Vibration Data Spike Filtering for Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for collecting and analyzing vibration data are inefficient in removing noise spikes, leading to false alarms and inadequate preventative maintenance, as they are time-consuming and not well-suited for monitoring multiple machines.
Innovation Solution
An apparatus and method that filters vibration data using a spike filter to identify and eliminate noise spikes by calculating a median value and comparing scalar values to determine actual spikes, producing a noise-free database for accurate fault analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mathematical or statistical analysis is used to identify and eliminate spikes from vibration data, then spike removal capability is improved, but processing time increases and productivity decreases
Solution Approach 1:
The patent transforms the vibration data parameter representation by converting time-domain vibration signals into frequency-domain spectra. This parameter transformation enables more efficient spike identification and removal algorithms to operate on spectral data rather than raw time-series data, thereby improving processing speed while maintaining spike removal effectiveness
Solution Approach 2:
The patent replaces complex mathematical and statistical analysis methods with a simplified spectral analysis approach. By substituting the mechanical computation of statistical outliers with frequency-domain transformation and comparison, the system achieves faster processing while effectively identifying and eliminating spikes from vibration data
2Measurement precision
If complex mathematical analysis is applied to detect anomalies in vibration data, then measurement precision is improved, but device complexity and processing time increase
Solution Approach 1:
The patent replaces complex mathematical anomaly detection algorithms with a simpler spectral comparison method. By transforming vibration data into frequency spectra and comparing spectral characteristics, the system achieves accurate fault detection without requiring complex statistical analysis, thereby reducing device and processing complexity
Solution Approach 2:
The patent transitions from analyzing vibration data in the time domain to analyzing it in the frequency domain. This dimensional transformation provides additional insight into fault characteristics while simplifying the detection algorithm, as spectral peaks and patterns are more easily identifiable and interpretable than time-domain waveforms
3Reliability
If prior art spike filtering methods are used, then noise removal is improved, but the methods are too slow for real-time preventive maintenance monitoring of multiple machines
Solution Approach 1:
The patent changes the data representation parameter from time-domain values to frequency-domain spectra. This parameter change enables more efficient computational algorithms that can process multiple machine vibration datasets simultaneously and in real-time, while maintaining effective noise and spike removal capability through spectral analysis
Data Source
AI summary
A machine monitor includes sensors producing a series of scalar values corresponding to sensed physical parameters. An analyzer produces a first database based on the scalar values and determines a median value of the scalar values for each sensor. It also sets a spike level that is offset from the median value by a predetermined multiple of the median value. A spike filter in the analyzer compares the scalar values to the spike level, and identifies a particular scalar value as a potential spike when the particular scalar value differs from the median value by an amount that is equal to or greater than the spike level. A potential spike is determined to be an actual spike if the first and second side values are within a predetermined range of the median value. A second database is produced with the actual spikes eliminated. Using the second database, corrected faults are identified by finding a data point that exceeds a danger level with a preceding data point exceeding a warning level and two trailing data points being less than an advise level.