# IoT Shop Floor Integration: Connecting Machines to Your Manufacturing ERP in Real Time
The gap between what happens on the shop floor and what your ERP shows is the root cause of most manufacturing inefficiencies. McKinsey's research on manufacturing operations estimates that real-time data integration can improve OEE by 10-20%. Operators know a machine has been down for an hour, but the ERP still shows the production order as "on schedule." Raw material was consumed, but inventory updates lag by a shift. IoT shop floor integration closes this gap — permanently.
Why Real-Time Connectivity Matters
Traditional ERP data entry is manual, delayed, and error-prone:
- Shift reports entered at end of shift — 8 hours of latency
- Production counts rounded or estimated — accuracy ±5–15%
- Downtime reasons recorded from memory — root cause lost
- Quality data transcribed from paper — transcription errors common
With IoT integration, data flows automatically from machine to ERP in seconds, not hours.
The Business Impact
| Metric | Manual Reporting | IoT-Integrated ERP | Improvement |
|---|---|---|---|
| Data latency | 4–8 hours | < 30 seconds | 99.9% |
| Production count accuracy | 85–95% | 99.5%+ | Significant |
| Downtime capture rate | 40–60% of events | 98%+ of events | Major |
| Inventory accuracy | 85–92% | 97–99% | Substantial |
| Order status visibility | Shift-level | Real-time | Transformative |
Understanding Shop Floor Communication Protocols
Before connecting machines, you need to speak their language.
OPC-UA (Open Platform Communications Unified Architecture)
The gold standard for industrial interoperability:
- Pros: Platform-independent, secure (built-in encryption and authentication), rich data modeling, widely supported by modern PLCs
- Cons: Higher complexity, requires OPC-UA server on each machine or gateway
- Best for: Direct PLC communication, new equipment, multi-vendor environments
- Vendors supporting natively: Siemens, Beckhoff, B&R, Rockwell (recent), Mitsubishi (recent)
MQTT (Message Queuing Telemetry Transport)
Lightweight publish-subscribe messaging:
- Pros: Extremely lightweight, low bandwidth, supports unreliable networks, huge ecosystem
- Cons: No built-in data model (you define your own), requires broker infrastructure
- Best for: Sensor data, edge-to-cloud communication, high-volume low-latency streams
- Popular brokers: Mosquitto, HiveMQ, EMQX, AWS IoT Core
Modbus (TCP/RTU)
The legacy workhorse:
- Pros: Simple, universally supported by older equipment, easy to implement
- Cons: No security, limited data types, polling-based (not event-driven), register-based addressing
- Best for: Connecting older PLCs and equipment that only support Modbus
MTConnect
Manufacturing-specific open standard:
- Pros: Purpose-built for CNC machines, standardized data dictionary, RESTful HTTP interface
- Cons: Limited to CNC/machine tool domain, read-only by design
- Best for: CNC machine monitoring, especially in job shops
Architecture Patterns
Pattern 1: Edge Gateway Aggregation (Recommended)
``` Machines (Modbus/OPC-UA) → Edge Gateway → MQTT → Cloud/On-Prem Broker → ERP API ```
When to use: Most scenarios. The edge gateway normalizes protocols, buffers data, and handles connectivity issues.
Edge gateway options: - Industrial: Advantech, Moxa, Siemens IOT2050, Hilscher - Open-source: Apache PLC4X on Raspberry Pi / industrial PC - Cloud-managed: AWS Greengrass, Azure IoT Edge, Google Cloud IoT
Pattern 2: Direct Machine-to-Cloud
``` Modern Machine (MQTT native) → Cloud Broker → ERP API ```
When to use: New machines with built-in MQTT/HTTP clients. Simpler but less resilient.
Pattern 3: MES Middleware
``` Machines → MES → ERP ```
When to use: When a full MES is already in place. The MES handles shop-floor orchestration and sends summarized data to ERP.
FlowSense Manufacturing ERP includes a built-in lightweight MES module that can serve as both the aggregation layer and the ERP, eliminating the need for separate middleware. Explore FlowSense MES.
Data Model Design
Machine State Tracking
Define a universal state model for all machines:
- RUNNING — producing parts within specification
- IDLE — powered on, not producing (waiting for material, operator, program)
- DOWN_UNPLANNED — unexpected stoppage (breakdown, fault, crash)
- DOWN_PLANNED — scheduled maintenance, changeover, cleaning
- SETUP — changeover between jobs
- OFFLINE — powered off or not communicating
Map each machine's native status codes to this universal model at the edge gateway level.
Production Counting
Capture counts at the source:
- Total count — all parts produced (good + bad)
- Good count — parts passing quality criteria
- Reject count — parts failing, subcategorized by defect type
- Cycle time — time per part or per batch operation
Critical design rule: Use event-based counting (increment on each part-complete signal) rather than periodic polling. Polling misses parts and creates inaccurate counts.
Downtime Event Logging
Every state change should generate an event:
```json { "machineId": "CNC-042", "timestamp": "2025-07-15T14:23:17.450Z", "previousState": "RUNNING", "newState": "DOWN_UNPLANNED", "faultCode": "E-4012", "faultDescription": "Spindle overload detected", "operatorId": null, "productionOrderId": "PO-2025-4478" } ```
The ERP receives this event and: 1. Updates the production order status 2. Logs the downtime event with duration tracking 3. Recalculates the estimated completion time 4. Alerts the supervisor if downtime exceeds threshold
ERP Integration Points
Real-Time Production Tracking
IoT data feeds directly into the ERP's production module:
- Work order progress updated automatically as parts are counted
- OEE calculated live — availability, performance, and quality components
- Schedule adherence visible in real time — early warning of delays
Automated Inventory Transactions
When a production order completes a step:
- Raw material consumption posted based on BOM and actual quantity produced
- WIP transfers between work centers recorded automatically
- Finished goods receipt triggered when last operation completes
- Scrap transactions posted with reason codes from quality sensors
Quality Data Capture
Inspection data flows from IoT devices to ERP quality module:
- SPC (Statistical Process Control) charts generated from sensor data
- Control limits monitored automatically — out-of-spec triggers alerts
- Inspection records created without manual entry
- Non-conformance reports auto-generated when rejection thresholds hit
Implementation Roadmap
Week 1–2: Audit and Plan - Inventory all machines, PLCs, and existing connectivity - Map data points available from each machine - Prioritize machines by business value and connectivity readiness - Select edge gateway hardware and MQTT broker
Week 3–4: Pilot Line Setup - Install edge gateways on 3–5 pilot machines - Configure protocol adapters (OPC-UA, Modbus) - Set up MQTT broker and verify data flow - Define data model and topic hierarchy
Week 5–8: ERP Integration - Build REST API endpoints in ERP for receiving IoT data - Implement real-time production tracking updates - Configure automated inventory transactions - Set up downtime event processing and alerting
Week 9–12: Validate and Expand - Compare IoT data against manual records for accuracy validation - Train operators on new real-time dashboards - Resolve edge cases and data quality issues - Begin rollout to remaining production lines
Security Considerations
Shop floor IoT introduces new attack surfaces:
- Network segmentation — Keep OT network separate from IT with a DMZ
- Encrypt in transit — TLS for MQTT, HTTPS for API calls
- Authenticate devices — X.509 certificates for edge gateways
- Firmware updates — Establish a patch management process for IoT devices
- Monitor for anomalies — Unexpected traffic patterns may indicate compromise
Total Cost of Ownership
For a 50-machine facility:
| Component | One-Time Cost | Annual Cost |
|---|---|---|
| Edge gateways (10 units) | $15,000–$30,000 | $2,000 (maintenance) |
| MQTT broker (on-prem) | $5,000–$10,000 | $1,500 (support) |
| Network infrastructure | $10,000–$20,000 | $3,000 (management) |
| ERP integration development | $20,000–$40,000 | $5,000 (updates) |
| Sensor add-ons | $10,000–$25,000 | $2,000 (replacement) |
| **Total** | **$60,000–$125,000** | **$13,500** |
Payback typically occurs within 8–14 months through improved OEE and reduced manual effort.
Next Steps
Real-time shop floor visibility is the foundation for everything else in smart manufacturing — predictive maintenance, AI-driven scheduling, and autonomous quality control all depend on connected machines.
Schedule a FlowSense IoT integration workshop and get a customized connectivity plan for your facility.
