Machine Tool Sensor Modeling for Early Malfunction Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting malfunctions in manufacturing machines, such as machine tools, struggle to accurately predict the onset of abnormal states due to variations in maintenance operations and the difficulty in generating learning models that cover all possible normal states, leading to inefficient maintenance and potential production stoppages.
Innovation Solution
An information processing method and apparatus that utilize machine learning to generate and update models based on sensor data from different operational periods, allowing for the detection of malfunctions by comparing measurement values against initial and updated models, thereby improving prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single learning model is used for abnormality detection in machine tools, then the detection method is simple, but it cannot accurately detect abnormal states across different processing kinds
Solution Approach 1:
The patent segments the detection system by creating multiple learning models, each specialized for a specific processing kind. The system divides the overall detection task into separate models (first learning model for initial state, second learning model for post-maintenance state) that can be selectively applied based on the processing type, thereby achieving high accuracy without requiring a single overly complex universal model
2Reliability
If maintenance operations are performed frequently to prevent abnormal states, then preventive safety increases, but the operation rate of the manufacturing machine decreases
Solution Approach 1:
The patent implements preliminary action by detecting signs of malfunction before the abnormal state actually occurs. The learning models identify early deviations from normal operation patterns, allowing maintenance to be scheduled proactively at optimal moments rather than through frequent preventive maintenance, thus maintaining high operation rates while ensuring reliability
Solution Approach 2:
The system continuously monitors operation status and provides feedback through the learning models that compare current state against learned normal patterns. This real-time feedback enables dynamic maintenance scheduling based on actual machine condition rather than fixed intervals, optimizing both reliability and productivity
3Adaptability or versatility
If a learning model is updated by adding data from erroneous detection periods, then the model adapts to new normal states, but it may fail to detect actual malfunction signs
Solution Approach 1:
The patent segments the learning data by creating distinct learning models for different operational phases (initial state model and post-maintenance state model). Each model is trained on data specific to its phase, preventing contamination from erroneous periods while maintaining adaptability to the specific characteristics of each operational state
Solution Approach 2:
The system dynamically selects which learning model to apply based on the current operational context and maintenance status. This dynamic approach allows the system to adapt to different normal states without compromising detection precision, as each model operates within its validated domain
Data Source
AI summary
An information processing apparatus includes a controller. The controller is configured to obtain a measurement value of a sensor provided in mechanical equipment. The controller is configured to generate a first model by machine learning using the measurement value of the sensor measured in a first period of the mechanical equipment and store the first model in a storage portion. The controller is configured to generate a second model by machine learning using the measurement value of the sensor measured in a second period after a trigger event has occurred in the mechanical equipment and store the second model in the storage portion. The controller is configured to determine a state of the mechanical equipment by using the measurement value of the sensor measured in an evaluation period and the first model and the second model stored in the storage portion.