01 — IIoT Foundations
Industrial IoT:
Architecture and Layers
The Industrial Internet of Things (IIoT) in manufacturing and agroindustrial processes is built
on a four-layer architecture: perception (sensors), connectivity (transmission networks),
computing (edge/cloud), and application (analytics and control). Proper implementation of each
layer determines the quality and usefulness of data available for ML models and real-time
decision systems.
04
Application Layer
Operational dashboards · Production ML models · Alerts and control · ERP/MES integration · Data APIs
↑ Analytics & AI ↑
03
Compute Layer (Edge + Cloud)
Local edge computing (latency <10ms) · Cloud processing · Time-series DB (InfluxDB, TimescaleDB) · MQTT broker · Industrial data lakes
↑ Data Processing ↑
02
Connectivity Layer
Industrial WiFi (IEEE 802.11) · LoRaWAN (long range, low power) · Modbus TCP/RTU · OPC-UA · 4G/LTE Cat-M1 for rural areas
↑ Transmission ↑
01
Perception Layer (Sensors)
Temperature · Relative humidity · Pressure · pH · Weight/mass · Vibration · Imaging/vision · Electrochemical sensors · NIR spectroscopy
Al-Fuqaha et al. (2015), in IEEE Communications Surveys & Tutorials, establish that
industrial IIoT reference architectures should prioritize transmission latency
below 100ms for control applications, redundancy in the connectivity layer, and
standardized protocols (OPC-UA as the de facto interoperability protocol in industrial
manufacturing). IEC 62541 regulates OPC-UA implementation in industrial automation environments.
[1] Al-Fuqaha A et al. (2015). Internet of Things: A Survey on Enabling Technologies, Protocols, and Applications. IEEE Commun Surv Tutorials, 17(4), 2347–2376. doi:10.1109/COMST.2015.2444095
[2] IEC 62541:2020. OPC Unified Architecture. Geneva: International Electrotechnical Commission.
[3] Sisinni E et al. (2018). Industrial Internet of Things: Challenges, Opportunities, and Directions. IEEE Trans Ind Inform, 14(11), 4724–4734.
02 — Instrumentation
Sensor Selection and
Validation
Industrial sensor selection must balance five dimensions: metrological accuracy, robustness under
process conditions (temperature, humidity, vibration, chemicals), communication protocol, total
cost of ownership (TCO), and availability of traceable calibration. In food-processing environments,
compliance with sanitary standards is also required (FDA 21 CFR, HACCP, NTC 512-1 for Colombia).
| Variable | Sensor technology | Typical range | Accuracy | Protocol | Approx. cost |
|---|
| Temperature | PT100 / Type K thermocouple | -40 to +500C | ±0.1C | 4-20mA / Modbus | USD 25-150 |
| Relative humidity | Capacitive (SHT40, HIH) | 0-100% RH | ±1.5% RH | I2C / Analog | USD 15-80 |
| Pressure | MEMS piezoresistive | 0-100 bar | ±0.1%FS | 4-20mA / HART | USD 50-300 |
| pH | Combined glass electrode | 0-14 pH | ±0.01 pH | Analog / RS485 | USD 80-400 |
| Weight/flow | Load cell / Coriolis | Variable | ±0.1% | Analog / Modbus | USD 100-2,000 |
| Vibration | MEMS accelerometer (ADXL345) | ±2g to ±16g | ±0.05g | SPI / I2C | USD 5-50 |
| NIR (moisture, Brix) | Inline NIR spectroscopy | 700-2500 nm | ±0.1% moisture | Ethernet / RS232 | USD 5,000-30,000 |
ISO/IEC 17025 defines competence requirements for calibration laboratories and is the mandatory
reference to ensure metrological traceability in industrial IIoT systems. Willmott & Matsuura
(2005) warn that unquantified sensor uncertainty propagates directly into
predictive models, degrading downstream ML reliability by 2-5x when formal measurement
system analysis (MSA) is not applied.
[4] ISO/IEC 17025:2017. General requirements for the competence of testing and calibration laboratories. Geneva: ISO.
[5] Willmott CJ, Matsuura K. (2005). Advantages of the Mean Absolute Error over the Root Mean Square Error. Climate Research, 30(1), 79–82.
[6] AIAG. (2010). Measurement Systems Analysis (MSA) Reference Manual (4th ed.). Automotive Industry Action Group.
[7] INVIMA. (2023). Good Manufacturing Practices Guide - Measurement System Validation. Bogota.
03 — Connectivity and Processing
Edge Computing and Protocols
for IIoT
In Colombian industrial and agroindustrial environments, high-speed internet connectivity is
not always available, especially in rural plants. The Edge Computing paradigm -
processing data locally on devices or plant servers without depending on permanent cloud
connectivity - is essential for operational resilience in LATAM IIoT systems.
Field Protocol
LoRaWAN for Rural Areas
Range of 2-15 km in open environments, low power consumption (5-10 year batteries), ideal for distributed sensors in agricultural fields. 30% CAGR in LATAM (MarketsandMarkets, 2024). Supported by TheThingsNetwork and private networks with Dragino/RAK gateways.
Industrial Protocol
Modbus TCP + OPC-UA
Modbus RTU/TCP is the most widely adopted industrial protocol globally (>95% of PLCs support it). OPC-UA provides semantics and interoperability across heterogeneous systems. Migration from Modbus to OPC-UA is the ISA-95 recommended standard for MES-ERP integration.
IoT Messaging
MQTT for Telemetry
MQTT (Message Queuing Telemetry Transport) is the reference messaging protocol for industrial IoT, optimized for low bandwidth and high reliability. MQTT 5.0 (OASIS, 2019) adds QoS 0/1/2 support, session messaging, and robust authentication.
Edge Platform
Stack OphirIAn
Raspberry Pi 4 / NVIDIA Jetson Nano as low-cost edge gateways. Node-RED for data-flow orchestration. InfluxDB for local time-series storage. Grafana for plant visualization. Mosquitto MQTT broker. Periodic synchronization to cloud (AWS IoT / Azure IoT Hub).
Shi et al. (2016), in IEEE Internet of Things Journal, propose that edge computing
reduces response latency in industrial control systems by 10-100x versus
cloud-only architectures, and reduces transmitted cloud data volume by 90% through local
filtering and aggregation, making IIoT viable even with intermittent 2G/3G connectivity common
in rural Colombian areas.
Without reliable data there is no reliable ML:
IoT instrumentation is the zero layer of all industrial analytics.
[8] Shi W et al. (2016). Edge Computing: Vision and Challenges. IEEE Internet Things J, 3(5), 637–646. doi:10.1109/JIOT.2016.2579198
[9] OASIS Standard. (2019). MQTT Version 5.0. OASIS Open. mqtt.org
[10] MarketsandMarkets. (2024). LoRaWAN Market: Global Forecast to 2029. Chicago: MarketsandMarkets Research.
[11] ISA-95. (2022). Enterprise-Control System Integration: ANSI/ISA-95.00.01-2022. Research Triangle Park: ISA.
[12] Cisco Systems. (2023). IoT at Work: Industrial IoT Technology and Architecture Guide. San Jose: Cisco.
04 — Technical Proposal
OphirIAn Stack
Industrial IoT
OphirIAn has standardized a low-cost, high-reliability IoT technology stack designed specifically
for Colombian industrial and agroindustrial MSMEs, with full metrological traceability, resilient
connectivity, and direct readiness to feed machine-learning models developed in optimization projects.
Sensors
PT100, SHT40, MEMS NIR by variable
Gateway
RPi 4 / Jetson Nano + Node-RED
Storage
InfluxDB 2.0 local + S3 sync
Visual
Operator Grafana dashboards
Deployed system capabilities
· Acquisition: Up to 64 simultaneous sensor channels at 1 Hz - 1 kHz depending on variable.
· Latency: <50ms for critical process alarms (edge processing).
· Storage: Compressed time series (~80% reduction vs CSV) with configurable retention.
· Connectivity: LoRaWAN / WiFi / 4G based on geographic availability.
· ML integration: REST API for direct consumption by Python models (scikit-learn, TensorFlow Lite).
· Calibration: Traceable calibration protocols with automated reminders.
The implementation cost of a basic IIoT node (temperature + humidity + pressure + gateway
+ dashboard) with the OphirIAn stack is between USD 800 and USD 2,500
per monitoring point, representing a 60-75% reduction versus traditional industrial solutions
(Siemens SIMATIC, Rockwell FactoryTalk), while maintaining metrological traceability required
by ISO/IEC 17025.
[13] National Instruments. (2023). Industrial IoT Reference Architecture. Austin: NI Corp.
[14] Amazon Web Services. (2024). AWS IoT Core for Industrial IoT. aws.amazon.com/iot-core
[15] InfluxData. (2024). InfluxDB 3.0 Technical Documentation. influxdata.com
[16] Grafana Labs. (2024). Grafana for Industrial Monitoring. grafana.com/solutions/manufacturing
[17] Raspberry Pi Foundation. (2024). Compute Module 4 Industrial Applications Guide. raspberrypi.com
OphirIAn