Robotic Arm Vibration Anomaly Detection by Activity Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial robotic arms face challenges in anomaly detection due to variability in operations, leading to high false alarm rates when predicting failures based on vibration data without considering operation-specific patterns.
Innovation Solution
A method involving a job categorizer to sort vibrations into activity-specific groups, followed by fluctuation-based and spectrum-based anomaly detection processes, which calculate anomaly scores and detect anomalies across multiple activities, reducing false alarms by accounting for operation-specific patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single anomaly detection process is used for all robotic arm operations, then the detection system is simple, but the false alarm rate increases due to operation variability
Solution Approach 1:
The anomaly detection system is segmented into multiple activity-specific detection processes, each trained on vibration data from a particular robotic arm activity. The system first classifies the current activity using a job categorizer, then routes the vibration data to the corresponding anomaly detection process. This segmentation allows each detection process to specialize in detecting anomalies for its specific activity, reducing false alarms caused by operation variability while maintaining manageable system complexity through modular architecture.
2Reliability
If operation-specific anomaly detection is implemented, then the false alarm rate decreases, but the system complexity increases
Solution Approach 1:
The system employs a universal job categorizer that can identify multiple different activities, and a unified anomaly detection framework that manages multiple activity-specific detection processes. This multi-functional design allows the system to handle diverse robotic arm operations through a single integrated platform, reducing false alarms by adapting to operation-specific patterns while avoiding the complexity of completely separate detection systems for each activity.
3Productivity
If vibration data is analyzed without considering activity context, then the processing is fast and simple, but the anomaly detection precision deteriorates
Solution Approach 1:
The system performs preliminary activity classification using the job categorizer before conducting anomaly detection. By first identifying what activity the robotic arm is currently performing, the system can then apply the appropriate activity-specific anomaly detection process. This preliminary action enables the subsequent anomaly detection to focus on activity-relevant patterns, improving detection precision without significantly impacting processing speed due to the efficient classification mechanism.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention defines an anomaly detection method for robotic apparatuses such as robotic arms using vibration data and involves anomaly detection, e.g., based on the fluctuations in the vibration measurements and/or frequency spectrum-based anomaly detection, e.g., based on the natural fluctuations in the vibration measurements.