Robotic Anomaly Detection With Self-Updating Parameter Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial robots in plants face inefficiencies in anomaly detection due to models being specific to certain parameters, requiring manual updates, and high data storage costs, often resulting in false alarms and lack of awareness among operators about model parameters.
Innovation Solution
A system utilizing a Distributed Control System (DCS) with sensors to measure configuration and process parameters, applying machine learning to detect anomalies, and updating analytics models based on validated setpoints to improve anomaly detection accuracy and reduce storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a separate analytics model is built for each parameter to detect anomalies, then the anomaly detection accuracy is improved, but the device complexity and cost increase significantly
Solution Approach 1:
The patent merges multiple separate analytics models into a single unified analytics model that can handle multiple parameters simultaneously. This consolidation reduces the overall system complexity while maintaining the ability to detect anomalies across different parameters like robot position, velocity, acceleration, and torque through a single integrated model.
Solution Approach 2:
The unified analytics model is designed to be universal and multi-functional, capable of analyzing various parameters (position, velocity, acceleration, torque) within a single model framework. This eliminates the need for parameter-specific models and reduces the burden of maintaining multiple separate models while providing comprehensive anomaly detection coverage.
2Reliability
If all raw data is stored for anomaly analysis, then the anomaly detection capability is improved, but the data storage cost increases
Solution Approach 1:
The system extracts only the essential features and parameters needed for anomaly detection from the raw data, rather than storing and analyzing all raw data. The analytics model processes extracted features such as position, velocity, acceleration, and torque deviations, significantly reducing storage requirements while maintaining effective anomaly detection capability.
Solution Approach 2:
Data preprocessing and feature extraction are performed before the anomaly detection analysis. By pre-processing the data to extract relevant features and compute deviations from nominal values, the system reduces the volume of data that needs to be stored and processed, while ensuring that the essential information for anomaly detection is preserved.
3Reliability
If manual model updates are performed by third-party vendors, then the model accuracy is maintained, but the time and cost for updates increase
Solution Approach 1:
The analytics model is designed to perform self-updates using machine learning techniques. The model can automatically learn from new data, adapt to changing conditions, and update its parameters without requiring manual intervention from third-party vendors. This self-service capability maintains model accuracy while eliminating the time and cost associated with manual updates.
Solution Approach 2:
The system enables dynamic parameter changes within the analytics model through automated machine learning processes. The model can adjust its internal parameters and adapt to new operating conditions automatically, allowing for continuous improvement of model accuracy without manual reconfiguration or vendor involvement.
4Measurement precision
If operators manually select parameters for analysis, then the analysis precision is improved, but the ease of operation decreases
Solution Approach 1:
The system implements automated feedback mechanisms that monitor multiple parameters simultaneously and use machine learning to determine which parameters are most relevant for anomaly detection. This automated parameter selection process maintains high analysis precision by focusing on the most informative parameters while eliminating the need for manual operator intervention in parameter selection.
Solution Approach 2:
The analytics system performs self-configuration by automatically selecting and prioritizing parameters for analysis based on their relevance to anomaly detection. The machine learning model identifies which parameters (position, velocity, acceleration, torque) are most critical for detecting anomalies in specific situations, eliminating the need for operators to manually select parameters while maintaining high detection accuracy.
Data Source
AI summary
The present invention relates to a method and a system for detecting anomalies in a robotic system in an industrial plant. The robotic system is associated with a computing system configured to detect an anomaly in the robotic system. The computer system monitors configuration parameters of the robotic system and process parameters associated with the robotic system. Further, the computing system detects an association between at least one configuration parameter and at least one process parameter for obtaining optimal configuration parameters and optimal process parameters. The optimal configuration parameters and optimal process parameters are analyzed for detecting an anomaly. At least one parameter among the configuration parameters and the process parameters is identified causing the anomaly. Thereafter, the detected anomaly is validated, valid setpoint is estimated and the estimated valid setpoint is updated in the analytics model. The updated analytics model is subsequently used to detect anomaly accurately.