Vibration Waveform DC Bias Removal via Moving Average
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vibration waveform analysis is hindered by asymptotically decaying DC bias, causing spurious low-frequency components in frequency spectra that can be misinterpreted as equipment issues, especially when the DC component changes erratically due to sensor placement or equipment jolts, and is difficult to distinguish from AC vibration changes.
Innovation Solution
A method involving computing a running average of the vibration waveform to identify and subtract the DC component, using techniques such as moving averages, linear least square fits, or polynomial/exponential equations to remove the asymptotically decaying DC bias, followed by a Fast Fourier Transform to produce a cleaner vibration spectrum.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Fast Fourier Transform is performed on the disturbed waveform, then frequency spectrum analysis is obtained, but spurious low frequency components are introduced due to DC disturbance
Solution Approach 1:
The patent extracts and removes the DC component from the vibration waveform before performing Fast Fourier Transform. By separating the DC component (which causes spurious low frequency components) from the AC vibration signal, the transformation can be performed without introducing harmful artifacts into the frequency spectrum.
Solution Approach 2:
The patent applies preliminary DC removal processing to the vibration waveform before conducting the Fast Fourier Transform. This preliminary action eliminates the source of spurious low frequency components, ensuring that the subsequent frequency spectrum analysis is not contaminated by DC-related artifacts.
2Reliability
If DC component changes are monitored, then bias shifts are detected, but it is difficult to distinguish amplifier bias changes from AC vibration component changes
Solution Approach 1:
The patent segments the vibration waveform into distinct DC and AC components. By calculating the DC component as the average value and separating it from the AC vibration signal, the system can independently monitor and analyze changes in each component, making it possible to distinguish amplifier bias changes from actual vibration changes.
Solution Approach 2:
The patent introduces an intermediary DC removal process that acts as a mediator between the raw waveform and the analysis stage. This intermediary step isolates the DC component effects, allowing technicians to determine whether DC changes are due to amplifier bias or other factors without confusion from AC vibration variations.
3Loss of information
If asymptotically decaying DC bias is present, then circuitry settling is reflected in the waveform, but low frequency components appear in the spectrum that can be misinterpreted as equipment problems
Solution Approach 1:
The patent converts the harmful effect of asymptotically decaying DC bias into a beneficial diagnostic tool. By removing the DC component and analyzing its decay characteristics separately, the system can identify circuitry settling events without allowing them to create misleading low frequency components in the vibration spectrum, thus transforming a potential error source into useful information.
Data Source
AI summary
A computer implemented method processes time waveform machine vibration data that are indicative of operational characteristics of a machine. The data, which were measured on the machine over a period of time having a begin time and an end time, are accessed from a memory or storage device. An integer number M of waveform samples are determined from the data to be averaged, and an asymptotically decaying DC bias component in the data is derived using a moving average of the M number of waveform samples. The DC bias component is extrapolated from the begin time of the waveform back to an earlier time and from the end time of the waveform forward to a later time. The DC bias component is then subtracted from the time waveform data, and a Fast Fourier Transform is performed on the data to generate a spectrum.