An optimal filter length selection method for MED based on autocorrelation energy and genetic algorithms

Area of Science:

• Signal Processing
• Mechanical Engineering
• Fault Diagnosis

Background:

• Minimal Entropy Deconvolution (MED) is sensitive to filter length selection.
• Improper filter length can hinder the accurate recovery of fault impulses.
• Detecting weak periodic impulses in machinery diagnostics is challenging.

Purpose of the Study:

• To develop an adaptive method for optimal filter length selection in MED.
• To enhance the detection of weak periodic impulses in rolling bearing fault diagnosis.
• To improve the performance of MED-based fault detection methods.

Main Methods:

• An energy ratio of autocorrelation is proposed to measure impulse salience.
• This index serves as an objective function for genetic algorithms (GA).
• An adaptive optimal filter length selection method is formulated.

Main Results:

• The proposed method successfully revealed periodic impulses from casing signals in rolling bearing fault experiments.
• The method demonstrated superior performance in detecting weak fault signals compared to other MED-based techniques.
• Experimental validation confirmed the effectiveness of the adaptive filter length selection.

Conclusions:

• The developed adaptive MED method accurately selects optimal filter length for impulse recovery.
• This technique offers improved sensitivity for detecting subtle fault signatures in rotating machinery.
• The proposed approach enhances the reliability of condition monitoring systems for rolling bearings.