Machine Vibration Monitoring Using Learned Anomaly Signatures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vibration monitoring methods for machines, relying on amplitude threshold exceedances and dominant frequency variations, are limited in detecting anomalies at unforeseen frequencies, leading to potential undetected issues and subsequent damage.
Innovation Solution
A method involving a learning phase to establish a knowledge base of normal vibratory signatures and a monitoring phase to detect deviations, using similarity analysis and thresholds to identify anomalies, with periodic data acquisition and wireless transmission of warnings, suitable for resource-constrained devices like microcontrollers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If amplitude threshold exceedances and dominant frequency variations are used for anomaly detection, then the detection method is simple to implement, but it cannot detect anomalies at unforeseen frequencies
Solution Approach 1:
The system performs a learning phase before actual monitoring to build a knowledge base of normal vibratory signatures. This preliminary action enables the system to adapt to the specific machine and its operating conditions, allowing detection of anomalies at unforeseen frequencies while maintaining implementation simplicity.
Solution Approach 2:
The system automatically learns and adapts to the machine's normal vibrations without requiring manual configuration of frequency ranges or thresholds. The autonomous learning process eliminates the need for expert intervention while expanding detection capabilities to cover unexpected anomaly frequencies.
2Reliability
If all vibratory signatures are saved in the knowledge base, then the learning quality is improved, but the data storage requirement increases
Solution Approach 1:
Instead of uniformly storing all vibratory signatures, the system selectively saves only those signatures that represent novel operational states or deviations from normal patterns. This targeted approach maintains learning quality by preserving important variations while minimizing redundant data storage.
Solution Approach 2:
The system discards redundant vibratory signatures that are similar to already stored patterns, keeping only distinctive signatures that provide new information. This selective retention strategy maintains the reliability of the knowledge base while reducing storage requirements for resource-constrained devices.
3Reliability
If continuous monitoring is performed to detect anomalies early, then the reliability is improved, but the energy consumption increases
Solution Approach 1:
The system performs monitoring at periodic intervals rather than continuously, acquiring vibratory signals at predetermined time points. This periodic approach maintains anomaly detection reliability by capturing relevant vibration patterns while significantly reducing energy consumption compared to continuous monitoring.
Solution Approach 2:
The learning phase is performed beforehand to establish the knowledge base, enabling the system to efficiently compare subsequent measurements against learned patterns. This preliminary preparation reduces the computational burden during actual monitoring, allowing reliable detection with lower energy consumption.
Data Source
AI summary
A method for monitoring the operation of a machine that generates vibrations, includes a learning phase in which a knowledge base containing vibrational signatures representative of the operation of the machine is generated, and a monitoring phase in which the vibrations of the machine are compared to the knowledge base so as to detect an anomaly in the machine.