Factory Island — Coffee Processing

The Factory Island is the didactic centrepiece of the supply chain: two Dobot robots demonstrate physical processing steps — receiving, sorting, roasting (simulated), quality control, and packaging — while ERPNext MES connects every step to production orders, material tracking, and Hyperledger Fabric.

→ Architecture overview | B2B communication | Decisions log

---

Physical Structure

Aluminium profile rack with three levels. Only the top level carries a stainless-steel perforated plate (Lochblech); the mid and lower levels are solid multiplex shelves.

┌─────────────────────────────────────┐
│  TOP LEVEL — Robot Workspace        │
│  ┌───────────────────────────────┐  │
│  │  Stainless-steel Lochblech    │  │  ← standardised hole grid
│  │                               │  │
│  │  ┌──┐ ┌──┐ ┌──┐ ┌──┐ ┌──┐   │  │  ← 3D-printed containers snap in
│  │  │  │ │  │ │  │ │  │ │  │   │  │     at fixed grid positions
│  │  └──┘ └──┘ └──┘ └──┘ └──┘   │  │
│  │  [IN] [SRT][RST][QC] [OUT]   │  │  ← named station zones
│  │                               │  │
│  │  Dobot #1 (linear axis)  ─── │──│  ← traverses all station zones
│  │  Dobot #2 (Vision Studio)    │  │  ← fixed at QC station
│  └───────────────────────────────┘  │
│                                     │
│  MID LEVEL — Storage (Multiplex)    │
│  ┌───────────────────────────────┐  │
│  │  Spare containers, tools,     │  │
│  │  Seeed Grove sensor modules   │  │
│  └───────────────────────────────┘  │
│                                     │
│  LOWER LEVEL — Compute (Multiplex)  │
│  ┌───────────────────────────────┐  │
│  │  Linux workstation            │  │
│  │  Network switch / router      │  │
│  └───────────────────────────────┘  │
│                                     │
│  FRONT — Touch display (mounted)    │
└─────────────────────────────────────┘

Station Zones on the Lochblech

The Lochblech is divided into named zones. Each zone holds exactly one 6 × 6 cm container at a defined grid coordinate:

Zone	Label	Description
IN	Goods receipt	Raw green beans arriving from Farm
SRT	Sort	Dobot #1 picks and places; defective beans removed
RST	Roast (simulated)	Container rests here during simulated roasting cycle
QC	Quality control	Dobot #2 (vision) inspects roast level and defects
OUT	Dispatch	Packaged, approved product ready for Distributor
REJ	Reject	Quarantine for non-conforming batches

Standardised Containers

All containers share a ~6 × 6 cm base footprint and snap into the Lochblech grid — they cannot slip and always occupy reproducible positions. 3D-printed variants:

Variant	Colour / marking	Content
Input container	green marking	Raw green beans (from Farm)
Sorted container	yellow marking	Pre-sorted beans, ready to roast
Roasted container	brown marking	Simulated roasted beans
Reject container	red marking	Defective / non-conforming beans
Output container	white/neutral	Packaged finished product

---

Dobot Hardware Capabilities

#	Configuration	Capability	Assigned role
Dobot #1	Mounted on linear axis	Traverses the full length of the Lochblech; can pick from any station zone and transport containers along the axis	Material transport: IN → SRT → RST → QC → OUT
Dobot #2	Fixed position + Vision Studio + camera	Identifies objects by colour, shape, and position; no pre-programmed XYZ needed for detected objects	QC station: visual inspection, roast-level classification, defect detection

Assumption: the linear axis spans all station zones so Dobot #1 can reach every container without repositioning. If axis length is insufficient, a manual handoff between zones may be required for some steps — tbc once axis dimensions are confirmed.

---

Didactic Concept: Simulating Roasting and MES

Core Learning Objectives

The Factory Island is designed so students can experience and experiment with the following MES functions in direct connection with physical robot actions:

1. Production order lifecycle — from demand signal to finished goods
2. Bill of materials (BoM) and material requirements — how much raw material is needed?
3. Work order routing — defining the sequence of operations and assigning work centres
4. Real-time shop-floor feedback — robot status and sensor readings visible in MES
5. Quality management — recording non-conformances, quarantining batches
6. Traceability — linking the finished product batch back to the specific Farm harvest
7. Overall Equipment Effectiveness (OEE) — tracking utilisation and downtime per robot

Simulating the Roasting Process

Since actual coffee roasting is not feasible in the lab, roasting is represented through a time-temperature profile simulation in combination with physical container movement:

Student places "green bean" container in IN zone
    ↓
Dobot #1 picks container → moves to SRT zone
    ↓
Dobot #2 (vision) inspects: counts beans, detects foreign objects
    ↓
Dobot #1 moves approved container to RST (Roast) zone
    ↓
MES starts roasting timer (configurable, e.g. 60 s for demo)
Grove temperature sensor at RST zone logs a mock temperature ramp
Touch display shows live temperature curve (Grafana)
    ↓
MES signals roasting complete → Dobot #1 moves container to QC zone
    ↓
Dobot #2 (vision) inspects roast level:
    - "Correctly roasted" token/marking → container moves to OUT
    - "Over-roasted" or defect detected    → container moves to REJ
    ↓
MES books production completion, records quality result
Fabric Peer Node writes batch event to ledger

Physical representation of roast level: The containers or the tokens inside them carry a colour-coded marker (green = raw, brown = roasted, dark red = over-roasted). Dobot #2's Vision Studio detects the colour and classifies the roast level. Students can deliberately place "wrong" tokens to trigger the reject path and observe MES behaviour.

Alternative considered: using actual heat-sensitive material to change colour during the roast phase — rejected as too fragile for repeated classroom use.

Seeed Grove Sensor Integration

Full sensor inventory to be confirmed. The following integration points are planned once the list is available:

Sensor type	Station zone	MES / Grafana use
Temperature	RST (roast zone)	Displays simulated roasting temperature curve
Colour / light	QC zone	Redundant roast-level check (backup to Vision Studio)
Proximity / distance	IN zone	Detects container arrival → triggers goods-receipt dialogue
Weight	OUT zone	Verifies container fill level before dispatch booking
CO₂ (if available)	RST zone	Narrative enrichment: "roasting gas" detection

---

MES Didactic Scenarios

The following scenarios are designed as self-contained exercises. Each scenario starts from a clean MES state and has a defined trigger, expected MES behaviour, and learning outcome.

Scenario A — Standard Production Run

Trigger: Distributor sends a B2B purchase order via REST API.

Flow: ERPNext receives order → checks raw material stock → releases production order → student scans incoming Farm container (QR/NFC) → MES books goods receipt → Dobot workflow executes → MES books production completion → Distributor notified via REST → Fabric records batch.

Learning outcome: Students follow the complete production order lifecycle and see how each physical action (scan, robot movement, sensor reading) generates a corresponding ERP event.

Scenario B — Quality Reject and Non-Conformance

Trigger: Dobot #2 vision detects a defect marker (deliberately placed by instructor).

Flow: Vision Studio classifies bean as "defective" → MES creates non-conformance record → batch moved to REJ zone → supervisor role in ERPNext must approve quarantine or rework decision → Fabric records quality event.

Learning outcome: Quality management workflow; difference between a production booking and a quality booking; what a non-conformance record contains.

Scenario C — Bottleneck and Scheduling

Trigger: Two production orders arrive simultaneously (instructor triggers second order mid-run).

Flow: MES must queue the second order because Dobot #1 (linear axis) is busy → students observe the order queue on the MES dashboard → first order completes → second order released automatically.

Learning outcome: Work-centre capacity, queue management, the concept of a bottleneck resource.

Scenario D — Material Shortage

Trigger: Production order is released but the IN zone container is empty or absent.

Flow: MES material availability check fails → production order placed on hold → MES creates a purchase requisition to Farm Island → students must trigger Farm shipment → goods receipt unblocks the production order.

Learning outcome: MRP logic (material requirements planning); how a missing component propagates through the system.

Scenario E — End-to-End Traceability (Recall Simulation)

Trigger: Instructor announces a quality complaint from the Coffee House for a specific batch.

Flow: Students use ERPNext traceability report → trace batch number back through QC records → back to Farm harvest batch on Fabric ledger → identify which raw material lot was used → simulate recall decision.

Learning outcome: Hyperledger Fabric as the immutable inter-island audit trail; why on-chain batch IDs matter for real-world supply chains.

---

Material Flow: Farm → Factory

Decision: manual transfer with ERP + Fabric tracking (see decisions log).

Farm ERPNext: create outbound shipment → print/attach QR or NFC label
    ↓  (student carries container to Factory Island)
Factory ERPNext: scan QR/NFC on container → goods receipt
    ↓
Fabric Peer Node: records inter-island handover event on ledger
    ↓
ERPNext: checks material availability → releases production order
    ↓
Dobot workflow starts (see roasting simulation above)

Open question: QR code vs. NFC tag on container — not yet decided. Either requires a reader at the Factory workstation.

---

Hardware

Component	Model / Spec	Function
Linux workstation	—	MES, robot control, Node-RED, all Docker services
Touch display	—	MES operator interface (ERPNext) + Grafana production dashboard
Dobot #1	Dobot Magician + linear axis	Material transport along all station zones
Dobot #2	Dobot Magician + Vision Studio + camera	Visual QC: roast-level classification, defect detection
Seeed Grove sensors	tbc (list pending)	Temperature, proximity, colour — integrated via Node-RED
Network switch / router	—	Island LAN, DHCP, routing

---

Services (Docker)

Service	Function
ERPNext Manufacturing (MES)	Production orders, BoM, routing, QC records, traceability
Node-RED	Dobot SDK orchestration, sensor ingestion, MES REST integration
Dobot Python SDK	Direct robot control via USB/TCP (called from Node-RED)
OPC-UA Server	Exposes robot and sensor status to ERPNext
Apache Kafka	Island-internal event processing
Grafana	Production dashboard: roasting curve, OEE, order queue
Fabric Peer Node	Writes processing batch events to the shared ledger
MariaDB	ERPNext database

---

Internal Data Flow

B2B REST API (Distributor order)
    ↓
ERPNext: production order + material check
    ↓
Node-RED: orchestrates Dobot workflow step by step
    ├─→ Dobot #1 (linear axis)  — transport between zones
    ├─→ Dobot #2 (Vision Studio) — QC inspection
    └─→ Grove sensors            — temperature, proximity, colour
         ↓
    OPC-UA → ERPNext: books each operation
         ↓
    ERPNext: production completion + quality record
         └─→ Fabric Peer Node (batch event on ledger)
              └─→ B2B REST API (delivery notification to Distributor)

---

Status

Planned — configuration will be created analogously to the Farm island.

The docker-compose.yml, config files, and scripts for this island are not yet created. Seeed Grove sensor inventory is pending; MES scenario scripts will be developed once hardware is confirmed.