Neural Sound Monitoring for Real-Time Abnormal Pressure Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ultrasonic cleaning processes in semiconductor manufacturing lack real-time monitoring and early warning for abnormal sound pressures, relying on offline and time-consuming manual sound pressure meter measurements.
Innovation Solution
A neural network model is trained using sound signals from a device under test to detect abnormalities in real-time, utilizing a sound receiving apparatus and an electronic device with a processor to analyze sound signals and output notification signals based on probability values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a sound pressure meter is used for offline detection, then measurement accuracy is improved, but real-time monitoring capability deteriorates
Solution Approach 1:
The patent replaces the mechanical sound pressure meter measurement system with an acoustic sensor coupled to a neural network model. The acoustic sensor continuously captures sound signals, and the neural network model processes these signals to detect abnormal sound pressures in real-time, eliminating the need for manual offline measurements while maintaining detection accuracy.
Solution Approach 2:
The patent introduces a neural network model as an intermediary between the acoustic sensor and the detection system. This intermediary processes the raw sound signals, extracts features, and identifies abnormal sound pressures, enabling real-time monitoring while preserving the measurement precision that would otherwise require manual sound pressure meter readings.
2Measurement precision
If manual sound pressure meter measurement is performed, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The patent replaces the manual mechanical measurement process with an automated system comprising an acoustic sensor and neural network model. The acoustic sensor continuously captures sound signals, and the neural network automatically processes these signals to detect abnormal sound pressures, eliminating manual intervention entirely and dramatically improving detection speed while maintaining measurement precision.
Solution Approach 2:
The system performs self-service detection through the neural network model, which automatically processes sound signals and identifies abnormal sound pressures without requiring manual operation. The model continuously monitors the acoustic environment and autonomously detects anomalies, improving both productivity and measurement precision simultaneously.
3Device complexity
If conventional detection methods are used, then device complexity is reduced, but loss of time increases
Solution Approach 1:
The patent replaces conventional manual detection methods with a neural network-based acoustic analysis system. The neural network model processes sound signals in real-time, providing rapid detection of abnormal sound pressures. While the system complexity increases due to the neural network, the detection time is dramatically reduced from manual offline measurement to continuous real-time monitoring.
Data Source
AI summary
Disclosed are a signal processing method and an abnormal sound detection system. First, a neural network model is trained, including: (a) randomly selecting a plurality of sample signals from a training database to obtain a combined signal, wherein the training database includes the first sample set belonging to the first classification label and the second sample set belonging to the second classification label, and the number of selected sample signals conforms to the preset number; repeating the said step to obtain a plurality of combined signals and using the combined signals to train the neural network model. Then, a sound signal received from a sound receiving apparatus is inputted to the trained neural network model to output a probability value, and a corresponding notification signal is outputted based on the probability value.