Protecting IIoT from EMI-induced phantom data

In industrial Internet of Things (IIoT) systems, data accuracy and reliability are critically important. However, electromagnetic interference (EMI) can generate “phantom” data, leading to automation failures and incorrect decisions. Addressing this challenge requires a multi-layered approach, encompassing physical protection, edge data filtering, and intelligent analytics on the platform.

Identifying and classifying EMI-induced phantom data

Phantom data in IIoT, caused by EMI, results from the unwanted influence of electromagnetic fields on sensors and communication lines. Typical sources of EMI in industrial environments include electric motors, welding equipment, variable frequency drives (VFDs), switching power supplies, digital circuits, clock signals, and wireless transmitters [21, 22, 27, 31, 32]. These sources can induce voltage on wires, leading to distortion of analog and digital signals, or cause radio frequency interference (RFI) affecting wireless communications [1, 27].

The impact of EMI can manifest as: distorted or delayed data, signal loss, reduced performance, false alarms, malfunctions, and complete system failure [8, 14, 21, 28, 30, 31, 32]. For example, a temperature sensor might suddenly show an abnormally high value, or a pressure sensor an unrealistic spike. The distinction of phantom data from other anomalies, such as sensor drift or malicious injection, lies in its unpredictability and direct correlation with the electromagnetic environment. For instance, Hall effect sensors used for current measurement can be sensitive to EMI, whereas resistive current measurement methods show greater resilience [10].

Architectural strategies for physical EMI protection

Physical protection is the first and most crucial line of defense against EMI. It aims to minimize the penetration of interference into IIoT devices and cables. The International Electrotechnical Commission (IEC) has developed the IEC 61000 series of standards, which define electromagnetic compatibility (EMC) requirements for electrical and electronic equipment used in industrial environments [5, 13, 18]. Specifically, IEC 61000-6-2 sets immunity requirements, and IEC 61000-6-4 sets emission requirements for industrial environments [7, 16].

Key architectural approaches include:

• Cable shielding: Using shielded cables, such as shielded twisted pair (STP) or optical fibers, significantly reduces susceptibility to EMI [21]. Optical fibers, for example, are completely immune to electromagnetic fields, making them ideal for highly sensitive applications [30].
• Device and cabinet shielding: Metal enclosures or plastic enclosures with conductive coatings or fillers (e.g., carbon fibers, graphene, or metal particles) create a barrier to electromagnetic waves, absorbing or reflecting them [4, 9, 11, 14]. It is important to ensure precise shaping and integration of shielding elements, as well as proper grounding, to avoid EMI leakage through gaps or loose connections [6, 11, 14].
• Grounding: Proper grounding of all system components provides a path for unwanted currents caused by EMI to dissipate, preventing their impact on sensitive electronics [11, 18].
• Surge protection devices: Installing surge protectors on each Ethernet port connecting field devices to network switches helps absorb transient voltages that can damage equipment or disrupt communication [21].

Data filtering and validation methods at device and gateway levels

Even with effective physical protection, some level of EMI may still reach sensors. Therefore, software filtering and data validation on edge devices or IIoT gateways are critically important. This allows for detecting and discarding EMI-induced anomalies as close to the source as possible, reducing the load on the central platform and ensuring faster response.

Effective methods include:

• Digital filtering: Applying digital filtering algorithms, such as moving average, median filter, or Kalman filter, can smooth noise and extract true signal values [8]. The Kalman filter, for example, is effective for smoothing sensor noise and predicting true values [8].
• Data validation: Range checks, rate-of-change checks, and comparison with threshold values allow for identifying data that falls outside expected parameters [8]. For example, if a temperature sensor suddenly reports a value exceeding the physically possible maximum, it could be a phantom reading.
• Implementation on controllers and gateways: Industrial controllers (PLC) and IIoT gateways often have sufficient computational power to perform these operations. This ensures low latency for critical control systems, as data is processed locally before transmission to the cloud [8, 19].

Anomaly detection and data correlation architecture on the IIoT platform

To detect more complex EMI-induced anomalies that might pass through initial filters, centralized analytics on the IIoT platform are necessary. This layer uses machine learning (ML) and deep learning (DL) to identify patterns that deviate from normal system behavior [23, 25].

Key architectural approaches:

• Data correlation: Analyzing data from multiple sensors allows for confirming anomalies. If one sensor shows an anomaly, but related sensors in the same physical environment do not confirm it, this may indicate phantom data [8].
• Machine learning for anomaly detection: Algorithms such as Isolation Forest or One-Class SVM can be used to detect unexpected sensor behavior without prior labeling of anomalous data [8, 24, 25]. Deep learning also shows significant improvements in anomaly detection, especially when dealing with large volumes of time series data [17, 23].
• Digital twins: Using digital twins allows for validating data by comparing it with a virtual model of the infrastructure, detecting deviations that may be caused by EMI [8].
• Integration with SCADA/MES: Detected anomalies should be integrated with SCADA/MES systems for automatic response, operator alerts, or initiation of corrective actions.

Operational strategies and monitoring EMI protection effectiveness

Implementing architectural solutions without proper operational strategies is insufficient. Continuous monitoring and testing ensure the long-term effectiveness of EMI protection.

• EMC testing: Regular EMC testing for deployed IIoT systems, including interference simulation, allows for verifying equipment immunity to various types of EMI [16, 26].
• Data quality monitoring: Continuous data quality monitoring helps identify new EMI sources or degradation of existing protection mechanisms.
• Response procedures: Developing clear procedures for responding to detected phantom data, including diagnostics, localizing the EMI source, and optimizing protection.
• Cost and trade-offs: Implementing comprehensive EMI protection systems requires significant investment. It is important to assess data criticality and potential failure risks against the cost of different protection levels. For example, using high-precision industrial sensors designed for extreme conditions can reduce drift and increase EMI resilience compared to standard commercial sensors [8].

AZIOT implements architectural solutions to combat EMI-induced phantom data by integrating data from various industrial protocols (MQTT, Modbus, BACnet, SCADA) and sensors. The platform leverages edge computing capabilities for primary data filtering and validation, applying rules and scenarios to detect anomalies on-site. Collected data is aggregated and analyzed on the central AZIOT platform, where dashboards and auditing monitor data quality and detect complex anomalies that may indicate EMI influence. This ensures data reliability and integrity for facility automation and management decision-making.

Matrix for selecting EMI protection strategies in IIoT

Ensuring data integrity in IIoT systems under EMI conditions is a complex but solvable problem. Integrating physical protection, intelligent edge filtering, and platform-level anomaly analytics allows for creating a resilient architecture that minimizes risks associated with phantom data. The right balance between implementation cost and protection level depends on the application's criticality and the potential consequences of failures. Intecracy solutions and inbase.com.ua solutions offer robust platforms for managing complex industrial data environments.