An integrated approach for model building, maintenance, reduction and updating in complex dynamic chemical systems

过程系统工程涉及动态模型的开发和更新，以及使用这些模型进行监测、控制和优化。复杂的系统对每个目标的标准方法的应用提出了许多挑战。复杂性本身来自许多来源，包括过程的多尺度或分布式性质、它们的网络结构、系统中存在的群体或单元集合、或系统中的绝对数量的变量。复杂系统提出的具体问题有三种：对如此大小和规模的模型应用标准分析技术(用于模型更新、估计、监测、优化和控制)的计算和数值问题；对物理(和/或化学)的不完全理解导致首先难以构建基本模型；以及从传感器获得和解释此类系统的有意义的数据的困难。 过去，我们的研究在不同的应用领域解决了这些挑战的例子，包括开发适用于大系统的模型更新和估计方法，开发复杂化学系统的混合伪动力学模型，以及开发油藏建模和预测中的替代建模和不确定性传播技术。在这份提案中概述的研究计划中，我们迈出了第一步，以开发一个通用的研究框架，以解决为系统建立混合模型或灰箱模型的挑战，在这些系统中，我们没有不完整的知识来建立基本模型，用于融合来自多个传感器的信息并使它们与所开发的模型相一致，以及开发模型降阶方法以降低复杂系统中系统特性分析的难度。 拟议计划所产生的预期贡献有可能显著推动复杂系统过程系统工程技术发展的最先进水平；此外，这些技术将在我的研究中调查的各种应用领域找到适用性，包括泡沫浮选、煤和石油储量建模和估计、催化剂和催化过程的优化和控制、沥青的热升级、利用微藻固定二氧化碳和生产生物燃料，以及在线光谱数据的化学计量学解释。

Process systems engineering is concerned with the development and updating of dynamic models, and the use of these models for monitoring, control and optimization. Complex systems present many challenges to the application of standard methods for each of objectives. Complexity itself arises from many sources, including the multiscale or distributed nature of processes, their networked structure, the presence of populations or ensembles of units in the system, or the sheer number of variables in the system. The specific problems complex systems pose are of three kinds: computational and numerical issues in applying standard techniques of analysis (for model updating, estimation, monitoring, optimization and control) to models of such size and scale, incomplete understanding of the physics (and/or chemistry) leading to difficulty in constructing fundamental models in the first place, and the difficulty in obtaining and interpreting meaningful data from sensors for such systems. In the past, our research has addressed examples of these challenges in various application areas, including developing methods for model updating and estimation suitable for large-scale systems, the development of hybrid pseudo-kinetic models for complex chemical systems, and the development of techniques for surrogate modeling and uncertainty propagation in reservoir modeling and prediction. In the research program outlined in this proposal, we take the first steps towards developing a general research framework for addressing the challenges of building hybrid or gray-box models for systems where we have incomplete knowledge to build a fundamental model, for fusing information from multiple sensors and reconciling them with the developed models, and the development of model reduction methods for reducing the difficulty of analysis of system characteristics in complex systems. The expected contributions arising from the proposed program have the potential to significantly advance the state of the art in development of process systems engineering techniques for complex systems; in addition, these techniques will find applicability in a wide variety of application areas under investigation in my research, including froth flotation, coal and petroleum reservoir modeling and estimation, optimization and control of catalysts and catalytic processes, thermal upgrading of bitumen, carbon dioxide fixation and biofuel production using microalgae, and chemometric interpretation of on-line spectroscopic data.