Grain Size Estimation Device Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The determination of grain size in metallographic inspection of steel products is highly dependent on the knowledge and experience of individual inspectors, making it challenging to standardize the process and maintain consistency when inspectors retire.
Innovation Solution
A grain size estimation device that uses machine learning to estimate grain sizes based on predictive models generated from images of metal surfaces, allowing for automated evaluation without relying on inspector expertise, incorporating a measurement/evaluation device connected to a microscope, camera, and X-Y stage for image acquisition and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inspection by inspectors is used to determine grain size, then the determination can be performed with human judgment and experience, but the process becomes highly dependent on individual inspector knowledge making standardization difficult
Solution Approach 1:
The patent replaces the mechanical visual inspection system with an automated image processing system that captures metallographic images and uses algorithm-based grain size measurement. The system substitutes human inspectors' visual judgment with computational methods, including automated edge detection, grain boundary identification, and statistical analysis of grain size distributions, thereby eliminating dependency on individual inspector expertise while maintaining measurement accuracy.
Solution Approach 2:
The patent transforms the qualitative visual assessment parameter into quantitative measurable parameters. Instead of relying on inspectors' subjective visual evaluation, the system converts grain size determination into objective numerical measurements through image processing algorithms that calculate grain area, perimeter, and equivalent diameter, enabling standardized and repeatable results across different inspectors and laboratories.
2Adaptability or versatility
If automated machine learning estimation is used for grain size determination, then standardization and consistency are improved, but the device complexity increases
Solution Approach 1:
The patent introduces an intermediary machine learning model that acts as a bridge between raw metallographic images and grain size estimates. The system uses a trained neural network or predictive model as the intermediary component, which has been pre-trained on labeled training data containing images and corresponding grain size values. This intermediary model handles the complex pattern recognition and estimation tasks, allowing the overall system to achieve high standardization without requiring complex real-time processing infrastructure.
Solution Approach 2:
The patent performs preliminary actions by pre-training the machine learning model offline using extensive training data before deployment. The model is trained in advance on a dataset of metallographic images with known grain sizes, allowing the system to learn complex patterns and relationships beforehand. This preliminary training phase separates the complex learning process from the actual inspection operation, simplifying the deployed system while maintaining high accuracy and standardization capability.
3Productivity
If more inspectors are hired to meet increased inspection demands, then inspection capacity increases, but labor costs and variability between inspectors increase
Solution Approach 1:
The patent implements a self-service automated inspection system that performs grain size determination without requiring human inspectors for each measurement. The system automatically captures images, processes them through the machine learning model, and generates grain size estimates independently. This self-service capability allows the system to handle increased inspection volumes without additional human resources, while maintaining consistent results through automated standardized processing of all samples.
Data Source
AI summary
According to one embodiment, a grain size estimation device includes an acquisition unit that acquires a captured image of a surface segment of an object inducing metal; and an estimation unit that estimates a grain size of the surface segment of the object indicated in the acquired image, based on a predictive model generated by machine learning using images of metal surfaces and grain sizes in the metal surfaces as training data.