The optimization of batch processes usually relies on the availability of a detailed knowledge-driven model. However, because of the great varieties of industrial batch processes and their small production rates, a knowledge-driven model might not always be available. In such a case, a data-driven model, developed after a limited number of experiments, is an attractive alternative. Here we apply, in an evolutionary manner, the design of dynamic experiments (DoDE) (Georgakis et al. Ind. Eng. Chem. Res. 2013, 52 (35), 12369) methodology to model the process behavior and minimize the batch cycle time of an industrial polymerization process. In evolutionary DoDE, the initial design is selected conservatively in the close vicinity of the previous operating conditions to minimize the risk of violating safety constraints of the industrial process. After the initial data-driven model has been estimated using the collected data, an optimal operating condition satisfying process constraints is calculated. In addition, the input domain is enlarged to seek conditions that further optimize the process. The above steps are iterated until the most optimal process performance is achieved. We examine this evolutionary DoDE approach in silico using a detailed simulation of a working polymerization process at Dow to produce that data. After three rounds of experiments are performed, a 17.2% reduction in batch cycle time is achieved while all constrains on safety and product quality are met. It is only 0.7% longer than the batch cycle time obtained using model-based optimization, assuming a 100% accurate model is available.

中文翻译：

---

动态实验方法设计的工业间歇聚合过程数据驱动优化

批处理过程的优化通常取决于详细的知识驱动模型的可用性。但是，由于工业批处理的种类繁多且生产率低，因此知识驱动模型可能并不总是可用。在这种情况下，经过有限次实验开发的数据驱动模型是一种有吸引力的选择。在这里，我们应用，在一个渐进的方式，动态实验设计（栗）（Georgakis等人工业主机。化学式RES。2013，52（35），12369）的方法来模拟过程行为并最小化工业聚合过程的批处理周期时间。在进化式DoDE中，在接近先前的操作条件的情况下保守地选择初始设计，以最大程度地减少违反工业过程安全约束的风险。在使用收集的数据估算了初始数据驱动模型之后，便会计算出满足过程约束的最佳运行条件。另外，输入域被扩大以寻找进一步优化过程的条件。重复执行上述步骤，直到获得最佳的过程性能。我们使用在陶氏生产聚合数据的工作聚合过程的详细模拟，对这种进化的DoDE方法进行了计算机分析。经过三轮实验，批次周期时间减少了17.2％，同时满足了安全性和产品质量的所有约束。假设可以使用100％准确的模型，它仅比使用基于模型的优化所获得的批处理周期时间长0.7％。