Induction Machine Fault Detection via Robust Damped Signal Demixing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection methods for electric machines struggle to distinguish fault signatures from normal operation signals and noise, especially in noisy environments and changing load conditions, due to the variability of fault signatures and interference from transient states.
Innovation Solution
A method that measures stator current signals using a sampling rate at least twice the fundamental frequency, demixes damped signals with non-zero amplitudes and noise using a Hankel matrix, and applies convex or non-convex robust parameter estimation to denoise and extract fault signals, enabling continuous fault detection during operation without machine restarts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classical methods such as Fourier transform and Wavelet transform are used to decompose signals, then they work well for static operations, but they perform poorly in extracting fault signature for electric machines working in transient state due to changing magnitude
Solution Approach 1:
The patent applies Hilbert-Huang transform (HHT) which uses empirical mode decomposition (EMD) to adaptively decompose nonstationary and nonlinear signals into intrinsic mode functions (IMFs). This method dynamically adjusts to changing signal characteristics during transient operations, unlike fixed-parameter Fourier or Wavelet transforms. The HHT method extracts instantaneous frequency and amplitude information that varies with time, enabling accurate fault signature detection during transient states when operating parameters change.
2Adaptability or versatility
If HHT method is used to decompose signal into intrinsic mode function, then it works well for nonstationary and nonlinear data, but it is more like an empirical approach without theoretical guarantee
Solution Approach 1:
The patent incorporates feedback mechanisms through iterative signal decomposition and reconstruction processes. The HHT method iteratively extracts IMFs from the signal, and each IMF is subsequently analyzed using Hilbert spectral analysis. This iterative feedback loop allows the system to progressively refine the decomposition results and improve the reliability of fault signature extraction from nonstationary and nonlinear signals.
3Measurement precision
If compressive sensing technique is used to extract fault signature in very short time, then the electric machine can be assumed to operate at steady condition, but it requires the machine to be at steady condition which limits continuous monitoring capability
Solution Approach 1:
The patent employs dynamic signal analysis methods that can handle both steady-state and transient operations continuously. The Hilbert-Huang transform and time-frequency analysis techniques enable real-time processing of signals with changing characteristics, allowing the system to maintain high measurement precision during transient states without requiring the machine to be at steady condition. This dynamic capability enables continuous monitoring during all operational phases including startup, shutdown, and load changes.
4Measurement precision
If stator current spectral analysis is used to detect fault, then it can identify fault signatures, but the magnitude of fault signatures varies at different loads making it difficult to distinguish from normal operation signal and noise
Solution Approach 1:
The patent applies empirical mode decomposition to segment the complex stator current signal into multiple intrinsic mode functions (IMFs), each representing different frequency components and signal characteristics. By decomposing the signal into separate IMFs, the method isolates fault signatures from normal operation signals and noise into distinct components. This segmentation enables targeted analysis of specific IMFs that contain fault information, improving detectability while reducing interference from other signal components.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A system for detecting faults of an electric machine is provided. The system includes an interface, a memory to store computer-implemented programs including a signal sampling program, a matrix formation program, an optimization formation program, a matrix pencil program, optimization solvers and lookup data including predetermined system parameters related to the faults, and a processor. The processor performs, using the computer- implemented programs, generating a signal matrix based on the acquired signals for the input time domain, forming an optimization problem with a low-rank constraint using the optimization formation program, demixing the signal matrix into a low-rank matrix, a spike interference matrix, and a Gaussian noise matrix by solving the optimization problem using one of the optimization solvers, extracting parameters of damped exponentials from the low-rank matrix using the matrix pencil program, and determining the faults with respect to the induction machine by identifying each of the measured system parameters of the induction machine based on the lookup data.