Research on Diesel Engine Fault Status Identification Method Based on Synchro Squeezing S-Transform and Vision...

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Diesel engine reliability and safety decline with operational time, causing frequent failures.
• Traditional fault identification methods struggle with accurate diesel engine fault detection.
• Existing signal time-frequency analysis methods exhibit low resolution and weak energy aggregation.

Purpose of the Study:

• To develop an advanced diesel engine fault status identification method.
• To overcome limitations of traditional methods in accurately identifying diesel engine faults.
• To leverage the strengths of SSST for signal processing and ViT for image classification.

Main Methods:

• Vibration signals from diesel engines were collected using sensors.
• Synchro squeezing S-transform (SSST) converted vibration signals into 2D time-frequency maps, enhancing resolution and energy aggregation.
• Vision Transformer (ViT) model extracted features from these time-frequency maps for fault status assessment.

Main Results:

• The proposed SSST-ViT method demonstrated superior performance compared to ST-ViT, SSST-2DCNN, and FFT spectrum-1DCNN.
• High fault status identification accuracy achieved: 98.31% on a public dataset.
• Excellent accuracy of 95.67% obtained on actual laboratory data.

Conclusions:

• The SSST-ViT method offers a robust and accurate approach for diesel engine fault status identification.
• This technique effectively addresses the challenges of non-linear and non-smooth signal processing in fault detection.
• The study provides a promising new direction for enhancing diesel engine condition monitoring and maintenance.