Acoustic Equipment Diagnosis Using Spectrogram-Based ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current equipment diagnosis methods are subjective and do not account for previous diagnoses, leading to inaccurate results and unnecessary repairs or equipment downtime, as they fail to objectively assess the operation of complex systems like vehicle components.
Innovation Solution
A system that records audio of equipment operation, transforms it into image data using techniques like mel spectrograms, and inputs this data into a machine learning model to determine if the operation is desired or undesired, allowing for accurate diagnosis and potential repair or operation adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If subjective diagnosis methods are used by individuals, then the diagnosis process is simple and quick, but the accuracy and reliability of the diagnosis deteriorates
Solution Approach 1:
The patent replaces subjective human mechanical diagnosis with an automated acoustic analysis system using microphones, processors, and machine learning algorithms to objectively detect equipment anomalies through sound signatures, thereby maintaining fast diagnosis speed while significantly improving accuracy
Solution Approach 2:
The patent introduces acoustic signals as an intermediary between the equipment and the diagnosis system. The audio recordings serve as objective mediators that capture equipment states without human subjectivity, enabling accurate and repeatable diagnosis through automated analysis of sound patterns
2Device complexity
If traditional diagnosis methods are used without considering previous diagnoses, then the diagnosis process is simple, but the reliability of determining correct diagnosis deteriorates
Solution Approach 1:
The patent implements feedback by storing and analyzing historical audio recordings and diagnosis results. The system compares current acoustic signatures with previous diagnoses to track equipment degradation over time and verify diagnosis consistency, thereby improving reliability without significantly increasing process complexity
Solution Approach 2:
The patent performs preliminary actions by continuously recording and archiving audio data during normal equipment operation. This pre-captured data is readily available for immediate analysis when diagnosis is needed, enabling reliable comparison with historical states without adding complex real-time monitoring infrastructure
3Ease of manufacture
If incorrect diagnosis occurs, then unnecessary part replacement is performed, but equipment downtime and repair costs increase
Solution Approach 1:
The patent replaces subjective human judgment in repair decision-making with objective acoustic analysis. By automatically detecting true equipment anomalies through sound signature analysis, the system prevents false positives that would lead to unnecessary part replacements and equipment downtime, while maintaining ease of repair through clear diagnostic guidance
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides objective and accurate diagnosis of equipment operation, reducing unnecessary repairs and downtime by leveraging deep learning to identify patterns and anomalies in audio signatures, thus improving the reliability of equipment assessment.
Implementation Method 1
an audio sensor to record operation of equipment to generate an audio file
Implementation Method 2
inputting the image data into a machine learning model to determine whether the image data is indicative of a desired operation of the equipment or an undesired operation of the equipment
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
A method includes recording operation of equipment into an audio file and transforming the audio file into image data. The image data is input into a machine learning model to determine whether the image data is indicative of a desired operation of the equipment or an undesired operation of the equipment. A system includes an audio sensor configured to record operation of equipment and create an audio file, and one or more processors. The one or more processors transform the audio file into image data and input the image data into the machine learing model to determine whether the image data is indicative of a desired operation of the equipment or an undesired opertion of the equipmnt.