Sensor Data Clustering for Operation-Specific Malfunction Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting malfunctions in apparatuses like robot arms, such as those used in manufacturing, are inefficient as they do not adapt to different operational types, leading to unnecessary maintenance stops and reduced productivity.
Innovation Solution
A processing apparatus that obtains and clusters sensor data in real-time, sectionalizing it based on operational conditions to differentiate between repetitive and random operations, allowing for tailored malfunction prediction without requiring information on the specific operation type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single malfunction prediction method is used for all operations, then the device complexity is reduced, but the measurement precision and reliability of malfunction prediction deteriorates
Solution Approach 1:
The patent segments the continuous sensor value data into discrete sections based on operation type detection. The processing portion divides the sensor data stream into first sections corresponding to first operations and second sections corresponding to second operations, enabling separate analysis for each operation type. This segmentation allows the system to maintain simple overall structure while achieving high prediction accuracy through operation-specific processing.
Solution Approach 2:
The processing apparatus uses a universal clustering algorithm that can handle both operation types with a single system architecture. The same clustering processing is applied to both first and second sections, but the reference values are operation-specific. This universal approach maintains device simplicity while adapting to different operation characteristics.
2Reliability
If operation-specific prediction methods are implemented, then the reliability of malfunction prediction is improved, but the device complexity increases
Solution Approach 1:
The system segments sensor data by operation type using automatic operation detection, creating separate data streams for different operations. This segmentation enables reliable operation-specific prediction without requiring separate physical hardware for each operation type, thus improving reliability while controlling complexity.
Solution Approach 2:
The processing apparatus automatically detects operation types and selects appropriate reference values without external input or manual configuration. The system self-adapts to different operations by analyzing sensor data patterns, eliminating the need for complex manual setup or external operation information inputs.
3Loss of time
If maintenance is performed based on general malfunction prediction, then the loss of time for unplanned stops is reduced, but the productivity deteriorates due to unnecessary maintenance stops
Solution Approach 1:
By segmenting maintenance prediction by operation type, the system identifies true malfunction risks specific to each operation. This prevents unnecessary maintenance stops during operations where the apparatus is functioning normally, thereby maintaining productivity while reducing unplanned downtime through targeted maintenance scheduling.
Solution Approach 2:
The system continuously monitors sensor values, compares them with operation-specific reference values, and provides feedback on malfunction risk. This feedback mechanism enables proactive maintenance scheduling based on actual apparatus condition, reducing unplanned stops while avoiding unnecessary maintenance interventions that would reduce productivity.
Data Source
AI summary
A processing apparatus includes a processing portion. The processing portion obtains, in time series, a sensor value of a first sensor output from a predetermined apparatus including the first sensor in correspondence with an operation of the predetermined apparatus, sectionalizes the sensor value of the first sensor on a basis of a predetermined condition, and clusters the sectionalized sensor value of the first sensor.