Gear Damage Detection via Cepstrum Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting gear damage in interlinking systems are subjective and rely on manual noise assessment, or they use automated spectral analysis methods that struggle with frequency overlap, making it difficult to accurately identify and localize gear damage, especially in multi-stage gears.
Innovation Solution
A procedure that involves recording sensor signals from interlinking systems, transforming them into cepstrum, determining characteristic quadruple values, and comparing scalar functional values to detect and localize gear damage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual noise assessment is used for gear damage detection, then the inspection can be performed with simple equipment, but the detection precision and reliability are low due to subjective operator perception
Solution Approach 1:
The patent replaces manual acoustic assessment with automated spectral analysis using vibration sensors and signal processing systems. The mechanical/subjective assessment process is substituted with electronic sensing and computational analysis, objectively measuring vibration signals to detect gear damage without operator subjectivity.
Solution Approach 2:
The patent introduces vibration sensors and spectral analysis algorithms as intermediaries between the gear system and the inspector. These intermediaries objectively capture and process vibration signals, translating physical gear conditions into measurable data that can be analyzed without direct human sensory involvement.
2Extent of automation
If spectral kurtosis assessments and envelope spectral analyses are used for automated inspection, then the automation extent increases, but the measurement precision deteriorates due to frequency overlap making damage localization impossible
Solution Approach 1:
The patent moves from traditional frequency-domain analysis to time-frequency domain analysis using spectrograms and wavelet transforms. By adding the time dimension to frequency analysis, the method can resolve overlapping frequencies at different time points, enabling both automated detection and precise localization of gear damage.
Solution Approach 2:
The patent segments the vibration signal analysis into different time-frequency components using wavelet transforms and short-time Fourier transforms. This segmentation allows separate identification of different gear mesh frequencies and their corresponding damage locations, resolving the overlap problem that prevents localization in conventional spectral analysis.
3Productivity
If automated spectral analysis methods are used for multi-stage transmissions, then productivity increases through automation, but the device complexity increases requiring extensive training data sets for each damage scenario
Solution Approach 1:
The patent develops a universal spectral analysis framework that can handle multiple gear stages and different damage scenarios using the same core algorithms. The time-frequency analysis methods and pattern recognition techniques are designed to be stage-agnostic, reducing the need for separate training data sets for each specific gear configuration while maintaining high detection accuracy.
Data Source
Figure 1
Figure 2
AI summary
The invention relates to a method for detecting gear damage on at least one gear (1, 2, 3), including the steps of: - generating (S2) a cepstrum of the sensor signal, - determining (S3) characteristic cross-frequency values of the cepstrum, wherein the characteristic cross-frequency values indicate those values of the cepstrum in the cross-frequency range at which the first gear and the second gear mesh at the same location, - defining (S4) a cross-frequency band around each of the characteristic cross-frequency values of the cepstrum, - generating (S5) a scalar function value (ICepstrum) for each of the characteristic cross-frequency values, and - detecting (S6) the gear damage on the at least one gear (1, 2, 3), each by comparing the scalar function value (ICepstrum) with at least one reference value.