Transient Electromagnetic Inversion Based on Improved Supervised Descent Method

Supervised descent method (SDM) is a machine learning method mainly used to solve the least squares problem, which is divided into a training phase and a prediction phase. The original SDM transient electromagnetic method (TEM) inversion is difficult to balance the speed and accuracy in the training phase, and deficiencies remain in the convergence stability of the inversion phase. In this article, to address the problems of the original algorithm, a relaxation factor is introduced in the training phase, and the iteration step size is dynamically adjusted to accelerate the training speed and improve the fitting accuracy at the same time. In the prediction stage, adaptive descent, iterative update amount constraints, and inversion termination conditions are used to improve the inversion accuracy and enhance the stability of the algorithm. In this article, we use the forward data of various geoelectric models to carry out inversion trial calculations and verify the inversion effect of the improved SDM algorithm by comparing three aspects of different geological structures, resistance values, and anomalies. The results elucidate that the improved SDM algorithm not only shows better performance in the inversion of stratigraphic models, but also accurately recognizes and distinguishes the buried depth position of the anomalies when dealing with independent anomalies in more complex multifeatured stratigraphy and improves the noise immunity. Finally, the inversion results of the measured data show that the results are consistent with the actual drilling results, which verifies the feasibility and effectiveness of the improved algorithm in practical application.