Machine Learning Model for Position-Dependent Component Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing measurement systems face variability due to factors like position-dependent components, making it challenging to accurately analyze and manage the state of test apparatuses and jigs, leading to potential errors and reduced yield in device testing.
Innovation Solution
An analysis apparatus utilizing machine learning to acquire and separate position-dependent components from measured values, enabling precise error detection and state management of test apparatuses and jigs by learning models of these components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If measurement is performed by bringing a jig into contact with the device under measurement, then measurement can be conducted, but position-dependent components cause variability in measurement results
Solution Approach 1:
The patent extracts position-dependent components from measurement results by performing separate measurements at multiple positions on the device under test. By isolating and analyzing the position-dependent portion independently, the system removes its influence from the final evaluation, thereby improving measurement accuracy while maintaining system reliability.
Solution Approach 2:
The measurement process is segmented into multiple position-specific measurements rather than a single overall measurement. Each position's measurement is analyzed separately to identify position-dependent components, allowing the system to distinguish between actual device characteristics and measurement artifacts, thus resolving the contradiction between precision and reliability.
2Device complexity
If position-dependent components are not separated from measured values, then analysis is simpler, but error detection and state management become difficult
Solution Approach 1:
The patent introduces an intermediary analysis process that separates position-dependent components from device characteristics. This intermediary step acts as a mediator between raw measurements and final evaluation, automatically identifying and isolating position-dependent portions through comparative analysis, thereby improving error detection without significantly increasing operational complexity.
3Measurement precision
If machine learning is used to learn position-dependent components, then separation accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary measurements at multiple positions and uses machine learning to pre-learn position-dependent components before actual device evaluation. By establishing the position-dependent model in advance through preliminary action, the system achieves high separation accuracy during actual measurements without requiring extensive processing time during production testing.
Data Source
AI summary
There is provided an analysis apparatus including: an acquisition unit configured to acquire a plurality of measured values obtained by measuring a device under measurement; a machine learning unit configured to use the plurality of measured values to learn, by machine learning, a model of a position-dependent component that depends on a measured position in the device under measurement; and an analysis unit configured to separate, from the plurality of measured values, the position-dependent component which is calculated by using the model learned by the machine learning unit. Further, there is provided an analysis method. Further, there is provided a recording medium having recorded thereon an analysis program.