Scholarlink

<p>As a simple and efficient method based on traditional phase-shift method, double-phase-shift filtering method is capable of dealing with numerous complicated linear scanning signals efficiently combined with a modified U-Net. In the meanwhile, along with the rapid development of exploration techniques in vibroseis, linear scanning signals, whose high-frequency components tend to be absorbed along with the process of propagation, are outshone by nonlinear scanning signals whose scanning frequency is not bound by linear constraints, which results in compensation to the seismic wave absorption. However, double-phase-shift method based on modified U-Net are theoretically unable to deal with a large amount of nonlinear scanning signals because it cannot automatically identify the type of fundamental sweep’s frequency changes with time, resulting in impossibility of constructing corresponding phase-shift curves or starting the filtering process. For improving this situation, a structural regularized support vector machine (SRSVM) structure is added after the process of U-Net to identify the type of fundamental sweep’s frequency, and make it possible of applying double-phase-shift method on nonlinear scanning signals. The inputs of SRSVM structure are signal frequency variations with time of the fundamental waves output by U-Net, and the outputs are types of frequencies of fundamental sweeps change with time, which are required by the process of constructing the phase-shift curves and using double-phase-shift method to obtain the fundamental sweep. Through monitoring the testing process of SRSVM and comparing the results with traditional SVM, it is proven that the effect of SRSVM is eminent and superior to that of traditional SVM, and the final filtering results demonstrate the feasibility of the whole algorithm. Apply this double-phase-shift method based on U-Net and SRSVM to nonlinear scanning signals chosen in actual industrial data, and the satisfactory results show that this method is worthy of promoting.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Harmonic elimination filtering method upon nonlinear scanning signal based on structural regularized support vector machine

  • 期刊论文

  • 发表日期：

    2025年2月22日

查看全文
  • Bolin Li

摘要

As a simple and efficient method based on traditional phase-shift method, double-phase-shift filtering method is capable of dealing with numerous complicated linear scanning signals efficiently combined with a modified U-Net. In the meanwhile, along with the rapid development of exploration techniques in vibroseis, linear scanning signals, whose high-frequency components tend to be absorbed along with the process of propagation, are outshone by nonlinear scanning signals whose scanning frequency is not bound by linear constraints, which results in compensation to the seismic wave absorption. However, double-phase-shift method based on modified U-Net are theoretically unable to deal with a large amount of nonlinear scanning signals because it cannot automatically identify the type of fundamental sweep’s frequency changes with time, resulting in impossibility of constructing corresponding phase-shift curves or starting the filtering process. For improving this situation, a structural regularized support vector machine (SRSVM) structure is added after the process of U-Net to identify the type of fundamental sweep’s frequency, and make it possible of applying double-phase-shift method on nonlinear scanning signals. The inputs of SRSVM structure are signal frequency variations with time of the fundamental waves output by U-Net, and the outputs are types of frequencies of fundamental sweeps change with time, which are required by the process of constructing the phase-shift curves and using double-phase-shift method to obtain the fundamental sweep. Through monitoring the testing process of SRSVM and comparing the results with traditional SVM, it is proven that the effect of SRSVM is eminent and superior to that of traditional SVM, and the final filtering results demonstrate the feasibility of the whole algorithm. Apply this double-phase-shift method based on U-Net and SRSVM to nonlinear scanning signals chosen in actual industrial data, and the satisfactory results show that this method is worthy of promoting.