Research on Defect Detection of Well Control Manifolds Based on Multi-Instrument Data Fusion

This paper proposes a multi-instrument data fusion model based on hybrid article warm optimization and kernel extreme learning machine for the problem of defect detection in offshore drilling well control manifolds. The principles of well control manifold defect detection are studied, and the characteristics of defect detection for individual instruments are analyzed. Ultimately, three types of precision instruments suitable for detecting well control manifold defects are optimized.Combining the output parameters of the instrument, the final determination is made using the coordinate positions x, y of the defect, as well as the defect size dimensions of length, width, and wall thickness as the input feature parameters for the data fusion model, with the model's output being the regression values for each feature parameter.The experimental analysis compared this model with classical extreme learning machines, kernel extreme learning machines, and ordinary particle swarm optimization-based kernel extreme learning machines. The results indicate that the model based on hybrid particle swarm optimization for kernel extreme learning machines has superior accuracy and generalization capabilities compared to the other models.