Machining State Estimation Using Adapted Acoustic Learning Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for determining the machining state of a workpiece using sound or vibration data are hindered by the need for extensive data collection for each device type, leading to high costs and reduced accuracy due to differences in acoustic characteristics between devices.
Innovation Solution
A data generation device that utilizes large-scale data from one machine tool to generate adapted learning data for another, compensating for differences in acoustic characteristics through feature quantity conversion, thereby reducing the need for extensive data collection for each device type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If learning data is collected for each device type separately, then determination accuracy is maintained, but cost and time resources increase significantly
Solution Approach 1:
The patent applies parameter changes by converting feature quantities of observation data to adapt to different device types. Instead of collecting device-specific data, the system transforms the features of general learning data to match the acoustic characteristics of target devices, thereby maintaining determination accuracy while avoiding time-consuming device-specific data collection
Solution Approach 2:
The patent implements universality by creating a general learning data set that can be adapted to multiple device types through feature quantity conversion. The determination model learns universal patterns that can be applied across different devices, eliminating the need for separate data collection for each device type
2Measurement precision
If learning data is collected for each device type separately, then determination accuracy is maintained, but cost resources increase significantly
Solution Approach 1:
The system transforms feature quantities of observation data to adapt to different device types, allowing a single general learning data set to serve multiple devices. This parameter transformation approach maintains determination accuracy while eliminating the need for expensive device-specific data collection
Solution Approach 2:
The patent creates adapted learning data by transforming features of general learning data to match target device characteristics. This copying and adaptation approach allows reuse of existing data across device types, significantly reducing data collection costs while maintaining accuracy
3Productivity
If a determination model is trained on one device type and applied to another, then resource requirements are reduced, but determination accuracy decreases due to acoustic characteristic differences
Solution Approach 1:
The patent resolves this contradiction by transforming feature quantities of observation data to account for acoustic characteristic differences between device types. This parameter adaptation allows the determination model to maintain high accuracy when deployed across different devices without requiring device-specific training data
Solution Approach 2:
The system introduces feature quantity conversion as an intermediary process between the determination model and devices with different acoustic characteristics. This intermediary transformation layer enables the model to interpret observation data from various devices accurately, bridging the gap between different acoustic environments
Data Source
AI summary
A data generation device includes a large-scale data acquisition unit that obtains large-scale data that is large-scale learning data used in learning of a first determination model for determining a machining state of a workpiece machined by a first machine tool; an adaptive data acquisition unit that obtains adaptive data for use in generation of learning data for use in learning of a second determination model for determining a machining state of a workpiece machined by a second machine tool; and a learning data generation unit that converts the large-scale data based on the adaptive data to generate adapted large-scale data for use in learning of the second determination model.