Visual Inspection System for Uncoupling Status of Control-Rod Drive Rod in Nuclear Reactor

Zhi Li, Jianqiang Hu, Zhangxiang Zhou, Jin Wang, Xianen Zhou, Qing Zhu, Shaonan Chen, Yaonan Wang2022 · DOI

TL;DR

This paper introduces a novel visual inspection system for nuclear reactors that accurately detects if control-rod drive rods have properly disengaged. By analyzing the elliptical shape of the drive rod head and its height, the system provides a reliable, automated method to ensure reactor safety during critical operations, overcoming the limitations of traditional, less accurate inspection techniques.

What Problem Does This Paper Address?

Nuclear reactors rely on control-rod drive mechanisms (CRDMs) to regulate power and shut down the reactor. A critical safety concern arises when a control-rod drive rod fails to disengage from its assembly during certain operations, such as refueling or in-service inspections. If this uncoupling failure goes undetected, the control rod could be inadvertently lifted out of the reactor core along with other components, posing a significant safety hazard and impacting operational efficiency. Traditional inspection methods for this uncoupling status are often plagued by issues like misjudgment, low accuracy, and high labor costs, necessitating a more robust and automated solution.

Methods

The proposed system employs a two-pronged approach:

  • Ellipse Detection: The system first identifies the drive rod head in the captured image. Due to the smooth, reflective surface of the drive rod, traditional feature matching is unreliable. Instead, the authors utilize template matching to locate the approximate area of the drive rod head. This is followed by an ellipse detection algorithm based on arcs to accurately extract the elliptical parameters of the drive rod's head. This method is designed to overcome noise interference and incomplete ellipse detection.
  • Height Measurement and Modeling: To determine the uncoupling status, the system measures the change in the drive rod's height. Recognizing that the drive rod head can oscillate, leading to measurement inaccuracies, a modeling-based approach is adopted. This involves establishing a polynomial model that relates the ellipse parameters (major and minor axis lengths) of the drive rod head to its vertical distance from the camera. By capturing images before and after the drive rod is expected to rise, the system calculates the initial and final vertical distances and then determines the height difference. This model accounts for the varying heights and oscillations, aiming to improve measurement accuracy.

Datasets

The paper describes experiments conducted using an experimental platform designed to simulate the uncoupling status of the control-rod drive rod. The authors collected image data from this platform. While specific named datasets are not mentioned, the experiments involved testing the system on 150 images to evaluate the ellipse detection and template matching performance, and further experiments with drive rods rising by 30 mm, 35 mm, and 45 mm were used to assess the height measurement accuracy.

Key Insights and Ideas

  • Template Matching as a Precursor: The paper highlights that while feature matching is insufficient for drive rod head detection due to limited features, template matching serves as an effective initial step to narrow down the region of interest.
  • Ellipse Parameters as a Proxy for Height: The core idea is that the elliptical representation of the drive rod head in the image is directly influenced by its distance from the camera. By accurately modeling this relationship, the system can infer height changes.
  • Polynomial Modeling for Robustness: The use of a polynomial model to relate ellipse parameters to distance is crucial for handling the inherent oscillations and variations in the drive rod's position, making the height measurement more reliable than simple geometric calculations.
  • Automated Safety Check: The system offers a pathway to automate a critical safety check in nuclear reactors, reducing human error and improving efficiency during maintenance and operational procedures.

Key Results

  • Ellipse Detection Accuracy: The ellipse detection algorithm, after template matching, achieved accurate detection in 138 out of 150 tested images.
  • Height Measurement Accuracy:
    • A monocular vision scheme (without modeling) resulted in errors generally exceeding 2 mm for height differences of 30-45 mm.
  • The modeling-based approach demonstrated significantly improved accuracy. For actual height differences ranging from 5 mm to 42 mm, the absolute value of the error was consistently below 2 mm in most cases, with some errors as low as 0.04054 mm. This indicates the model's effectiveness in compensating for oscillations.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the drive rod head can oscillate, which can lead to variations in measured height differences even when the actual rise is the same. While the modeling approach mitigates this, it's an inherent challenge.
  • The system relies on a controlled experimental environment. Its performance in the actual harsh underwater environment of a nuclear reactor, with potential for debris, varying lighting, and stronger water currents, would require further validation.
  • Future Work:
    • The authors suggest that the developed system can be applied to other similar scenarios requiring precise visual inspection and measurement in challenging environments.
  • Further validation and testing in real-world nuclear reactor conditions would be a logical next step.

Summary updated

BibTeX
@misc{li2022visual,
  author = {Zhi Li and Jianqiang Hu and Zhangxiang Zhou and Jin Wang and Xianen Zhou and Qing Zhu and Shaonan Chen and Yaonan Wang},
  title = {{Visual Inspection System for Uncoupling Status of Control-Rod Drive Rod in Nuclear Reactor}},
  year = {2022},
  doi = {10.3390/s22228805},
}
Figure