Inverted ML Modeling for Semiconductor Process Input Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional manufacturing process modeling methods are inefficient as they require numerous experiments and simulations to identify optimum input settings, especially when dealing with complex processes involving multiple inputs and varying configurations, leading to increased resource consumption and processing time.
Innovation Solution
The use of a set of homogeneous inverted machine learning models that perform linear extrapolation to predict input data for semiconductor device manufacturing processes based on expected output data, reducing the need for physical experiments and simulations by taking the end result as input and outputting the required input settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional manufacturing process modeling methods are used to identify optimum input settings, then manufacturing precision can be achieved, but the number of experiments and simulations increases significantly, leading to increased loss of time and loss of energy
Solution Approach 1:
The patent inverts the traditional modeling approach by training machine learning models to predict input settings from output specifications rather than predicting outputs from inputs. This inversion allows the system to directly determine optimum input settings for desired outputs without requiring extensive forward simulations and experiments, thereby reducing time loss while maintaining manufacturing precision
Solution Approach 2:
The patent performs preliminary action by pre-training multiple machine learning models on historical manufacturing data before actual production. These pre-trained models can then quickly predict optimum input settings for new output specifications without requiring real-time experiments or simulations, significantly reducing the time needed to identify optimal parameters while ensuring manufacturing precision
2Manufacturing precision
If traditional manufacturing process modeling methods are used to identify optimum input settings, then manufacturing precision can be achieved, but numerous experiments and simulations are required, leading to increased loss of energy
Solution Approach 1:
By inverting the prediction direction to determine inputs from outputs, the system avoids the need for numerous energy-consuming forward simulations and physical experiments that are required in traditional methods, thereby reducing energy loss while maintaining the ability to achieve manufacturing precision
Solution Approach 2:
The patent creates virtual copies of the manufacturing process through machine learning models that replicate the relationship between inputs and outputs. These digital twins can predict optimal settings without requiring physical experiments or energy-intensive simulations, reducing energy consumption while maintaining prediction accuracy for manufacturing precision
3Manufacturing precision
If process engineers manually select and customize settings based on domain expertise, then manufacturing precision can be achieved, but the complexity of the process increases and productivity decreases
Solution Approach 1:
The system enables self-service by allowing the machine learning models to automatically determine optimum input settings based on desired outputs without requiring manual intervention from process engineers. This automation maintains manufacturing precision through algorithmic optimization while dramatically improving productivity by eliminating time-consuming manual analysis and iteration
Solution Approach 2:
The patent replaces the mechanical system of manual engineering analysis and iterative experimentation with an automated computational system based on machine learning. This substitution maintains the precision of expert-driven settings selection while vastly improving productivity through rapid automated prediction and optimization
Data Source
AI summary
Disclosed herein is technology for performing predictive modeling to identify inputs for a manufacturing process. An example method may include receiving expected output data defining an attribute of a semiconductor device manufactured by at least one semiconductor device manufacturing process performed within at least one processing chamber, wherein the expected output data corresponds to an unexplored portion of a process space associated with the at least one semiconductor device manufacturing process, and identifying expected input data by using the expected output data as input to a plurality of homogeneous inverted machine learning models, wherein each inverted machine learning model of the plurality of homogeneous inverted machine learning models is trained to determine, by performing linear extrapolation based on the expected output data, a respective set of input data of a plurality of sets of input data for configuring the semiconductor device manufacturing process to manufacture the semiconductor device.