Abstract: In order to solve the problem of cementing quality evaluation, we proposed a cementing quality evaluation method based on CNN-SVM and integrated learning. Firstly, the method adopted improvement measures such as reducing the number of network layers, adding multi-scale convolutional layers, and embedding convolutional attention modules for the DenseNet model to improve the training speed and evaluation accuracy of the model. Secondly, the InceptionV1 module and dilated convolution were used to construct an Inception-DCNN model with relatively small model complexity and relatively high evaluation accuracy. Thirdly，three classic convolutional neural network models (ResNet50, MobileNetV3-Small and GhostNet) were selected. By utilizing the powerful feature extraction capabilities of convolutional neural networks and the structural risk minimization capabilities of support vector machines, the above  models were combined with a support vector machine to synthesize a new CNN-SVM model to improve the generalization ability of the model. Finally, the Bagging method was used to integrate the five new CNN-SVM models into a strong learner, thereby improving the accuracy of the evaluation results and enhancing the anti-interference ability of the model. The experimental results show that the accuracy of  the method for 3 types of evaluation samples in the test set is 97.69%, which is 1—9 percentage points higher than that of  a single model and other methods, thus verifying  the feasibility of using  methods based on CNN-SVM and ensemble learning for cementing  quality evaluation.

Key words: cementing quality evaluation, sector cement cement logging, integrated learning, convolutional neural network, support vector machine