Substrate Processing Scheduling Using Neural Network Tool Pacing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current substrate processing systems face challenges in optimizing scheduling parameters for semiconductor substrate processing, leading to inefficiencies such as idle times and reduced throughput.
Innovation Solution
A system utilizing a processor and memory to simulate various processing scenarios and scheduling parameters, train a model using data from actual processing and simulation, and predict optimum scheduling parameters to minimize idle times and maximize throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional scheduling methods are used for substrate processing, then the system operation is simple, but idle times increase and throughput decreases
Solution Approach 1:
The system performs preliminary simulation of multiple processing scenarios before actual substrate processing to predict and determine optimal scheduling parameters in advance. This allows the system to pre-calculate ideal transfer timing, processing durations, and chamber allocation to minimize idle times and maximize throughput before the actual processing begins.
Solution Approach 2:
The system implements a feedback mechanism where actual processing data from the substrate processing tool is continuously collected and used to refine and retrain the machine learning model. This closed-loop feedback enables the system to learn from real-world operations and continuously improve scheduling predictions, thereby reducing idle times and enhancing throughput over time.
2Productivity
If optimal scheduling parameters are implemented, then throughput is maximized, but system complexity increases
Solution Approach 1:
The system introduces a machine learning model as an intermediary between the complex processing parameters and the scheduling decisions. This model acts as a smart mediator that automatically translates multiple input parameters (processing times, transfer durations, chamber availability) into optimal scheduling decisions, thereby managing system complexity while maximizing throughput without requiring manual optimization of each parameter.
Solution Approach 2:
The system creates a virtual copy or simulation environment that mirrors the actual substrate processing tool. This digital twin allows for scenario simulation and model training without affecting real production, enabling complex optimization calculations to be performed in the virtual copy while the actual system operates efficiently based on the learned schedules.
3Measurement precision
If simulation of multiple processing scenarios is performed, then scheduling accuracy is improved, but computational time increases
Solution Approach 1:
The system performs comprehensive simulation and model training in advance, before actual production needs the scheduling decisions. By pre-simulating multiple processing scenarios and training the machine learning model beforehand, the system accumulates knowledge that enables rapid, accurate scheduling decisions during actual operations without requiring time-consuming real-time simulations.
Data Source
AI summary
For etching tools, a neural network model is trained to predict optimum scheduling parameter values. The model is trained using data collected from preventive maintenance operations, recipe times, and wafer-less auto clean times as inputs. The model is used to capture underlying relationships between scheduling parameter values and various wafer processing scenarios to make predictions. Additionally, in tools used for multiple parallel material deposition processes, a nested neural network based model is trained using machine learning. The model is initially designed and trained offline using simulated data and then trained online using real tool data for predicting wafer routing path and scheduling. The model improves accuracy of scheduler pacing and achieves highest tool/fleet utilization, shortest wait times, and fastest throughput.