Polylactic acid (PLA) is an attractive environment‐friendly thermoplastic that is bio‐sourced and biodegradable. PLA is industrially produced by the ring‐opening polymerization of lactide. This reaction is sensitive to drifts in the operating conditions and impurities in the raw materials that may affect the reaction rate as well as the polymer properties, which can be very costly in continuous processes. It is therefore crucial to employ a control strategy that allows recovering the nominal conditions and maintaining the desired properties and conversion level in case of drift. Three control strategies are discussed in this paper: proportional‐integral (PI) controller, dynamic optimization, and model predictive control (MPC). The proposed approaches are validated by simulation of a continuous PLA process constituted of three cascade reactors including one loop reactor in the middle. Besides the coupling of inputs and outputs, the process model is highly nonlinear, and the control is done only on the boundaries. The results show that the open‐loop optimization strategy provides better performance compared to the PI controller if the disturbance is assumed to be measured. The MPC also shows superior performances provided that the disturbance is first estimated. A polynomial model is developed to predict the nonmeasured disturbance based on the measured outputs.

中文翻译：

---

连续丙交酯开环聚合过程的模型预测控制

聚乳酸（PLA）是一种有吸引力的环保热塑性材料，可生物来源且可生物降解。PLA是由丙交酯的开环聚合反应工业生产的。该反应对操作条件的漂移和原料中的杂质敏感，这些杂质可能影响反应速度以及聚合物性能，这在连续过程中可能会非常昂贵。因此，至关重要的是采用一种控制策略，该策略允许恢复标称条件并在发生漂移的情况下保持所需的特性和转化水平。本文讨论了三种控制策略：比例积分（PI）控制器，动态优化和模型预测控制（MPC）。通过模拟由三个级联反应器（中间包括一个环流反应器）组成的连续PLA工艺，验证了所提出的方法。除了输入和输出的耦合之外，过程模型是高度非线性的，并且仅在边界上进行控制。结果表明，如果假定要测量干扰，则与PI控制器相比，开环优化策略可提供更好的性能。如果首先估计干扰，MPC还将显示出优异的性能。开发了多项式模型以基于测量的输出预测未测量的干扰。结果表明，如果假定要测量干扰，则与PI控制器相比，开环优化策略可提供更好的性能。如果首先估计干扰，MPC还将显示出优异的性能。开发了多项式模型以基于测量的输出预测未测量的干扰。结果表明，如果假定要测量干扰，则与PI控制器相比，开环优化策略可提供更好的性能。如果首先估计干扰，MPC还将显示出优异的性能。开发了多项式模型以基于测量的输出预测未测量的干扰。