Identification of the defective transmission devices using the wavelet transform

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Identifying failure modes in transmission devices is crucial for cost reduction and operational efficiency.
• Traditional methods may be intrusive or lack automation.
• Acoustic data offers a non-invasive approach to monitoring device health.

Purpose of the Study:

• To develop an automated system for identifying specific failure modes in defective transmission devices.
• To utilize external acoustic sensors for data acquisition.
• To improve the accuracy of failure mode identification using an enhanced machine learning approach.

Main Methods:

• Employing the wavelet transform for signal analysis.
• Using external acoustic sensors to capture operational sound data.
• Developing a feature extraction process with expert input.
• Implementing an improved learning vector quantization (LVQ) method with normalized feature vectors.

Main Results:

• The system successfully identified specific failure modes in defective transmission devices.
• The acoustic data analysis proved effective in pinpointing causes of failure.
• Experimental results validated the system's performance in a manufacturing environment.

Conclusions:

• The developed system provides an effective, automated solution for diagnosing transmission device failures using acoustic data.
• The integration of wavelet transform and improved LVQ enhances identification accuracy.
• This non-invasive technique offers practical benefits for industrial applications.