A Study on the Application of Machine and Deep Learning Using the Impact Response Test to Detect Defects on the...

Area of Science:

• Automotive Engineering
• Materials Science
• Non-Destructive Testing

Background:

• Quality control of critical automotive components like shock absorber piston rods and steering racks is essential.
• Current inspection methods (visual, sampling, tensile tests) struggle to identify defect type and location, potentially leading to premature part failure.
• Defects can cause surface cracks, accelerating deterioration and posing significant safety risks in vehicles.

Purpose of the Study:

• To develop and present a non-destructive shock response test and analysis method for efficient and accurate defect detection in automotive piston rods and steering racks.
• To propose a suitable test method and excitation frequency range for measuring defect-related changes.
• To establish a defect discrimination model using machine and deep learning.

Main Methods:

• Collected shock response data from piston rods and steering racks with varying defects.
• Engineered features from time and frequency domains of the collected data.
• Constructed defect discrimination models using machine learning and deep learning algorithms.

Main Results:

• The analysis demonstrated that defect presence and location could be effectively distinguished using frequency domain features.
• The proposed method showed superior performance in identifying defect characteristics compared to time domain analysis.
• The developed models accurately classified components based on defect presence and severity.

Conclusions:

• The shock response test method offers a viable non-destructive approach for detecting defects in automotive components.
• Frequency domain analysis is more effective than time domain analysis for this specific application.
• This technology can significantly improve quality control in automotive manufacturing, contributing to smart factory initiatives and enhanced vehicle safety.