Adaptive Thermal Processing Control for Wafer Temperature Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Thermal processing systems face challenges in accurately tracking temperature setpoints during semiconductor wafer processing, leading to inefficiencies and errors in heat treatment processes.
Innovation Solution
A control system that uses sensors to determine actual workpiece temperatures, compares them to setpoints, and adjusts system parameters through a machine-learned system model to improve temperature tracking, incorporating submodels for radiometric, optical, and thermal simulations to refine heat control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional control methods are used for thermal processing, then the system is simpler to implement, but temperature setpoint tracking precision deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where sensors continuously measure the actual workpiece temperature, and the control system compares this measurement with the desired setpoint. The error signal is fed back to adjust the heating power, creating a closed-loop control system that automatically corrects temperature deviations and improves tracking precision.
Solution Approach 2:
The patent replaces traditional mechanical control methods with an intelligent control system that uses sensors, processors, and algorithms. Instead of relying on simple mechanical thermostats or manual control, the system employs electronic sensing and computational processing to achieve superior temperature control accuracy.
2Manufacturing precision
If adaptive control with machine learning models is implemented, then temperature tracking accuracy improves, but system complexity and computational requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models offline using historical process data. The neural network is trained beforehand to understand the complex thermal dynamics and relationships between process parameters. During actual production, the pre-trained model makes real-time predictions and control decisions without requiring complex online computations, thus improving accuracy while keeping real-time system complexity manageable.
Solution Approach 2:
The patent uses a virtual model or digital twin of the thermal processing system, created through machine learning. This copied model replicates the behavior of the actual physical system, allowing the control algorithm to operate on the virtual model's predictions rather than directly managing the complex physical system, thereby simplifying the control architecture while maintaining high precision.
3Measurement precision
If multiple submodels (radiometric, optical, thermal) are integrated, then system modeling accuracy improves, but computational load and processing time increase
Solution Approach 1:
The patent merges multiple submodels (radiometric, optical, and thermal) into a single integrated machine learning framework. Instead of running separate computational models for each physical phenomenon, the neural network learns the combined effects of all these interactions simultaneously from training data, reducing computational overhead while maintaining the accuracy benefits of multi-physics modeling.
Solution Approach 2:
The patent transforms the complex multi-physics modeling problem into a parameter estimation problem for the neural network. By changing the approach from solving differential equations for each submodel to learning parameter relationships directly from data, the system achieves high accuracy temperature estimation with significantly reduced computational time suitable for real-time control.
Data Source
AI summary
A control system operable to train a control tuner to generate temperature setpoint tracking improvements for a thermal processing system is provided. In one example implementation, temperature setpoint tracking improvements are achieved by generating system controller parameter adjustments based on a difference between a simulated workpiece temperature estimate and an actual workpiece temperature estimate. For example, a system model can generate a simulated workpiece temperature estimate simulating an actual workpiece temperature estimate, and based on the difference between the simulated and actual workpiece temperature estimates, generate clone controller parameter adjustments. The clone controller parameter adjustments can be used to generate system controller parameter adjustments, which can improve temperature setpoint tracking for the thermal processing system.