Scholarlink

<p>Achieving consistent product purity and high recovery rates is a&#xa0;critical operational hurdle for the recycling industry. This challenge stems directly from the inherent heterogeneity of incoming waste streams, which causes high variability in process performance and undermines stable operations. This paper introduces the “Smart Twin”, a&#xa0;probabilistic digital twin framework designed to manage this uncertainty. Unlike deterministic models that provide single-point predictions, the Smart Twin leverages a&#xa0;Bayesian statistical approach to generate a&#xa0;full distribution of likely outcomes. This provides operators with a&#xa0;direct measure of confidence for their predictions, enabling more robust, risk-informed decision-making. We present the framework’s high-level concept and demonstrate its application through a&#xa0;detailed use case in a&#xa0;multi-stage copper recovery process. By enabling operators to simulate the impact of operational adjustments on final product quality distributions, the Smart Twin acts as a&#xa0;powerful decision-support tool for quality control, process optimization, and root cause analysis. The framework offers a&#xa0;practical pathway for implementing AI-driven control to enhance resource efficiency and economic viability in the broader recycling and secondary materials industry.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Smart Twin: Ein Bayessches Framework für die probabilistische Modellierung von mehrstufigen industriellen Prozessen

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Sabrina Meindl,
  • Du Nguyen Duy,
  • Valeria Fonseca Diaz,
  • Roman Rainer,
  • Alexia Tischberger-Aldrian

摘要

Achieving consistent product purity and high recovery rates is a critical operational hurdle for the recycling industry. This challenge stems directly from the inherent heterogeneity of incoming waste streams, which causes high variability in process performance and undermines stable operations. This paper introduces the “Smart Twin”, a probabilistic digital twin framework designed to manage this uncertainty. Unlike deterministic models that provide single-point predictions, the Smart Twin leverages a Bayesian statistical approach to generate a full distribution of likely outcomes. This provides operators with a direct measure of confidence for their predictions, enabling more robust, risk-informed decision-making. We present the framework’s high-level concept and demonstrate its application through a detailed use case in a multi-stage copper recovery process. By enabling operators to simulate the impact of operational adjustments on final product quality distributions, the Smart Twin acts as a powerful decision-support tool for quality control, process optimization, and root cause analysis. The framework offers a practical pathway for implementing AI-driven control to enhance resource efficiency and economic viability in the broader recycling and secondary materials industry.