Early Damage Recognition Using Frequency-Transformed Signal Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for early damage recognition in machines, particularly fluid machines, are limited by the need for working loads, restricted applicability to specific types of machines, and reliance on pressure pulsations, which makes them inefficient and unsuitable for detecting preliminary damage or damage in non-dominant excitation components.
Innovation Solution
The method involves transforming signals into a frequency range before filtering, allowing for digital signal processing and differentiation of amplitude-overlaid oscillation components, thereby enabling the detection of minor changes caused by early damage mechanisms by filtering out dominant excitations and using techniques like fast Fourier transforms and bandstop filters for improved recognition and remaining lifetime prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pressure pulsations are measured to detect damage, then damage detection is possible, but the method is restricted to specific machine types and damage patterns
Solution Approach 1:
The signal is segmented into different frequency components through Fourier transformation, allowing separate analysis of dominant excitations and damage-related vibrations. This enables the method to extract damage information from complex signals across different machine types by isolating relevant frequency bands.
Solution Approach 2:
The method transforms the signal from time domain to frequency domain, changing the parameter representation from temporal pressure variations to spectral frequency components. This parameter transformation enables universal application across different machine types by focusing on frequency characteristics rather than machine-specific temporal patterns.
2Reliability
If dominant excitations are present in the signal, then the signal represents normal operation, but early damage signals are masked and undetectable
Solution Approach 1:
The method extracts and removes dominant excitations from the signal spectrum through targeted filtering in the frequency domain. By taking out these overwhelming frequency components, the previously masked early damage signals become detectable without losing the ability to represent normal operation when present.
Solution Approach 2:
The analysis moves from the time dimension to the frequency dimension, where dominant excitations and damage signals occupy different frequency spaces. This dimensional change allows simultaneous representation of normal operation (through dominant excitations) and early damage (through subtle frequency components) without mutual masking.
3Reliability
If working loads are applied to generate pressure pulsations, then damage detection is enabled, but energy consumption increases
Solution Approach 1:
The method uses the machine's own operational vibrations and ambient signals as the detection source, rather than requiring external excitation or additional working loads. The system processes signals already present during normal operation, enabling damage detection without additional energy consumption from forced excitations.
4Reliability
If pressure sensors are used to measure pressure pulsations, then damage detection is possible, but the method is limited to disturbances detectable in the pressure curve
Solution Approach 1:
The frequency-domain analysis method provides a universal detection framework that can process signals from various sensor types (pressure, vibration, acoustic) and detect multiple damage patterns including bearing defects, gear damage, and structural issues. The same spectral analysis approach adapts to different damage types and sensor modalities.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the recognition rate of machine states, improves damage diagnosis, and predicts remaining lifetimes more accurately, while avoiding false-positive damage recognition by accounting for operating parameters like speed and pressure.
Implementation Method 1
The control unit transforms the signal detected by the sensor in a frequency range by means of a fast Fourier transform
Implementation Method 2
from which excitations having a frequency in a frequency range are filtered out by means of a filter arrangement
Data Source
AI summary
A method for early damage recognition of a machine, and program and control unit for executing the method are disclosed. The method is disclosed for early damage recognition, wherein a frequency-transformed signal, filtered of dominant excitations, is supplied to a comparative early damage recognition, and wherein, after the filtering, damage of the machine is recognized by comparing the signal to a comparison value.