Intelligent Identification of Centrifugal Pump Damage by Combining Methods Independent Component Analysis and Particle Swarm Optimization
Subject Areas :
Mohammad sadegh
Aalaei
1
(Department of Mechanical Engineering, Ahvaz Branch, Islamic Azad University, Ahvaz, Iran.)
Mehdi
Shekarzadeh
2
(Department of Mechanical Engineering, Ahvaz Branch, Islamic Azad University, Ahvaz, Iran.)
Keywords: Bearing, particle swarm optimization, Intelligent Identification, Pump Damage, Independent Component Analysis,
Abstract :
Due to the progress of technical and engineering sciences and the more complex equipment and machinery in recent years, the maintenance and repair technology based on condition monitoring and defect estimation, under different titles such as performance-based logistic (PBL) and condition-based maintenance (CBM) is considered. These methods are used to prevent human and financial losses and to increase the production rate. This thesis presents an intelligent troubleshooting system to diagnose centrifugal pump-bearing faults. As a result, to design this intelligent troubleshooting system, a test set including shaft, bearings and real support conditions was designed and implemented in the laboratory. In this setup, three bearings with Normal wear and fault conditions (defect on the outer race) were examined, and vibration data were obtained. Then, the vibration data in extraction time and statistical features were calculated. After that, these features were used as classifier input data for intelligent troubleshooting. To identify the defect, the independent component analysis method was used. Also, the accuracy of fault detection was improved by using the particle batch optimization method. Finally, it was found that the statistical feature of Percentile can detect bearing defects by combining independent component analysis and particle swarm optimization methods.
_||_
