A new study introduces a revolutionary method for the real-time non-destructive testing (NDT) of wire ropes, a critical advancement for safety in ports, mines, elevators, and heavy-lifting operations. The research outlines the development of a system that overcomes the significant limitations of traditional inspection techniques and previous deep learning models by integrating a novel, lightweight AI model with an energy-efficient embedded platform.
While traditional NDT methods, such as optical and electromagnetic testing, often struggle with surface contaminants, instability, or noise, the new approach leverages a deep learning paradigm to achieve superior performance. The system, called Mini-YOLO, is based on the popular YOLOv8 framework and is specifically optimized for deployment on edge computing devices. Unlike earlier models that required extensive processing time and expensive, cloud-based servers, Mini-YOLO is designed for efficiency and speed.
Key to the system's success is its integration with the Rockchip RK3588 embedded platform. This hardware, with its powerful neural processing unit (NPU), allows for rapid, on-site data analysis. The system achieves a remarkable inference time of just 18.5 milliseconds per image, a performance metric that is essential for real-time applications where every second counts. This capability represents a significant improvement over most commercially available electromagnetic detection devices, which can take up to 20-30 minutes for inspection after data collection.
The development team incorporated several technical innovations into Mini-YOLO, including the lightweight MobileNetV3 architecture and a Coordinate Attention (CA) mechanism to enhance feature extraction. These enhancements ensure that the model maintains high accuracy in detecting critical defects like broken wires, wear, and corrosion, all while operating efficiently on the edge. Furthermore, the system is optimized with C++ and a thread pool to fully utilize the NPU’s capabilities, simplifying deployment and data management for industrial use.
This new system offers a cost-effective, real-time, and highly accurate solution for wire rope inspection, directly contributing to enhanced safety and reliability across a range of high-stakes industries.
Reference: https://www.nature.com/articles/s41598-025-16043-z
