Identification and Prediction of Casing Collar Signal Based on CNN-LSTM

To address the issue of casing collar localization and ensure the accuracy of well depth measurement, a method combining convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, known as the CNN-LSTM model, has been proposed for the identification and prediction of casing collars. This approach enables intelligent detection of casing collars. Establish simulation models of casing collar locator (CCL) and casing, obtaining CCL signals under various operating conditions. Utilize a sliding window processing method to generate datasets for identification and prediction, and input the processed datasets into a CNN-LSTM deep learning model for training. The evaluation metrics indicate that the CNN-LSTM model outperforms other models in recognizing and predicting casing collars, with an accuracy rate of 99.8127%, precision of 100%, recall of 99.4858%, and an F1 score of 0.9974. The average absolute error, root-mean-square error, and correlation coefficient between the predicted and actual values are 0.8417, 1.5891, and 0.99248, respectively. These results demonstrate the model's high accuracy and versatility, capable of rapidly and accurately identifying and predicting casing collar signals, thereby providing an effective method for well depth measurement.