Articulated Robot Failure Prediction Using Part-Level Vibration AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to accurately predict failures in articulated robots, particularly at the part level, which hinders effective maintenance and repair.
Innovation Solution
Measuring individual vibration and current values for each part of an articulated robot, analyzing correlations between these values and unit position pattern information, and using AI algorithms to predict failures based on learned data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vibration component extraction is used to determine reducer abnormality, then abnormality detection capability is improved, but failure prediction accuracy for each part is insufficient
Solution Approach 1:
The patent segments the failure prediction task by part type (motor, reducer, robot arm) and uses different prediction models for each part. Each model is trained on specific data relevant to that part, enabling accurate part-level failure prediction rather than generic abnormality detection.
Solution Approach 2:
The patent transforms raw sensor data into meaningful prediction parameters through learned feature extraction. Different parameter sets are used for different part types (e.g., current values for motors, vibration values for reducers), optimizing prediction accuracy for each specific component.
2Reliability
If comprehensive data collection and analysis is performed for accurate failure prediction, then prediction accuracy is improved, but system complexity increases
Solution Approach 1:
The system is divided into multiple independent prediction models, each dedicated to a specific part type. This segmentation allows each model to be simpler and more specialized, while collectively providing comprehensive failure prediction coverage for the entire robot system.
Solution Approach 2:
The patent uses learned data sets and models that can be copied and applied to predict failures for each occurrence of a given part type. Once a prediction model is trained for a specific part, it can be reused for similar parts, reducing overall system complexity.
Data Source
AI summary
According to one embodiment of the proposed invention, individual vibration values for each part of an articulated robot are measured, individual current values of respective motors are measured, correlations between a plurality of pieces of unit position pattern information, vibration value data, and current value data is analyzed, and a failure or possibility of failure for each part of an articulated robot is predicted.