Intelligent method for diagnosing structural faults of rotating machinery using ant colony optimization

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Structural faults in rotating machinery, including unbalance, misalignment, and looseness, are common causes of serious accidents and significant production losses.
• Early and accurate detection of these faults is crucial for preventing catastrophic failures and minimizing economic impact.

Purpose of the Study:

• To propose an intelligent method for early-stage detection and classification of structural faults in rotating machinery.
• To introduce novel 'relative ratio symptom parameters' (RRSPs) for characterizing vibration signal features.
• To evaluate the effectiveness of the proposed method against conventional techniques like neural networks (NN).

Main Methods:

• Development of 'relative ratio symptom parameters' (RRSPs) to capture vibration signal features under different fault conditions.
• Definition of a 'synthetic detection index' (SDI) based on statistical theory to assess RRSP suitability for ant colony optimization (ACO).
• Application of ACO combined with RRSPs for intelligent fault diagnosis and comparison with conventional neural network methods.

Main Results:

• The proposed method, utilizing ACO and RRSPs, demonstrated high effectiveness in identifying common structural faults in centrifugal fans, such as unbalance, misalignment, and looseness.
• The synthetic detection index (SDI) proved useful in evaluating the applicability of RRSPs for the ACO algorithm.
• Practical examples confirmed the superior performance of the proposed method compared to conventional neural networks for detecting these specific faults.

Conclusions:

• The intelligent method integrating ant colony optimization (ACO) and relative ratio symptom parameters (RRSPs) offers a robust solution for early and accurate diagnosis of structural faults in rotating machinery.
• The developed RRSPs and SDI provide a valuable framework for feature extraction and selection in fault diagnosis.
• This approach significantly outperforms conventional neural networks in detecting challenging structural faults in rotating machinery, particularly in centrifugal fan applications.