Robot Fault Detection Using Motion-Insensitive Signal Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Factory robots in production lines face unforeseen failures that can disrupt entire operations, leading to significant downtime and economic losses, as existing maintenance methods are either inefficient or require predictive detection of machine deterioration.
Innovation Solution
A fault-monitoring and detection system that extracts motion-insensitive features from current, acoustic, or vibration signals using signal-processing techniques like Hilbert transform and short-time Fourier transform, combined with unsupervised machine-learning methods like k-means clustering, to differentiate between normal and abnormal robot operations, enabling condition-based maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If schedule-based maintenance is performed at predetermined time intervals, then maintenance can be planned in advance, but unnecessary maintenance increases downtime and reduces asset availability
Solution Approach 1:
The patent transforms the maintenance approach by changing the parameter from time-based intervals to condition-based thresholds. Motion-insensitive features are extracted from sensor data and compared against predetermined thresholds to determine when maintenance is actually needed, rather than following a fixed schedule. This allows maintenance to be performed only when necessary, reducing unnecessary downtime while ensuring reliability.
Solution Approach 2:
The patent replaces the mechanical time-scheduling system with an intelligent detection system using motion-insensitive feature extraction from sensor data. Instead of relying on predetermined time intervals, the system uses signal processing and pattern recognition to detect actual machine conditions, substituting a rigid temporal mechanism with an adaptive condition-based mechanism.
2Reliability
If condition-based maintenance is implemented to reduce unnecessary maintenance, then asset availability increases, but detecting machine deterioration before failure requires sophisticated sensing and analysis
Solution Approach 1:
The patent extracts motion-insensitive features from complex sensor data by removing the motion component. Standard sensors (accelerometers, gyroscopes) continue to be used, but the patent extracts specific features from their data that are insensitive to motion, thereby simplifying the analysis while maintaining the ability to detect deterioration. This extraction approach reduces complexity compared to using specialized sensors.
Solution Approach 2:
The patent makes standard motion sensors multi-functional by extracting both motion information and motion-insensitive vibration information from the same sensors. This eliminates the need for separate specialized sensors, reducing system complexity while enabling condition-based maintenance detection.
3Device complexity
If standard vibration sensors are used to detect machine deterioration, then the system is simple to implement, but the sensors cannot distinguish between motion-induced vibrations and fault-induced vibrations
Solution Approach 1:
The patent introduces motion-insensitive features as an intermediary between standard vibration sensors and fault detection. These features act as a mediator that filters out motion-induced vibrations while preserving fault-induced vibration signals, allowing standard sensors to achieve precise fault detection without needing to be replaced with specialized equipment.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
One embodiment can provide a system for detecting faults in a machine. During operation, the system can obtain a dynamic signal associated with the machine, apply one or more signal-processing techniques to the dynamic signal to obtain frequency, amplitude, and/or time-frequency information associated with the dynamic signal, extract motion-insensitive features from the obtained frequency, amplitude, and/or time-frequency information associated with the dynamic signal, and determine whether a fault occurs in the machine based on the extracted features.