Three methods are supported:
  1. Alarm during input (equipment alarm, message to operator)
  2. Interlock applied—input allowed only after inspection or rework
  3. Auto-registration for rework, discard, or inspection
Input status is determined by:
  1. Start/end scan (needs 2 scanners)
  2. Time-based: e.g., if input at 13:00 and 3h set, then issue by 16:00 triggers interlock
      (Time can vary by equipment or product model)

UDMTEK addresses potential data overload through multiple strategies:
  A. Selective (Dieted) Data Collection: Out of thousands of parameters (e.g., 1,700), only critical ~100 parameters are transmitted to the database. The rest are stored in files for manual lookup when needed.
  B. Horizontally Scalable Storage: Introduces a horizontally expandable and parallel-processing-based data storage infrastructure to manage large volumes efficiently.
  C. Adaptive Data Collection Cycles: Data acquisition intervals are customized per equipment and aligned with their individual tact times to optimize system performance.

- UDMTEK offers several configurable options for DAS, tailored to reliability and budget requirements. Options C and E include centralized monitoring at no additional cost.
  A. Standard PC: Low reliability, low cost.
  B. Industrial PC: Moderate reliability and cost.
  C. Redundant Industrial PCs: High reliability; supports dual configuration.
  D. Thin-client/server multi-point EDGE: High cost, moderate reliability, easier management.
  E. Dual thin-client/server EDGE: Highest reliability and management convenience.
- Monitoring Functions Included (for C and E):
  · Centralized dashboard for monitoring DAS status.
  · Real-time data loss rate tracking.
  · Remote reset of non-responsive DAS units.
  · Automatic alert messaging to administrators upon abnormal conditions.

Unlike traditional FDC or preventive maintenance systems that rely mainly on sensor data and thresholds, the UDMTEK Platform analyzes both real-time sensor signals and the actual PLC control logic. This allows the platform not only to detect faults but also to identify their root causes at the control level.

By applying AI-based trend and pattern analysis to historical control data, UDMTEK can:
  - Detect abnormal timing or behavior in control sequences,
  - Pinpoint the exact I/O point or logic segment responsible,
  - Trace the issue to a physical component (like a valve or sensor),
  - And guide targeted, preemptive maintenance before failure occurs.

This logic-level insight makes the platform significantly more accurate, robust, and reliable—providing explainable diagnostics by combining sensor signals with deep control characteristic patterns.

AI in the UDMTEK Platform is applied through Trend Analysis and Pattern Analysis.

Trend Analysis uses AI to learn normal timing patterns of control signals and detects subtle shifts before they cause failures. Pattern Analysis builds a Master Pattern from repeated control cycles and identifies deviations in logic flow or sequence.

By combining control logic with digital, analog, and image data, the AI models detect anomalies early and provide explainable, cycle-based insights—enabling predictive maintenance and precise diagnostics.

AI models (e.g., CNN autoencoder) are trained on master pattern cycles to detect abnormal signal flows and statistical sensor anomalies. The integration of digital, analog, and visual data ensures robust anomaly detection and explainable results.

To activate the AI features of the solution, an initial data training period is required. We recommend collecting data for at least 1,000 operation cycles. Assuming each cycle takes approximately 60 seconds, this corresponds to around 16 hours of data collection. However, the required number of operation cycles may vary depending on the application and the specific characteristics of the manufacturing process.

Yes. The UDMTEK Platform integrates seamlessly with existing PLCs — including Mitsubishi, Siemens, Rockwell, and others — via standard industrial protocols such as OPC UA, Modbus, and TCP/IP. There is no need to replace or modify existing hardware, allowing for non-disruptive installation, rapid deployment, and cost-effective integration across your production line.

The data collection speed may vary depending on the network configuration, such as the number of devices connected to the PLC and whether a hub is used. The more devices simultaneously request data or place control workload on the PLC CPU, the more the data collection speed may degrade. The most reliable way to assess actual performance is through validation in the real operating environment.

The UDMTEK Platform enables high-speed data acquisition optimized for smart manufacturing environments. It supports data collection at millisecond-level intervals—as fast as 10–50 ms, depending on equipment specifications and network bandwidth.

In terms of scalability and data capacity:
- Designed to handle tens of thousands of tags per facility, including digital I/O, analog signals, image streams, and log data—with no fundamental limitations on tag count.
  · The system is built on a distributed edge computing and parallel data processing architecture, ensuring high-speed and uninterrupted performance without bottlenecks.
  · Specifically, when using CC-Link IE Field or CC-Link IE Control with a dedicated communication interface card, data acquisition is decoupled from the PLC CPU load. This architecture allows reliable data collection at intervals as fast as 5 ms—even under high CPU utilization or large data volumes.

Data is structured into two categories:
  · Real-time streaming data: Used for monitoring dashboards, FDC, alerts, and rapid operational response.  · Historical batch data: Utilized for trend analysis, traceability, auditing, and long-term process improvement.

The platform also provides flexible data retention and archiving policies, which can be customized based on each project's operational needs and compliance requirements.

The number of I/O points and PLCs that a single Edge PC can handle depends primarily on the I/O load per PLC and the real-time data requirements.

As a general guideline:
  - One Edge PC is recommended to handle up to 10,000 I/O points in total.
  - We typically recommend one PLC per Edge PC to ensure high-speed performance, especially when each PLC controls approximately 3,000 I/O points (bit and word-type tags included).
  - If a PLC has a smaller I/O footprint (e.g., under 1,000 points), a single Edge PC may handle multiple PLCs—typically up to 3–4 PLCs, depending on scan rate, protocol, and network latency.

However, the optimal configuration may vary based on field conditions, such as:
  - PLC brand and communication method (e.g., Ethernet/IP vs. serial)
  - Network bandwidth and stability
  - Required data acquisition interval (e.g., 10ms vs. 100ms)
  - The PLC's control workload and whether it is running other concurrent services

Therefore, we strongly recommend conducting a field assessment to validate data volume, timing, and network characteristics in order to define the most suitable Edge PC deployment strategy.

The UDMTEK Platform primarily uses real-time signal logs (bit-level ON/OFF data) and selected word data collected from the PLC. To enable advanced features such as control logic analysis and pattern recognition, the PLC program must be extracted and uploaded via the system’s configuration interface.

While the extraction method may vary slightly depending on the PLC manufacturer and model, the upload process is straightforward and well-guided.

In addition to PLC data, the platform can also integrate external inputs such as sensor data and NVR-based video streams. These multi-source datasets enhance the system’s ability to build Master Patterns and support deeper diagnostic insights.