Scholarlink

<p>Sheet Molding Compound (SMC) compression molding enables the cost-efficient manufacturing of lightweight structural components for automotive applications. However, the final part quality is highly sensitive to complex interactions between various manufacturing parameter that are difficult to optimize through trial-and-error. Physics-based Finite Element simulations support process understanding and parameter optimization, but their computational cost and modeling assumptions limit their applicability during production. This work introduces a co-deployed decision-support framework combining a Reduced Order Model (ROM) derived from physics-based simulations with complementary Data-Driven Models (DDMs) trained on real manufacturing measurements. The ROM provides instantaneous predictions of internal process fields, while the DDMs infer quality indicators not reliably obtainable from simulation alone. Both models are integrated into the manufacturing machine’s Process Control Manager and connected to the press via OPC-UA, enabling real-time insight for operators. The methodology is demonstrated on the industrial production of an automotive upper suspension control arm manufactured by MARELLI, yielding reduced design time, lower scrap rates, enhanced quality, and actionable feedback. These results highlight the potential of complementary digital-twin approaches for optimizing composite manufacturing processes.</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

Real-Time optimization of sheet molding compound production using complementary reduced-order and data-driven modeling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月11日

查看全文
  • Mikhael Tannous,
  • Konstantinos Tzimanis,
  • Iason Tzanetatos,
  • Jon Larreina,
  • Chady Ghnatios,
  • Salvatore Sottile,
  • Francisco Chinesta

摘要

Sheet Molding Compound (SMC) compression molding enables the cost-efficient manufacturing of lightweight structural components for automotive applications. However, the final part quality is highly sensitive to complex interactions between various manufacturing parameter that are difficult to optimize through trial-and-error. Physics-based Finite Element simulations support process understanding and parameter optimization, but their computational cost and modeling assumptions limit their applicability during production. This work introduces a co-deployed decision-support framework combining a Reduced Order Model (ROM) derived from physics-based simulations with complementary Data-Driven Models (DDMs) trained on real manufacturing measurements. The ROM provides instantaneous predictions of internal process fields, while the DDMs infer quality indicators not reliably obtainable from simulation alone. Both models are integrated into the manufacturing machine’s Process Control Manager and connected to the press via OPC-UA, enabling real-time insight for operators. The methodology is demonstrated on the industrial production of an automotive upper suspension control arm manufactured by MARELLI, yielding reduced design time, lower scrap rates, enhanced quality, and actionable feedback. These results highlight the potential of complementary digital-twin approaches for optimizing composite manufacturing processes.