A distribution-based selective optimization method for eliminating periodic defects in harmonic signals.

Area of Science:

• Signal Processing
• Metrology
• Data Analysis

Background:

• Measurement signals are often corrupted by environmental noise and periodic defects.
• Existing methods struggle to accurately differentiate defect components from valid signals.
• This limitation hinders precise defect detection and signal analysis.

Purpose of the Study:

• To propose a novel distribution-based selective optimisation method (SOM) for mitigating noise and defect effects.
• To develop a technique capable of simultaneously removing periodic defect components and performing signal-fitting regression.
• To validate the effectiveness, accuracy, and feasibility of the SOM in both theoretical and real-world scenarios.

Main Methods:

• The selective optimisation method (SOM) is introduced as a signal classifier based on error distribution.
• It operates as a binary- or multiple-class classifier to identify and isolate defect components.
• The method integrates signal-fitting regression alongside defect component elimination.

Main Results:

• Theoretical simulations demonstrate the SOM's capability to separate defect components from measurement signals.
• The method achieves satisfactory signal-fitting regression, producing accurate curve fits.
• Criteria for selecting operation variables under various parameter conditions are detailed.

Conclusions:

• The proposed SOM effectively mitigates noise and periodic defects in measurement signals.
• It offers accurate separation of defect components and robust signal fitting.
• The SOM shows broad applicability in signal processing and defect detection for mechanical, electronic, and instrumentation fields.