Smart Factory Architecture: IoT Sensors, Edge Computing & ERP Integration

A factory floor with 200 machines running 16 hours per day generates approximately 400 GB of raw sensor data daily. Sending all of that data to a cloud server, processing it, and returning instructions to the machines introduces latency measured in seconds. For a safety interlock that needs to respond in milliseconds, seconds might as well be hours.

Smart factory architecture solves this problem by distributing intelligence across three tiers: edge devices that process data at the machine, local servers that aggregate and analyze data within the facility, and cloud platforms that handle cross-site analytics and long-term storage. Getting this architecture right determines whether IoT investments generate real manufacturing value or just generate data nobody uses.

This article is part of our Manufacturing in the AI Era series.

Key Takeaways

• Edge computing processes 80-90% of manufacturing IoT data locally, reducing cloud costs and enabling sub-millisecond response times
• Sensor selection depends on the failure mode you are monitoring, not the machine type, and mismatched sensors are the leading cause of IoT project failure
• MQTT is the standard protocol for manufacturing IoT due to its lightweight footprint and publish-subscribe model
• Odoo ERP integration transforms raw sensor data into business actions through work orders, quality alerts, and inventory adjustments

---

Sensor Types for Manufacturing Environments

Choosing the right sensors is the most consequential decision in a smart factory project. Selecting the wrong sensor type for a given application produces unreliable data that undermines the entire system's credibility.

Comprehensive Sensor Comparison

Matching Sensors to Failure Modes

The common mistake is asking what sensors should go on a specific machine. The correct question is what failure modes need to be detected and what physical parameter changes before that failure occurs.

For example, a bearing failure on a conveyor motor progresses through predictable stages:

1. Early degradation: Ultrasonic acoustic emissions increase (detectable 3-6 months before failure)
2. Developing fault: Vibration signature changes at bearing fault frequencies (1-3 months before failure)
3. Advanced damage: Temperature rises above baseline (weeks before failure)
4. Imminent failure: Current draw increases as friction rises (days before failure)

A comprehensive monitoring strategy deploys sensors that detect the earliest possible stage for the most critical equipment, and later stages for less critical assets, based on the cost of unplanned downtime.

---

Edge Computing Architecture

Three-Tier Processing Model

Tier 1: Machine Edge (Response time: <10ms)

Machine-level edge devices sit directly on or adjacent to individual machines. They handle:

• Safety interlocks (emergency stops, overload protection)
• Real-time process control loops (temperature regulation, speed control)
• Data filtering and compression (sampling reduction from 10kHz to meaningful events)
• Local anomaly detection (threshold violations, sudden changes)

Hardware at this tier typically uses industrial-grade single-board computers or PLCs with embedded Linux. Power consumption is 5-15W, and the devices must withstand vibration, temperature extremes, and electromagnetic interference common in manufacturing environments.

Tier 2: Factory Edge (Response time: <1 second)

Factory-level edge servers aggregate data from multiple machines and perform more complex analytics:

• Cross-machine correlation (detecting when one machine's output quality affects downstream operations)
• Predictive maintenance model inference (running trained ML models against incoming sensor data)
• Quality trend analysis (statistical process control calculations)
• Production tracking and OEE calculation

This tier typically runs on rack-mounted industrial servers with GPU acceleration for machine learning inference. Storage capacity ranges from 1-10 TB for retaining 30-90 days of detailed data.

Tier 3: Cloud/Data Center (Response time: minutes to hours)

Cloud platforms handle workloads that benefit from massive compute and storage:

• Model training and retraining (updating ML models with new data)
• Historical analytics and trend analysis (multi-year data retention)
• Cross-facility benchmarking and best practice identification
• ERP integration and business intelligence

Data Flow Architecture

Sensors (10kHz) → Machine Edge (filter to events) → Factory Edge (analyze/store) → Cloud (train/archive)
                                                   ↕                              ↕
                                              Local Dashboard                  Odoo ERP

The critical design principle is that each tier reduces data volume while increasing data value. Raw vibration data sampled at 10,000 times per second is compressed to frequency-domain features at the machine edge, reducing volume by 95%. The factory edge further summarizes this into health indicators and alerts, reducing volume by another 80%. The cloud receives only business-relevant insights and model training datasets.

---

Communication Protocols

MQTT: The Manufacturing IoT Standard

MQTT (Message Queuing Telemetry Transport) has emerged as the dominant protocol for manufacturing IoT due to several characteristics that align with factory requirements:

• Lightweight: Minimal overhead makes it suitable for constrained devices
• Publish-Subscribe: Decouples data producers from consumers, enabling flexible architectures
• Quality of Service Levels: QoS 0 (fire and forget) for high-frequency monitoring, QoS 1 (at least once) for alerts, QoS 2 (exactly once) for critical commands
• Retained Messages: New subscribers immediately receive the latest state without waiting for the next publish cycle
• Last Will and Testament: Automatic notification when a device disconnects unexpectedly

Topic Hierarchy for Manufacturing

A well-designed MQTT topic hierarchy makes data discoverable and manageable:

factory/{site}/line/{line}/machine/{machine}/sensor/{type}
factory/plant-a/line/assembly-1/machine/cnc-001/sensor/vibration
factory/plant-a/line/assembly-1/machine/cnc-001/sensor/temperature
factory/plant-a/line/assembly-1/machine/cnc-001/status/oee
factory/plant-a/line/assembly-1/machine/cnc-001/alert/maintenance

Protocol Comparison

---

ERP Integration with Odoo

Integration Architecture

The bridge between IoT data and business processes runs through the ERP system. Odoo provides several integration mechanisms for manufacturing IoT:

Odoo IoT Box: Odoo's hardware gateway connects directly to USB and Bluetooth devices. It handles barcode scanners, scales, printers, and simple sensors out of the box. For more complex sensor networks, custom integration through Odoo's REST API is more appropriate.

REST API Integration: Factory edge servers push summarized data to Odoo through its JSON-RPC or REST API:

• Machine status changes create or update work order records
• Quality measurements trigger inspection records with pass/fail determination
• Maintenance alerts generate work orders with predicted failure components
• Energy consumption updates track costs per production order
• Production counts update work order quantities in real time

Webhook-Based Events: For time-sensitive events, Odoo can be configured to receive webhook notifications:

• Equipment alarm triggers immediate maintenance dispatch
• Quality threshold violation stops downstream processing
• Inventory sensor detects low material and triggers reorder

Data Mapping: Sensor to Business Object

---

Network Design and Security

Manufacturing networks require segmentation that balances operational access with cybersecurity. The Purdue Model provides a standard architecture:

Level 0-1 (Process Network): Sensors, actuators, PLCs, machine edge devices. Isolated from business networks. Only communicates upward to Level 2.

Level 2 (Control Network): Factory edge servers, HMI panels, SCADA systems. Communicates with Level 0-1 below and Level 3 above through a DMZ.

Level 3 (Site Network): Manufacturing execution systems, local databases, engineering workstations. Communicates with Level 4 through another DMZ.

Level 4 (Enterprise Network): ERP (Odoo), email, business applications. Standard IT security applies.

Key security measures for manufacturing IoT:

• All MQTT traffic encrypted with TLS 1.3
• Device certificates for mutual authentication
• Network segmentation with industrial firewalls between levels
• Regular firmware updates through managed deployment
• Intrusion detection monitoring for anomalous traffic patterns

---

Implementation: A Practical 90-Day Plan

Days 1-30: Assessment and Design

• Identify 3-5 critical machines for initial deployment
• Document failure modes and select appropriate sensors
• Design network architecture and select edge computing hardware
• Plan Odoo integration points and data mapping

Days 31-60: Deployment and Integration

• Install sensors and edge devices on selected machines
• Configure MQTT broker and topic hierarchy
• Develop Odoo integration connectors (REST API or webhook)
• Build initial monitoring dashboards

Days 61-90: Validation and Optimization

• Validate sensor data accuracy against manual measurements
• Tune alert thresholds to minimize false positives
• Train maintenance and operations teams on new tools
• Document standard operating procedures for IoT-driven workflows

---

Frequently Asked Questions

How many sensors does a typical machine need?

It depends on the machine's criticality and failure modes. A critical CNC machine might have 6-10 sensors (vibration on spindle and axes, temperature on spindle bearing and coolant, current on main motor, acoustic on spindle). A simple conveyor might need only 2-3 (vibration on drive motor, temperature on motor bearing, current draw). Start with the most critical equipment and the failure modes that cause the most downtime.

What is the total cost of an IoT sensor network for a 50-machine factory?

A typical deployment costs $50,000-200,000 including sensors ($15,000-50,000), edge computing hardware ($10,000-40,000), network infrastructure ($10,000-30,000), MQTT broker and software ($5,000-20,000), and integration development ($10,000-60,000). Ongoing costs include cloud services ($500-2,000/month) and sensor replacement (5-10% annually). The investment typically pays for itself within 8-14 months through reduced downtime alone.

Can IoT sensors work with older machines that have no digital interfaces?

Yes. Most industrial IoT sensors are external devices that do not require any digital interface with the machine itself. Vibration sensors attach magnetically or with adhesive. Temperature sensors clamp onto surfaces or pipes. Current sensors clip around power cables. The machine does not need to know the sensors exist. This makes retrofitting older equipment practical and cost-effective.

---

What Is Next

Building a smart factory architecture is the foundation for every advanced manufacturing capability, from predictive maintenance to AI quality inspection to digital twins. The architecture decisions you make now determine what is possible in the future.

ECOSIRE helps manufacturers design and implement IoT-connected factory architectures with Odoo ERP at the center. Our team brings expertise in sensor selection, edge computing design, and ERP integration that turns raw factory data into business value.

Explore our related guides on predictive maintenance and digital twins for manufacturing, or contact us to discuss your smart factory roadmap.

---

Published by ECOSIRE — helping businesses scale with AI-powered solutions across Odoo ERP, Shopify eCommerce, and OpenClaw AI.