Semiconductor Substrate Quality Evaluation Using Neural Network Reliability Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI-based evaluation methods for semiconductor substrates may not always provide accurate quality assessments, leading to a need for improved accuracy in evaluating the reliability of machine learning outputs to enhance overall product quality.
Innovation Solution
A system comprising an evaluation device with a neural network trained on images of semiconductor substrates and sensor data, which determines the probability of a machine learning device outputting an erroneous result, allowing for improved accuracy in quality evaluation by providing reliability information on the outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI-based evaluation methods are used for semiconductor substrate quality assessment, then productivity is improved, but measurement precision deteriorates due to potential erroneous outputs
Solution Approach 1:
An evaluation device acts as an intermediary between the machine learning device and the quality decision-making process. This device receives outputs from the machine learning device, evaluates their reliability using a trained neural network, and only accepts outputs that meet predetermined reliability thresholds, thereby filtering out erroneous assessments while maintaining high productivity
Solution Approach 2:
The system implements feedback by using the evaluation device to assess the reliability of machine learning outputs and feed this reliability information back into the quality evaluation process. The neural network in the evaluation device learns from labeled data about erroneous outputs, creating a feedback loop that continuously improves the system's ability to identify and reject unreliable assessments
2Measurement precision
If a neural network is trained to determine probability of erroneous outputs, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The evaluation device merges the quality assessment function with the reliability evaluation function into a single integrated system. The neural network within the evaluation device simultaneously processes substrate images and determines the probability of erroneous outputs from the machine learning device, combining multiple functions into one unified device rather than requiring separate systems
Solution Approach 2:
The system uses a second neural network in the evaluation device that is trained to replicate or predict the error patterns of the first machine learning device. This copied error-detection capability allows the system to identify potential failures without needing to understand the complex internal workings of the original machine learning model
Data Source
AI summary
An evaluation device may include: a receiving unit that receives an image of the semiconductor substrate, the image captured by an imaging device provided on the semiconductor substrate manufacturing apparatus; a determination unit that determines, using a neural network, at least one value representative of a probability of a machine learning device outputting an erroneous output for the image, the machine learning device configured to: (i) receive the image of the semiconductor substrate, (ii) perform computation using the received image, and (iii) output information indicating the quality of the semiconductor substrate based on a result of the computation; and an output unit that outputs an output based on the at least one value representative of the probability. The neural network has been trained using: images of manufactured semiconductor substrates; and information indicating, for each one of the images, a level of erroneousness for an output from the machine learning device.