Deep Learning for Precise Perforation Positioning with CCL Signals in Complex Environments

The Casing Collar Locator (CCL) is critical for ensuring precise depth correlation during perforating operations. However, in complex downhole environments, CCL signals are susceptible to distortion from various sources of interference, which can severely degrade their reliability and compromise depth positioning accuracy. To overcome the challenge of robust collar identification under these conditions, this paper introduces BCL-Net, an intelligent model that synergistically combines a ConvNeXt backbone with a Bi-level Routing Attention (BRA) mechanism and a Bidirectional Long Short-Term Memory (BiLSTM) network. This hybrid architecture is designed to effectively capture both spatial features and temporal dependencies within CCL signals, thereby achieving robust identification performance in noisy downhole conditions while maintaining computational efficiency amenable to field deployment. Experimental results demonstrate that at a signal-to-noise ratio (SNR) of −8 dB, the proposed model attains a recognition accuracy of 94.18% and exhibits minimal depth mis-tie across repeated trials. The proposed approach effectively mitigates the limitations of existing methods, offering a highly reliable and interference-resistant solution for depth positioning in automated perforation systems.