Thermal Conductivity Estimation via Machine Learning Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating thermal conductivity in semiconductor crystal growth units require actual crystal growth and are limited by the need for specialized measurement apparatus and the temporal changes of high-temperature components, making accurate heat-transfer simulations challenging and impractical.
Innovation Solution
A method and apparatus using a measurement sample from the production apparatus, where a regression model is created through machine learning based on heat-transfer simulations with varying thermal conductivities and heating conditions, allowing for the estimation of thermal conductivity without specifying the material, enabling easy analysis of heat-transfer patterns during semiconductor crystal production.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If heat parameters are estimated based on actual crystal growth measurements, then the thermal conductivity estimation accuracy is improved, but the measurement process becomes complex and requires specialized apparatus
Solution Approach 1:
The invention creates a virtual copy (simulation model) of the crystal growth apparatus and processes. Instead of measuring thermal conductivity directly in the complex actual apparatus, the patent uses computational models that replicate the physical system's behavior. This allows thermal conductivity estimation without requiring specialized measurement apparatus, resolving the contradiction between measurement accuracy and device complexity.
Solution Approach 2:
The patent replaces physical measurement systems with computational analysis. Instead of using specialized thermal conductivity measurement apparatus, the invention uses heat transfer simulations and image processing algorithms to estimate thermal conductivity from temperature distribution data obtained during normal crystal growth operations. This substitution eliminates the need for complex measurement devices while maintaining estimation accuracy.
2Adaptability or versatility
If heat parameters are modified to match large-diameter monocrystal properties, then the simulation applicability is improved, but the measurement time and productivity are reduced
Solution Approach 1:
The patent performs preliminary actions by pre-processing images to extract temperature distribution data before conducting thermal conductivity estimation. The system prepares simulation models and calibration data in advance, allowing rapid estimation once measurements are taken. This preliminary preparation significantly reduces the overall measurement time while maintaining the ability to adapt to different crystal sizes and growth conditions.
Solution Approach 2:
The invention changes key parameters from direct thermal conductivity measurement to temperature distribution analysis. By measuring temperature fields during normal crystal growth and using these data in heat transfer simulations, the system can estimate thermal conductivity without stopping production for specialized measurements. This parameter transformation enables both high adaptability to different crystal types and maintained productivity.
3Measurement precision
If temperature distribution measurement is performed during actual crystal growth, then the measurement accuracy is improved, but the production time is extended
Solution Approach 1:
The patent merges the temperature measurement function with the normal crystal growth process. Temperature distribution is measured using existing imaging equipment during routine production operations, rather than as a separate measurement step. This integration allows accurate temperature data collection without extending production time, as the measurement occurs concurrently with crystal growth operations.
Solution Approach 2:
The system uses its own operational data (temperature fields during crystal growth) for thermal conductivity estimation purposes. The crystal growth process itself generates the measurement data needed, eliminating the need for separate measurement operations. This self-service approach maintains both measurement accuracy and production efficiency by utilizing data already available during normal operations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach improves the accuracy of thermal conductivity estimation, allowing for precise calculation without the need for actual crystal growth or specialized measurement setups, facilitating the production of semiconductor crystal products with optimized thermal properties.
Implementation Method 1
heating a part of the measurement sample under a predetermined heating condition and measuring a temperature distribution of a surface of the measurement sample under a steady state
Data Source
AI summary
A thermal conductivity estimation method includes: measuring temperature distribution of a measurement sample surface in a steady state by partially heating the measurement sample under predetermined heating conditions; calculating temperature distribution of a sample model surface by performing a heat-transfer simulation on the sample model of the same shape as the measurement sample for a plurality of combinations of provisional thermal conductivities and heating conditions; making a regression model, whose input is temperature distribution of the measurement sample surface and whose output is a thermal conductivity of the measurement sample, by a machine learning technique using training data in a form of a calculation result of the plurality of combinations and the temperature distribution obtained from the plurality of combinations; and estimating the thermal conductivity of the measurement sample by inputting a measurement result of the temperature distribution of the measurement sample surface into the regression model.