The ARMxy BL450 Series, an industrial-grade ARM embedded computer is designed for flexible and robust industrial applications. By integrating OpenPLC and OpenCV/YOLO, the BL450 becomes a powerful platform for automated product quality inspection on production lines, detecting defects such as scratches, stains, or dimensional deviations. This article outlines the technical implementation, communication methods, visual system selection, and PLC logic design for such applications, with case studies in electronic component soldering and pharmaceutical packaging inspection.
In modern manufacturing, ensuring product quality is critical. The BL450, combined with OpenPLC and OpenCV/YOLO, enables real-time defect detection and automated sorting on production lines. Typical use cases include:
Electronic Component Soldering Quality Inspection: Detecting incomplete solder joints, excess solder, or misalignments.
Pharmaceutical Packaging Integrity Check: Identifying damaged, misprinted, or incomplete packaging.
The system uses a vision-based approach to detect defects and communicates results to a PLC, which controls mechanical arms to remove defective items or redirect them to a separate line.
The BL450’s flexible I/O configuration and robust hardware make it ideal for industrial environments. Key hardware features include:
Processor: Rockchip RK3588 (based on datasheet), offering high computational power for running OpenCV/YOLO and OpenPLC.
Interfaces:
Ethernet ports (1–3, 10/100/1000M) for industrial communication protocols.
USB ports (2) for connecting cameras.
I/O slots (X and Y boards) for direct I/O signals or RS485 communication.
Dimensions: Compact design (e.g., 42×83×110mm for single Ethernet port models) suitable for production line integration.
Optional Modules: WiFi (BL450W) or 4G (BL450L) for enhanced connectivity.
Example Configuration:
Model: BL450-SOM450-X10 (1 Ethernet port, 32GB eMMC, 4GB LPDDR4X, 2 RS485 ports).
Camera: 2D industrial camera for surface defect detection or 3D camera for depth-based inspection (e.g., bin picking).
The BL450 supports multiple communication methods to integrate the vision system with the PLC and mechanical systems:
Direct I/O Signal:
Used for simple scenarios, such as sending OK/NG (pass/fail) signals.
Configured via the BL450’s 6-pin I/O slot (X board) for direct connection to OpenPLC.
Example: A high/low signal indicates whether a product passes inspection.
Industrial Communication Protocols:
Ethernet/IP, Profinet, Modbus TCP: Suitable for complex data interactions, such as transmitting defect coordinates or confidence scores from YOLO.
The BL450’s Ethernet port supports these protocols, enabling seamless integration with industrial networks.
Example: Modbus TCP transmits defect type and location to OpenPLC for precise mechanical arm control.
The vision system is critical for defect detection. The BL450 supports both traditional and deep learning-based approaches:
Traditional Algorithms (OpenCV):
Techniques: Threshold segmentation, contour detection, edge detection.
Use Case: Suitable for regular objects with predictable defects (e.g., scratches or stains on flat surfaces).
Implementation: OpenCV runs on the BL450’s Debian or Ubuntu OS, processing images from a 2D camera to identify defects based on predefined thresholds.
Advantages: Fast, computationally efficient, and suitable for well-defined inspection tasks.
Deep Learning (YOLO):
Techniques: YOLO (You Only Look Once) for real-time object detection and classification.
Use Case: Complex defect detection (e.g., irregular solder joints or varied packaging defects).
Implementation: YOLO models are deployed on the BL450’s Cortex-A53 cores, leveraging the NPU for accelerated inference. Pre-trained models can be fine-tuned for specific defects.
Advantages: High accuracy for complex patterns, adaptable to diverse defect types.
Camera Selection:
2D Camera: Used for surface-level defect detection (e.g., scratches, stains). Connected via USB or Ethernet.
3D Camera: Used for depth-based inspection (e.g., bin picking or dimensional verification). Provides depth data for complex geometries.
OpenPLC, running on the BL450’s Linux-based OS, handles control logic for mechanical systems based on vision system outputs.
Trigger Mechanism:
The vision system (OpenCV/YOLO) processes images and sends a pulse signal to OpenPLC upon completing defect detection.
Example: A detected defect triggers a high signal on a designated I/O pin, which OpenPLC interprets to actuate a mechanical arm.
Safety Redundancy:
Timeout Detection: If the vision system fails to send a signal within a specified time (e.g., 2 seconds), OpenPLC assumes a failure and halts the production line.
Exception Handling: OpenPLC monitors for vision system errors (e.g., camera disconnection) and executes fallback actions, such as redirecting products to a manual inspection station.
Implementation: OpenPLC ladder logic includes timers and error-checking routines to ensure robust operation.
The BL450’s development environment supports seamless integration of OpenPLC and OpenCV/YOLO:
Operating System: Debian or Ubuntu, with pre-installed libraries for OpenCV and Python for YOLO.
Development Examples (from datasheet):
OpenCV development cases for image processing.
Node-RED for IoT integration and data visualization.
Docker containers for deploying YOLO models.
MQTT for communication between vision and PLC systems.
Customization: The BL450’s flexible SOM and I/O board design allows tailored configurations for specific inspection needs.
Setup: BL450-SOM450-X10 with a 2D camera and OpenCV for contour detection.
Process:
The camera captures images of solder joints.
OpenCV applies edge detection to identify irregularities (e.g., incomplete joints).
Results (OK/NG) are sent via Modbus TCP to OpenPLC.
OpenPLC controls a mechanical arm to remove defective components.
Outcome: Achieves high-speed inspection with minimal latency, suitable for high-throughput lines.
Setup: BL450W-SOM450-X10 with a 3D camera and YOLO for defect classification.
Process:
The 3D camera captures depth and surface data of packaging.
YOLO classifies defects (e.g., dents, misprints) with high accuracy.
Defect coordinates are sent via Ethernet/IP to OpenPLC.
OpenPLC directs defective packages to a reject line.
Outcome: Robust detection of complex defects, with WiFi enabling remote monitoring via BLiotLink.
The ARMxy BL450, combined with OpenPLC and OpenCV/YOLO, offers a versatile and powerful solution for product quality inspection. Its flexible hardware, support for industrial protocols, and rich development ecosystem make it ideal for automating defect detection and sorting in manufacturing. Whether using traditional OpenCV algorithms for simple tasks or YOLO for complex defect classification, the BL450 ensures reliable performance in demanding industrial environments.
