Professional Manufacture 0 432 193 420 Diesel Injector Common Rail Injector Engine Parts Vehicle Parts 0432193420

Abstract:

Accurate and efficient fault diagnosis of fuel injectors is essential for ensuring stable engine performance, improving fuel economy, and reducing emissions. Traditional diagnostic approaches based on pressure and flow measurement often fail to detect early-stage or transient faults. To address this limitation, this study proposes a vibration signal-based diagnostic framework that integrates advanced feature extraction and pattern recognition techniques for injector fault identification under varying operating conditions.

A high-pressure common-rail injector test bench was established to collect vibration data under normal and fault conditions, including needle sticking, leakage, delayed opening, and partial blockage. Vibration signals were acquired using high-sensitivity piezoelectric accelerometers mounted near the injector body, sampled at 100 kHz to capture transient impact events during each injection cycle.

In the signal processing stage, wavelet packet decomposition (WPD) and empirical mode decomposition (EMD) were applied to separate time–frequency components and isolate characteristic features related to valve impact, needle motion, and fuel hammer effects. Extracted features included energy entropy, kurtosis, root mean square (RMS), and dominant frequency band energy ratios. Principal component analysis (PCA) was used to reduce dimensionality and eliminate redundant information.

For fault pattern recognition, three machine learning classifiers—Support Vector Machine (SVM), Random Forest (RF), and Convolutional Neural Network (CNN)—were evaluated. The CNN model achieved the highest recognition accuracy of 97.3%, demonstrating superior capability in capturing nonlinear feature interactions and temporal dynamics compared to traditional algorithms. The diagnostic framework also maintained robustness under varying injection pressures (80–160 MPa) and noise conditions, with less than 3% accuracy loss.

The results show that vibration signal analysis provides a reliable, non-intrusive, and cost-effective approach for injector condition monitoring. The proposed method enables real-time fault identification and classification, particularly suitable for on-board diagnostics (OBD) and predictive maintenance applications in intelligent diesel engine systems.

This research establishes a data-driven diagnostic paradigm that combines vibration feature analysis with intelligent recognition algorithms, offering theoretical and practical foundations for the next generation of smart, self-learning injector health monitoring systems.