Knock Sensor Noise Fingerprinting via ADSR Envelopes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Knock sensors in combustion engines often detect unidentifiable noises that are not characterized, leading to a log of unsorted data, which reduces the utility of the noise recordings for analysis.
Innovation Solution
A method involving the use of Attack-Decay-Sustain-Release (ADSR) envelope and joint time-frequency techniques, such as cepstrum, quefrency, chirplet, and wavelet techniques, to create a sound fingerprint of engine noises, allowing for characterization and classification of unidentifiable noises through networked engine control units and external systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If knock sensors detect and record all engine noises, then the quantity of noise data increases, but the utility of the data for analysis decreases due to lack of characterization and sorting
Solution Approach 1:
The patent applies spectral analysis to transform noise signals into frequency domain representations, effectively 'coloring' or categorizing noises by their frequency characteristics. This allows systematic identification and characterization of different noise types, converting unsorted data into structured, analyzable information with distinct frequency signatures
Solution Approach 2:
The patent changes the parameter representation of noise data from raw time-domain signals to characterized frequency-domain parameters through spectral analysis. By transforming the data representation and extracting key frequency parameters, the system maintains large quantities of noise data while significantly improving their utility through systematic characterization
2Ease of manufacture
If unidentifiable noises are logged without characterization, then data collection is simple, but analysis efficiency decreases due to unsorted data requiring manual review
Solution Approach 1:
The patent performs preliminary characterization of noise data at the time of collection by applying spectral analysis and creating fingerprints. This preliminary processing automatically categorizes noises during data acquisition, eliminating the need for manual review later and significantly improving analysis efficiency while maintaining ease of data collection
Solution Approach 2:
The system performs self-characterization of noise data through automated spectral analysis and fingerprinting algorithms. The noise detection system serves itself by automatically categorizing and tagging recorded noises with their frequency characteristics, removing the burden of manual analysis and improving overall productivity
3Measurement precision
If detailed spectral analysis is applied to characterize noises, then noise characterization improves, but processing complexity increases
Solution Approach 1:
The patent extracts only the most relevant frequency domain parameters from the full spectral analysis to create compact noise fingerprints. By taking out and retaining only the essential characteristic frequencies and amplitudes needed for identification, the system achieves precise noise characterization while minimizing processing complexity and data storage requirements
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of analyzing a noise signal includes receiving, via a local engine control unit (ECU), a noise signal sensed by a knock sensor disposed in a reciprocating device. The method further includes processing the noise signal via at least one of the local ECU, a remote ECU, or an external system. The processing includes preconditioning the noise signal to derive a preconditioned noise signal, and applying an ADSR envelope to the preconditioned noise signal. The processing additionally includes extracting tonal information from the preconditioned noise signal and creating a fingerprint of the noise signal based on the ADSR envelope, the tonal information, or a combination thereof.