In modern manufacturing processes, various types of data—including equipment sensor readings, environmental information, and production metrics—are continuously generated. However, this data is often dispersed across equipment and devices from multiple manufacturers, creating a critical need for a consistent architecture capable of collecting this data in a standardized manner and processing it in real-time.
In this project, we have chosen the Mobius4 platform, an oneM2M-based IoT platform, as the core of our IoT data network to manage this integrated data flow. Mobius4 offers a standardized container structure, allowing data to be handled uniformly regardless of the type or manufacturer of the equipment/sensors. Furthermore, its structure facilitates easy scaling from a single production line to multiple factories and regional sites.
Building upon this foundation, our project aims to establish a Mobius4-based smart factory platform capable of real-time monitoring, anomaly detection, prediction, and comprehensive analysis.
Service OverviewThis project provides the foundation for stable and efficient management of factory operations by collecting and analyzing production, equipment, and environmental data generated in the factory in real-time.
The structure is designed not just for simple data collection, but to understand the entire process as a single, cohesive flow, enabling immediate decision-making and quick response capabilities based on changing situations.
The platform's key features consist of three main components:
- Heterogeneous Equipment Data Collection : Standardizing equipment and devices from different manufacturers via TAS and nCube, allowing for consistent storage in the Mobius4 platform structure.
- Real-time Anomaly Detection and Notification : Detecting anomalies by setting specific policies for each piece of equipment or device, and delivering notifications through the centralized control dashboard.
- Decision Support through Anomaly Analysis : Assisting management decision-making by utilizing diverse data to analyze the scope of impact during anomalous situations and presenting clear response priorities.
The overall architecture is designed to collect and standardize various equipment/sensor data generated from multiple factories, and utilize it for real-time analysis, control, and prediction.
The structure largely consists of the Edge Layer (TAS), nCube Gateway, Mobius4 Platform, Data Processing Server, and Dashboard.
The Mobius platform's container structure is stored hierarchically according to the oneM2M standard.
At the top level resides the AE, which signifies the service domain. Below the AE, Factory containers representing each factory and an Event container managing various events are created. Inside the Factory container, Unit containers representing production lines/units are created, and below each Unit, Machine containers that generate the actual data are sequentially placed.
Within each Machine container, data periodically collected from the equipment is stored chronologically in CIN form. This structure allows data to be clearly categorized by factory-unit-equipment, and facilitates easy expansion in the same hierarchical form when adding new equipment or factories.
Service Description1. Factory Edge Layer (TAS)
This layer is the first to collect raw data generated from equipment and sensors within each factory.
- Collects raw data from equipment or sensors.
- Converts data (based on manufacturer/protocol) into the oneM2M standard format.
- Delivers the normalized data to nCube.
This process enables heterogeneous equipment data to be processed in a uniform manner.
2. nCube Gateway
The nCube, located in each factory, aggregates all TAS data from the factory and transmits it to Mobius.
- Collects and aggregates all TAS data within the factory.
- Arranges data according to the defined container structure.
- Publishes data to Mobius4 based on MQTT topics.
- Maintains a stable, factory-level connection.
Using an MQTT-based structure minimizes network load and allows for easy scalability, even with multiple factories.
3. Mobius4 Platform
Mobius serves as the central hub for all data.
- Stores data in a container structure categorized by factory, unit, and equipment.
- Provides subscription-based notification features.
- Manages integrated data regardless of equipment or sensor type.
- Allows expansion without changing the structure when adding new factories or sensors.
It acts as the main hub connecting the Edge Layer and the Analysis/Control Layers.
4. Data Processing Server
The Data Processing Server receives data from Mobius and performs various tasks.
- Real-time data aggregation and anomaly detection.
- SPC-based process analysis.
- Time-series based anomaly detection and alerting.
- ESG-related prediction and optimization analysis.
Analysis results are generated in this layer, along with event data when necessary.
5. Dashboard
The Dashboard visualizes the following information via APIs from Mobius and the Data Processing Server.
- The status of each machine.
- Event notifications.
- Overall utilization rate.
Managers can instantly check the factory's status and respond using the dashboard.
Detailed ImplementationThe implementation phase focused on building an end-to-end data flow that transfers generated data to Mobius4 and processes/visualizes it in real-time. The overall configuration follows the oneM2M standard structure and is designed to be fully scalable to real-world factory environments.
1. Simulation Environment and Data Source Implementation
Since obtaining real factory equipment data was challenging, we constructed a simulation line based on Factorio to mimic a manufacturing process. This environment generates information—such as equipment status, production volume, and power consumption—similar to data handled in a real process, which was utilized to validate the entire data pipeline and anomaly detection logic.
The simulation environment includes the following components:
- 12 Offshore Pumps
- 12 Boilers
- 12 Steam Engines
- 18 Electric Mining Drills
- 42 Electric Furnaces
- 48 Assemblers
This line has a single process structure consisting of energy production, raw material mining, processing, assembly, and final product production (Automobiles).
The following image shows the configured simulation environment.
The type of data generated by each machine varies depending on the equipment type.
The following is a list of data fields generated by each equipment type.
2. TAS Agent Implementation
In the project, we utilized a Factorio mod to extract equipment status, production volume, and power consumption from all machines within the game.
The data collected by the mod was transferred to the bridge software via RCON, and subsequently converted into a JSON structure usable by the platform.
This implementation ensures that the data collection flow applied in the simulation environment is identical to that of real factory equipment.
3. Data Transmission and Collection Structure
The data converted by the bridge was transmitted to the Mobius platform via MQTT.
The MQTT topic structure was designed to be identical to the container hierarchy within the platform, ensuring that the data from each piece of equipment was naturally mapped to its corresponding location.
4. Mobius Container Deployment
The following hierarchical structure was created in the Mobius4 platform:
AE - Factory - Unit - Machine - CIN
Each message transmitted via MQTT is stored as a new CIN beneath its corresponding Machine container, accumulated chronologically, and utilized for subsequent analysis.
5. Data Processing Pipeline Configuration
The Data Processing Server subscribes to Mobius events and performs the following tasks in real-time.
- Real-time anomaly detection
- SPC-based process analysis
- Time-series based trend analysis
- Event generation and re-transmission to Mobius upon anomaly detection
The results analyzed in this layer are immediately reflected in the Dashboard.
6. Dashboard Implementation
The Dashboard visualizes the following information based on APIs from Mobius and the Data Processing Server.
- The status of each machine.
- Event notifications.
- Overall utilization rate.
This allows the operator to quickly grasp the factory status and take necessary immediate action.
Results and DemoThe implemented system was validated against three scenarios.
- Real-time data collection and visualization
- Anomaly notification (inoperability of certain raw material equipment)
- Overall equipment utilization adjustment notification when certain intermediate material production lines become inoperable
1. Integration and Scalability: Complete Standardization and Integration of Heterogeneous Data
Through oneM2M standard conversion via TAS, all equipment data, regardless of manufacturer or protocol, is collected and managed in a consistent structure. Based on the Mobius4 hierarchical container structure, easy expansion is possible without structural changes when adding new factories, lines, or equipment.
2. Operational Efficiency: Establishing a Proactive Response System Based on Real-time Monitoring
Real-time data collection and anomaly detection ensure process transparency and allow early identification of initial signs of equipment failure or defects. It aids accurate decision-making by analyzing the scope of impact upon anomaly detection, minimizing unplanned downtime.
3. Technical Stability: Securing Stable Data Flow and High Interoperability
The nCube's MQTT-based lightweight transmission reduces network load and maintains stable data flow. Compliance with the oneM2M standard ensures interoperability with various IoT systems, allowing flexible adaptation for future system integration and expansion.
4. Productivity Enhancement: Immediate Recommendation for Adjusting Line Bottlenecks and Equipment Abnormalities
When an abnormal situation occurs, such as a mid-process line stoppage, the system analyzes the entire process data flow in real-time and recommends sections where equipment utilization needs adjustment. This supports management in taking immediate action, reducing production loss and encouraging efficient operation.
5. Prediction-based Operational Advancement: Securing Foundational Data Structure for Future AI/Big Data Analysis
By continuously accumulating standardized time-series data, the necessary foundation is established for applying advanced technologies like predictive maintenance and quality forecasting. It can also be utilized for extended analyses such as ESG and energy optimization.
ConclusionThrough this project, we have established an operational environment that integrates diverse factory data into a single flow, enabling real-time process monitoring and rapid anomaly confirmation. Furthermore, we have laid the groundwork for flexible scalability to accommodate the future addition of factories and equipment. This successfully realized a practical smart factory environment that enhances process stability and operational efficiency.




Comments