MIMO RF Front-End Fault Diagnosis with Complex-Field Attention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional fault diagnosis methods for RF front-end circuits in MIMO systems face challenges in accurately extracting fault features due to high-frequency loss, environmental interference, and complex circuit parameters, leading to decreased accuracy and reliability.
Innovation Solution
A fault diagnosis method using Synchronous Enhancement Extracting Transform (SEET) for pre-processing fault signals, combined with a complex field-based asymmetric convolutional neural network and multi-head attention module, to enhance time-frequency feature extraction and fault identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional time-frequency analysis methods (STFT, WVD, EMD) are used for fault feature extraction, then the analysis can be performed with standard algorithms, but the time resolution and frequency resolution cannot be obtained simultaneously with high accuracy
Solution Approach 1:
The patent applies Synchroextracting Transform (SET) algorithm which changes the analysis parameters by using adaptive window functions and iterative frequency extraction to simultaneously achieve high time resolution and high frequency resolution, overcoming the fundamental limitation of traditional fixed-parameter methods
Solution Approach 2:
The patent introduces a dynamic adaptive window function in the SET algorithm that adjusts its parameters iteratively during the transformation process, allowing the analysis to adapt to the local characteristics of the signal and achieve optimal resolution in both time and frequency domains
2Productivity
If classical SET algorithm with linear phase function and constant amplitude is used, then the transformation can be computed efficiently, but it has very strong limitation for non-stationary signals with strong time-varying characteristics
Solution Approach 1:
The patent transforms the classical static SET algorithm into a dynamic version by introducing time-varying phase functions and amplitude functions that adapt to the instantaneous characteristics of non-stationary signals, enabling effective analysis of signals with strong time-varying properties
Solution Approach 2:
The patent modifies the SET algorithm by changing the parameters from constant values to time-varying functions, allowing the transformation to track and adapt to the changing characteristics of non-stationary signals while maintaining computational efficiency through iterative optimization
3Device complexity
If convolutional neural network with fixed sizes is used for fault feature extraction, then the model structure is simple and training is fast, but the perception range is limited and key fault information cannot be obtained
Solution Approach 1:
The patent extends the feature extraction from traditional fixed-size 2D convolutions to variable-size 3D convolutions that operate across time, frequency, and amplitude dimensions simultaneously, enabling the model to capture complex fault features with broader perception range
Solution Approach 2:
The patent introduces dynamic convolution kernels with variable sizes and receptive fields that adapt during training to capture fault features at multiple scales, overcoming the limitation of fixed-size convolutions while maintaining model efficiency through parameter sharing
4Measurement precision
If convolutional neural network is used for fault identification, then the deep learning architecture can transform fault features to higher-dimension space, but noise may be taken as feature and transmitted to follow-up diagnosis
Solution Approach 1:
The patent applies Synchroextracting Transform as a preliminary signal processing step before feeding data to the neural network, which pre-separates and enhances fault features while suppressing noise, preventing noise from being misidentified as useful features in subsequent diagnosis stages
Solution Approach 2:
The patent incorporates attention mechanisms that provide feedback loops within the neural network, allowing the model to dynamically adjust feature weights and suppress noise components that do not contribute to fault identification, thereby improving reliability
Data Source
AI summary
A fault diagnosis method for the RF front-end circuit of a MIMO system includes using a Synchronous Enhancement Extracting Transform (SEET) to pre-process the acquired fault signal to extract fault feature, creating a fault identification model fused by a complex field based asymmetric convolutional neural network and a complex field based multi-head attention module to assign fault feature weights, extract key feature and identify fault status. The SEET can extract fault feature components, and calculate its real field feature and imaginary field feature to obtain a real field two-dimensional matrix i and an imaginary field two-dimensional matrix q, thus an enhanced time-frequency feature is obtained. The fault identification model is used for a transform from a complex field feature space to a high dimensional space and realizing the assignment of fault feature weights, key features extraction and fault status identification.