Manufacturing Quality Inspection Using Labeled Process Sections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for recording operational data from manufacturing devices only allow for continuous, uninterrupted data collection, making it difficult to prepare and analyze data for machine learning models, and do not enable retrospective assignment of quality or status data, limiting predictive maintenance and optimization efforts.
Innovation Solution
A method and apparatus that obtain and label operational data with relevant status data, allowing for the creation of training datasets for machine learning models to generate quality indicators, enabling the selection and visualization of individual process sections and facilitating predictive maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If operational data is continuously recorded without interruption, then data completeness is improved, but data usability for machine learning models deteriorates due to inability to segment and label process sections
Solution Approach 1:
The patent segments continuous operational data into discrete process sections using trigger values. When a trigger value is detected in the time series data, the system creates a new data section, allowing the continuous data stream to be divided into meaningful, labelable units that can be used for machine learning training.
Solution Approach 2:
The patent performs preliminary labeling of operational data sections during the data collection phase by associating status data with corresponding time periods. This pre-labeling creates ready-to-use training datasets before machine learning model training begins, eliminating the need for manual post-processing.
2Measurement precision
If status data is manually assigned to operational data, then data accuracy for quality inspection is improved, but productivity deteriorates due to manual work requirements
Solution Approach 1:
The system automatically performs data labeling by detecting trigger values in the operational data and autonomously creating sections with associated status data. This self-service approach eliminates manual intervention while maintaining accurate labeling, allowing the system to prepare training datasets automatically at scale.
3Measurement precision
If data collection is implemented individually for each manufacturing device, then device-specific data accuracy is improved, but system complexity increases and scalability deteriorates
Solution Approach 1:
The patent implements a universal data collection and labeling framework that can be applied to multiple manufacturing devices with different types of operational data. The system handles various data sources (sensor data, machine logs, quality measurements) and trigger types through a single unified approach, enabling consistent implementation across diverse devices without requiring device-specific customization.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A computer-implemented method for quality inspection of a component (1) of a manufacturing device (2, 3, 4), obtaining operational data (11) relating to the operation of the manufacturing device (2, 3, 4), the operational data (11) comprising time series of one or more physical properties of the manufacturing device (2, 3, 4), obtaining status (12) data relating to a component (1) of the manufacturing device (2, 3, 4), the status data (12) comprising events relating to and/or characteristic properties relevant for the utilization of the component (1) within the manufacturing device (2, 3, 4), labelling one or more subsets of the operational data (11) by associating one or more of the events and/or characteristic properties to the one or more subsets, providing the one or more subsets as labelled training data for training a machine learning model (ML), wherein the machine learning model (ML) serves for outputting a quality indicator (Q) based on the labelled training data input, and providing the trained machine learning model (ML) for quality inspection.