FFT Peak Detection Using Spectral Root Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the magnitude and frequency of periodic, sinusoidal signals, such as those used in condition monitoring and power generation, face challenges with accuracy due to high computational loads and spectral leakage, especially when frequencies are not exact integers of the sample rate, leading to errors in fault detection and calibration processes.
Innovation Solution
The implementation of an optimized Fast Fourier Transform (FFT) method using Welch's technique and interpolation, specifically determining an optimal transform root, such as the 4.3093th root, to improve the estimation of magnitude and frequency by minimizing errors through parabolic interpolation and reducing spectral leakage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Fast Fourier Transform is used for spectral analysis, then computational load is reduced and processing speed is improved, but measurement precision and frequency estimation accuracy deteriorate due to spectral leakage and binning errors
Solution Approach 1:
The patent replaces the traditional FFT mechanical computation system with a hybrid approach that substitutes the final peak detection step with a mathematical interpolation model. Instead of relying solely on the discrete binning of FFT, the system uses parabolic, logarithmic, or spectral root interpolation to estimate the true peak location between bins, thereby substituting a more precise mathematical method for the approximate mechanical FFT process at the critical measurement stage.
Solution Approach 2:
The patent creates a composite frequency estimation method by combining FFT with interpolation techniques. The system first obtains the spectral data through FFT, then applies an interpolation algorithm (parabolic, logarithmic, or spectral root) to the peak region. This composite approach leverages the computational efficiency of FFT while adding the precision of interpolation, similar to how composite materials combine different substances to achieve superior properties.
2Device complexity
If traditional FFT methods are used, then computational complexity is low and processing is fast, but magnitude and frequency estimation accuracy deteriorates due to spectral leakage
Solution Approach 1:
The patent segments the spectral analysis process into two distinct stages: (1) FFT computation for obtaining the spectral profile, and (2) interpolation-based peak refinement for accurate frequency and magnitude estimation. By dividing the process, the system maintains the low computational complexity of FFT for the bulk processing while applying the more complex interpolation method only to the peak region, minimizing overall computational burden.
Solution Approach 2:
The patent changes the parameter representation by transforming the discrete FFT bin values into a continuous interpolated function. Instead of working directly with the discrete frequency bins and their magnitudes, the system fits a continuous curve (parabolic, logarithmic, or spectral root) through the peak region points, allowing for sub-bin frequency resolution and more accurate magnitude estimation by evaluating the function at the optimal peak location.
Data Source
AI summary
The magnitude and frequency of a peak in detected vibrational energy emitted from a component that includes a periodic portion is provided to monitor the component for faults. A Fast Fourier Transform of the signal is taken to generate a spectrum. A maximum index in the spectrum is found between a determined frequency bound. An array of parameters is generated and a determined optimized transform root is used on the array of parameters. An interpolated peak location is estimated based on the array and an interpolated frequency is found based on the peak location. An interpolated magnitude is then determined. The use of the optimized transform root in the processing results in significant improvements to the magnitude and frequency estimations, which can improve, for example, detection of defects from vibration spectra of rotating components, estimates of energy radiated from an electronic component, or analysis of spectral content of ionizing radiation.