CSAMT Time-Series Denoising for Fast Target Frequency Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CSAMT data denoising methods face high computational costs, long processing times, limited application scope, and poor robustness due to sensitivity to signal shape and noise nature, with existing solutions like cross-correlation algorithms and enhanced receiver capabilities failing to effectively address complex noise interference.
Innovation Solution
A noise suppression method using a denoising network constructed from an improved temporal convolutional network (TCN), bidirectional long short-term memory (BiLSTM) network, and fully connected layer, combined with preprocessing and standardization, to suppress noise in CSAMT data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cross-correlation algorithm is used for denoising CSAMT data, then noise suppression capability is improved, but computational cost increases and processing time extends
Solution Approach 1:
The patent replaces the traditional cross-correlation algorithm (mechanical/mathematical computation system) with a deep neural network system that uses convolutional layers, LSTM units, and attention mechanisms to achieve noise suppression. This substitution reduces computational complexity and processing time while maintaining or improving denoising performance through learned representations from training data.
Solution Approach 2:
The patent changes the fundamental parameters of the denoising approach by transitioning from explicit mathematical operations (cross-correlation) to implicit learned transformations through neural network parameters (weights and biases). This parameter transformation enables the system to capture complex noise patterns and signal characteristics more efficiently, reducing processing time while maintaining suppression capability.
2Reliability
If cross-correlation algorithm is used for denoising, then noise suppression is achieved, but sensitivity to signal shape changes and complex waveforms increases
Solution Approach 1:
The patent introduces dynamic elements through LSTM (Long Short-Term Memory) units and attention mechanisms that can adapt to varying signal shapes and temporal patterns. The LSTM units dynamically adjust their internal states based on input sequences, enabling the model to handle complex waveforms and varying signal characteristics without requiring manual feature engineering or assuming fixed signal shapes.
Solution Approach 2:
The neural network architecture is designed to be universal, handling multiple types of noise patterns and signal shapes through a single unified model. The combination of convolutional layers for spatial features, LSTM units for temporal dependencies, and attention mechanisms for important pattern identification creates a multi-functional system that adapts to various signal characteristics without requiring separate processing paths.
3Reliability
If cross-correlation algorithm is used, then denoising is performed, but assumption about noise nature limits effectiveness when signal and noise differ significantly in frequency domain
Solution Approach 1:
The patent replaces the frequency-domain based cross-correlation approach with a time-domain based neural network approach. The neural network processes signals in the time domain, capturing temporal patterns and dependencies that are not captured by frequency-domain methods. This substitution eliminates the limitation of assuming similar spectral features between signal and noise, as the model learns direct temporal relationships from training data regardless of frequency characteristics.
Solution Approach 2:
The patent transitions from analyzing signals in the frequency domain to analyzing them in the time domain. By using temporal convolutional layers and LSTM units, the model captures temporal dynamics and sequential patterns that are invisible to frequency-domain methods. This dimensional transformation enables effective denoising even when signal and noise have overlapping or significantly different frequency spectra.
4Object-affected harmful factors
If transmitter operating frequency is increased to suppress noise, then noise interference is reduced, but power consumption increases and radio frequency interference issues arise
Solution Approach 1:
The patent replaces the physical approach of increasing transmitter frequency (which consumes more energy and creates RF interference) with a signal processing approach using deep neural networks. The neural network learns to distinguish signal from noise patterns through training, enabling noise suppression without changing the operating frequency. This substitution eliminates the energy consumption and RF interference problems associated with frequency increase while maintaining noise suppression effectiveness.
Data Source
AI summary
A noise suppression method for extracting a target frequency signal from a CSAMT time series, and aims at solving the problems that an existing CSAMT data denoising method is long in denoising time, small in application range and poor in robustness. The method includes the following steps: acquiring CSAMT data to be subjected to noise suppression as input data, the CSAMT data being the CSAMT data with noise; preprocessing the input data to obtain preprocessed data; performing noise suppression on the preprocessed data through a trained denoising network to obtain noise-suppressed CSAMT data; wherein the denoising network is constructed based on an improved temporal convolutional network, a bidirectional long short-term memory network and a fully connected layer which are connected in sequence. According to the method, the denoising time of the CSAMT data is shortened, the application range is expanded, and the robustness is improved.