components, sub-assemblies, and raw materials needed to make a product, with quantities and units. Types: 1) Engineering BOM (EBOM): Design-centric, created by R&D — shows all parts as designed. 2) Manufacturing BOM (MBOM): Production-centric — shows how to actually build the product, with manufacturing sub-assemblies (phantom assemblies, by-products). May differ from EBOM due to alternate sourcing, manufacturing variants. 3) Sales BOM: Shows configurable product options (configure-to-order). 4) Multi-level vs Single-level: Single-level shows immediate children; multi-level explodes all levels to raw material. BOM explosion in MRP: System walks down every level of BOM, applying lead times and lot sizes, to calculate net requirements for every component. BOM accuracy is critical — a 1% BOM error can cause ₹Crores in material shortages or excess inventory. Change management (ECO — Engineering Change Order) must be tightly controlled when BOM is updated.

Architecture: 1) Data collection: Each machine's PLC connected to edge gateway via OPC-UA. Edge agent reads machine status (running/stopped), speed, and counter at 1-second intervals. Heartbeat-based downtime detection. 2) Edge processing: Local aggregation to reduce data volume. Downtime events detected at edge and tagged with reason code (by operator via HMI or ANDON). 3) Cloud ingestion: Edge gateway publishes to MQTT broker (AWS IoT Core). Telemetry stored in TimescaleDB (time-series). 4) OEE calculation engine: Streaming processor (Apache Flink) calculates OEE components in real time: Availability from downtime events, Performance from speed ratio, Quality from rejection counts from MES. 5) Dashboard: Grafana/custom React dashboard showing per-machine, per-line, per-shift OEE. Andon board displayed on shop floor TV. 6) Alerting: PagerDuty / SMS alert when OEE drops below threshold. 7) Historical analytics: Data warehouse (BigQuery) for trend analysis, bottleneck identification. Scale: 50 machines × 1 msg/sec = 4.3M messages/day — manageable on mid-tier cloud instance.

Discrete manufacturing: Products are made of distinct, countable components assembled together. Examples: cars, phones, furniture, appliances. Uses BOMs, routing, work orders. Can rework or disassemble. Track by serial number. Process manufacturing: Products result from a formula/recipe using ingredients mixed, transformed, or chemically processed. Examples: pharmaceuticals, chemicals, food, paint, cement. Uses formulas/recipes (not BOM), batch manufacturing, cannot easily disassemble. Track by batch/lot. Key differences in systems: 1) MES: Discrete uses routing-based work order dispatch; process uses recipe-driven batch execution. 2) Traceability: Discrete = serial number genealogy; Process = batch/lot genealogy (critical in pharma for recalls). 3) Quality: Process manufacturing has tight GMP (Good Manufacturing Practice) requirements — electronic batch records (EBR), 21 CFR Part 11 for audit trails. 4) Inventory: Process has yield variations, by-products, co-products — harder to plan. 5) Scheduling: Process has sequence-dependent changeover (e.g., going from white paint to dark paint requires more cleaning than reverse).