A data driven modeling and control tool

数据驱动的建模和控制工具

• 批准号: 548804-2020
• 负责人: Mhaskar, Prashant
• 金额: $ 9.11万
• 依托单位: McMaster University
• 依托单位国家: 加拿大
• 项目类别: Idea to Innovation
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=700938
• 关键词: data; driven; modeling; control; tool

The increased availability of sensor data and advanced computing tools has made some very lucrative opportunities for the manufacturing sector. In particular, the new large data sets we are now collecting can now be used to build accurate models for the process, which in turn can be utilized to dramatically improve on-spec production, and enable rapid new product design. This emerging technology is well suited for batch and batch like operation (specialty chemical production, pharmaceuticals or virtually in every continuous operation in startup mode) which are by definition transient in nature, and witness significant variation over a process cycle. Most of the existing commercial software focuses on operation at a steady state. Limited tools exist for modelling and controlling batch operation, but the existing tools cannot optimally control batches of varying duration to, for instance, minimize production time. There is new work in the area of machine learning/artificial intelligence based models, but those are currently limited to classification kind of problems, and not at the commercialization stage for a full-scale industrial plant. Recent results from the PI's, Prashant Mhaskar and Michael Thompson's group, have demonstrated the capability of a subspace model based data driven modelling and control approach to handle a myriad of applications, ranging from semi-continuous distillation to steel refining to hydrogen production startup to polymer processing to wastewater treatment. The present proposal aims to commercialize this framework. Hydromantis Inc. (a wastewater treatment plant models), Praxair Inc- Now Linde Inc., Imperial Oil and Rescraft Plastic Products Inc. intend to increase their competitiveness by updating their facilities with more intelligent systems and have expressed a desire to evaluate our approach once a working prototype is available. Easy access to other industrial members is readily enabled, through the PI's involvement in the McMaster Advanced Control Consortium (an academy industry consortium with 7 annual membership paying members) and McMaster Manufacturing Research Institute which serves the community offering contract and research services.

传感器数据和先进计算工具可用性的增加为制造业带来了一些非常有利可图的机会。特别是，我们现在收集的新的大型数据集现在可以用于为该过程构建准确的模型，从而可以用于显着提高按规格生产，并实现快速的新产品设计。这一新兴技术非常适合于间歇式和间歇式操作（特种化学品生产、制药或几乎在启动模式下的每个连续操作），这些操作本质上是瞬时的，并且在工艺周期内发生显著变化。大多数现有的商业软件侧重于稳态运行。有限的工具存在于建模和控制批量操作，但现有的工具不能最佳地控制不同持续时间的批次，例如，最小化生产时间。在基于机器学习/人工智能的模型领域有新的工作，但这些工作目前仅限于分类问题，而不是在商业化阶段的全面工业工厂。 来自PI、Prashant Mhaskar和Michael Thompson小组的最新结果已经证明了基于子空间模型的数据驱动建模和控制方法处理无数应用的能力，从半连续蒸馏到钢铁精炼到制氢启动到聚合物加工到废水处理。本提案旨在使这一框架商业化。Hydromantis公司（废水处理厂模型），普莱克斯公司-现在林德公司，Imperial Oil and Rescraft Plastic Products Inc. 打算通过使用更智能的系统更新其设施来提高其竞争力，并表示希望在获得工作原型后对我们的方法进行评估。通过PI参与McMaster Advanced Control Consortium（一个拥有7个年度会员付费成员的学术行业联盟）和McMaster Manufacturing Research Institute（为社区提供合同和研究服务），可以轻松访问其他工业成员。