An innovative algorithm known as IDA-RCF (Intelligent Detection Algorithm for Rock Core Fissure) has emerged, promising to revolutionize the way rock core integrity is evaluated. By leveraging deep learning techniques, this method offers automated, precise assessments of rock cores, addressing significant inefficiencies rooted in traditional manual processes. With accuracy metrics hovering around 93.8%, the algorithm is positioned to transform safety evaluations in geotechnical projects.
Rock core integrity (RCI) is pivotal for ensuring the stability and safety of underground constructions such as tunnels and mines. Accurate assessment helps engineers identify potentially unstable zones, enabling them to implement preventive measures against geological hazards like landslides and rockbursts. Traditionally, RCI evaluation has returned to manual observation methods, which are time-consuming and can lead to subjective errors due to human oversight. The new method—IDA-RCF—promises to mitigate these issues through reliable automation.
The IDA-RCF algorithm employs groundbreaking technological approaches, employing two distinct branches of feature extraction: one utilizes Deformable Convolution to capture detailed fissure characteristics, and the second relies on EfficientViT, which focuses on extracting global image information through self-attention mechanisms. By merging these features, the algorithm enhances its ability to discern the fissures effectively, even amid complex and varying geological backgrounds.
With evaluation metrics including F1, [email protected], and [email protected]:0.95 reaching 93.09%, 94.44%, and 84.61% respectively, IDA-RCF outperforms many existing models used for the same purpose. This advanced model shows its potential to not only streamline the evaluation process but also deliver accuracies previously thought unattainable. "The prediction accuracy of rock core integrity automatically calculated according to the fissure identification results of IDA-RCF is 93.8%," note the authors of the article.
To validate the method, more than 7000 diverse rock core images were collected for training, testing, and evaluation. The performance results demonstrated IDA-RCF’s superior capability to handle geological variability, from color and texture differences to varying levels of background clutter. This robustness allows for its application across various geological contexts without significant bias.
The broader significance extends to operational efficiency. "The proposed method can accurately evaluate the degree of RCI, which can greatly reduce the manpower and time consumed by manual evaluation means," the authors explain, pointing to substantial potential cost and labor savings for future geotechnical engineering projects.
Although IDA-RCF shows great promise, the authors acknowledge it is just the beginning. Future studies aim to refine this algorithm's capabilities. High-precision semantic segmentation methods are on the horizon, targeting other key evaluation metrics, such as Rock Quality Designation (RQD), to complement fissure rate analyses. This expansion will pave the way for more comprehensive assessments, enhancing overall predictive accuracy.
Through algorithmic advancements like IDA-RCF, the automation and efficiency of evaluating rock core integrity signal significant progress for the field of geotechnical engineering. The integration of deep learning technologies not only streamlines existing practices but also sets new standards toward improving the safety and reliability of underground constructions.