Model Predictive Control for Zone Melting Stability

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional PID controllers in the zone melting process for producing high-purity single-crystalline silicon face limitations when dealing with new or different system components, particularly in growing larger crystals and automating other process phases, leading to suboptimal control and stability issues.

Innovation Solution

The implementation of model predictive control (MPC) that uses a mathematical model to predict the future behavior of the zone melting process, optimizing manipulated variables such as generator power, stock rod speed, and crystal speed to minimize a cost function while adhering to system and manipulated variable limits, ensuring precise control and stability across various growth phases.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional PID controllers are used for zone melting process control, then good results are achieved under same process conditions, but the controllers reach operational limits when using new components, growing larger crystals, or automating other process phases

Engineering Contradiction:
Improveadaptability to new components and process conditionsVSAvoidcontrol stability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent transitions from conventional PID controller parameters to model predictive control parameters based on a mathematical process model. This allows the control system to adapt to new components and larger crystal sizes by using a fundamental model rather than empirically tuned parameters, resolving the contradiction between adaptability and reliability

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the empirical PID control mechanism with a model-based predictive control mechanism. This substitution enables the system to handle new process conditions and larger crystals through physics-based predictions rather than fixed empirical rules, improving both adaptability and reliability

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If model predictive control is implemented to improve adaptability and automation, then control precision and stability are enhanced, but the system complexity increases

Engineering Contradiction:
Improvecontrol precisionVSAvoidcontrol system complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent introduces a mathematical process model as an intermediary between the control inputs and the actual physical process. This model serves as a virtual representation that predicts system behavior, enabling precise control without directly increasing physical system complexity. The model acts as a mediator that translates control objectives into actionable manipulated variables

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a virtual copy of the physical zone melting process through a mathematical model. This digital twin or virtual model replicates the essential dynamics of the physical system, allowing control decisions to be made in the virtual domain and then applied to the physical system, thereby achieving high precision without proportionally increasing physical complexity

Inventive Principle:
Principle #26Copying

3Stability of the object's composition

If the prediction horizon is increased to improve system stability, then better control behavior is achieved, but the computational time and complexity increase

Engineering Contradiction:
Improvesystem stabilityVSAvoidcomputational time
Core Design Contradiction:
Stability of the object's compositionVSLoss of time

Solution Approach 1:

The patent applies partial action by using a finite prediction horizon rather than an infinite one. The control strategy predicts system behavior over a limited time window (prediction horizon) and applies control adjustments based on this partial future view. This approach achieves sufficient stability improvement without the computational burden of predicting the entire future trajectory, balancing stability gains with computational efficiency

Inventive Principle:
Principle #16Partial or excessive action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enables the production of high-purity single-crystalline silicon with improved axial shape control and stability, allowing for automation of the entire crystal growth process, reducing operator intervention, and enhancing productivity by maintaining defined growth conditions throughout all phases.

Implementation Method 1

uses induction heating to melt and maintain the melting zone under a protective gas atmosphere

Methodology Applied
Scientific EffectInduction heating: Induction Heating

Implementation Method 2

The contactless heating source meets the highest purity requirements

Methodology Applied
Scientific EffectElectromagnetic induction: Electromagnetic Induction

Implementation Method 3

the melt crystallizes into a dislocation-free single crystal

Methodology Applied
Scientific EffectCrystallisation: Crystallisation

Implementation Method 4

From a certain distance below the induction coil, the melt crystallizes into a dislocation-free single crystal

Methodology Applied
Scientific EffectSolidification: Freezing

Data Source

PatentEP2890834B1Model predictive control of the zone-melting process
Publication Date: 2018.06.13 TOPSIL GLOBALWAFERS AS
  • EP2890834B1 patent drawingFigure 1
  • EP2890834B1 patent drawingFigure 2
  • EP2890834B1 patent drawingFigure 3

AI summary

The invention relates to a method for the controlled production of crystalline molded bodies in the zone-melting process by influencing manipulated variables. The nature of the influencing is determined on the basis of a comparison of measured actual values with target values and predicted course values of system states. The actual values of the system states are measured directly and/or are measured indirectly by means of a state observer. The method is characterized in that the predicted course values are calculated in a model-based manner on the basis of the directly or indirectly measured actual values, and that at least one manipulated variable, selected from generator power (PGen), stock rod speed (vF), and crystal speed (vC), is changed in such a way that the deviation from the target values and the predicted course values is reduced.