Noninvasive Methods for Fault Detection and Isolation in Internal Combustion Engines Based on Chaos Analysis

Area of Science:

• Automotive Engineering
• Signal Processing
• Noninvasive Diagnostics

Background:

• Traditional on-board diagnostics (OBD) methods are inadequate for complex automotive mechanical failures.
• Existing sensing techniques often involve invasiveness and limited physical range.

Purpose of the Study:

• To develop a fully noninvasive system for fault detection and isolation in internal combustion engines using sound signal processing.
• To validate the use of fractal dimensions and artificial neural networks for engine diagnostics.

Main Methods:

• A data acquisition system transmitting sound signals to a smartphone for processing.
• Analysis of chaotic behavior using fractal dimensions to diagnose engine misfire and alternator belt problems.
• Fault classification using an artificial neural network and comparison with wavelet multiresolution analysis.

Main Results:

• The system successfully diagnosed engine misfires and alternator belt issues with 99% accuracy.
• Fractal dimensions proved effective for diagnosing specific engine faults.
• The artificial neural network achieved high classification accuracy based on extracted fractal dimension data.

Conclusions:

• The proposed noninvasive system offers a low-cost, real-time solution for automotive fault detection.
• This approach eliminates the need for physical contact or sensor installation, enhancing diagnostic convenience.
• The method demonstrates high efficacy and feasibility for practical application in vehicle maintenance.