Machine learning-based model predictive control of hybrid dynamical systems

This article presents a machine learning-based model predictive control (MPC) scheme for stabilization of hybrid dynamical systems, for which the evolution of states exhibits both continuous and discrete dynamics described by differential and difference equations, respectively. We first present the development of two recurrent neural networks (RNNs) for approximating continuous- and discrete-time dynamics of hybrid dynamical systems, respectively, and then construct a unified hybrid RNN by integrating the two RNN models to capture both continuous and discrete dynamics. The hybrid RNN is used as the prediction model in Lyapunov-based MPC (RNN-LMPC), under which closed-loop stability of hybrid dynamical systems is established. Finally, two case studies including a bouncing ball example and a chemical process are utilized to illustrate the open- and closed-loop performance of the proposed RNN-LMPC scheme.     

An Inverse QSAR Method Based on a Two-Layered Model and Integer Programming

A novel framework for inverse quantitative structure–activity relationships (inverse QSAR) has recently been proposed and developed using both artificial neural networks and mixed integer linear programming. However, classes of chemical graphs treated by the framework are limited. In order to deal with an arbitrary graph in the framework, we introduce a new model, called a two-layered model, and develop a corresponding method. In this model, each chemical graph is regarded as two parts: the exterior and the interior. The exterior consists of maximal acyclic induced subgraphs with bounded height, the interior is the connected subgraph obtained by ignoring the exterior, and the feature vector consists of the frequency of adjacent atom pairs in the interior and the frequency of chemical acyclic graphs in the exterior. Our method is more flexible than the existing method in the sense that any type of graphs can be inferred. We compared the proposed method with an existing method using several data sets obtained from PubChem database. The new method could infer more general chemical graphs with up to 50 non-hydrogen atoms. The proposed inverse QSAR method can be applied to the inference of more general chemical graphs than before.     

Graph machine learning for design of high-octane fuels

Fuels with high-knock resistance enable modern spark-ignition engines to achieve high efficiency and thus low CO₂ emissions. Identification of molecules with desired autoignition properties indicated by a high research octane number and a high octane sensitivity is therefore of great practical relevance and can be supported by computer-aided molecular design (CAMD). Recent developments in the field of graph machine learning (graph-ML) provide novel, promising tools for CAMD. We propose a modular graph-ML CAMD framework that integrates generative graph-ML models with graph neural networks and optimization, enabling the design of molecules with desired ignition properties in a continuous molecular space. In particular, we explore the potential of Bayesian optimization and genetic algorithms in combination with generative graph-ML models. The graph-ML CAMD framework successfully identifies well-established high-octane components. It also suggests new candidates, one of which we experimentally investigate and use to illustrate the need for further autoignition training data.     

Automated synthesis of steady-state continuous processes using reinforcement learning

Automated flowsheet synthesis is an important field in computer-aided process engineering.The present work demonstrates how reinforcement learning can be used for automated flowsheet synthesis without any heuristics or prior knowledge of conceptual design.The environment consists of a steady-state flowsheet simulator that contains all physical knowledge.An agent is trained to take discrete actions and sequentially build up flowsheets that solve a given process problem.A novel method named SynGameZero is developed to ensure good exploration schemes in the complex problem.Therein,flowsheet synthesis is modelled as a game of two competing players.The agent plays this game against itself during training and consists of an artificial neural network and a tree search for forward planning.The method is applied successfully to a reaction-distillation process in a quaternary system.     

从分子筛上纳尺度离散行为控制到宏观煤化工过程

化学工业基于对化学反应和化学键的操纵,在横跨超12个数量级的空间和时间尺度上解决传递问题,为人类社会创造和生产新物质。石油的发现和利用将化学工业带入了一个全新的时代,支撑了现代社会的运行,但对中国而言,“富煤贫油少气”的资源特点决定了发展煤化工在中国有独特的战略意义。与石化行业常见的烃类裂化过程不同,现代新型煤化工依赖于小分子原料的自组装,依靠择形分子筛的控制作用高选择性制备燃料和化学品。因此,在亚纳米尺度下,即小分子在择形分子筛上的反应和传递行为不能再被视为连续,而是离散的,这导致了宏观上的拟相变失活等一系列现象。因此,需建立新的方法论理解分子筛上的离散反应和传递行为,用以认识离散尺度在十几个数量级空间和时间尺度延展后对宏观化学过程的影响。本文以纳尺度下分子筛内的离散传递现象为基础,综述了运用图论、小世界网络和先进表征手段分析分子筛内失活与传递现象的研究及发现的一系列新现象,并介绍了基于此进行的分子筛原子级精准结构调变及其对宏观煤化工过程带来的革新。面向未来,本文建立的分析分子筛内离散行为的范式将对精确调变分子筛结构、开发下一代煤化工工艺流程提供全新的思路。     

From nanoscale discrete diffusion behavior control to macroscale coal chemical process

Based on the manipulation of chemical reactions and chemical bonds, the chemical industry solves transport problems on a spatial and temporal scale that exceeds 12 orders of magnitude, creating and producing new substances for human society. The discovery and use of crude oil have brought the chemical industry into a new era and supported the operation of modern society. However, for China, the resource endowment determines that the development of the coal chemical industry has unique strategic significance. Unlike the hydrocarbon cracking process, the modern new coal chemical industry relies on the self-assembly of small molecular from raw materials and the shape-selective effect of zeolites to produce fuels and chemicals with high selectivity. Therefore, at the sub-nanometer scale, that is, the reaction and transfer behavior of small molecules on shape-selective zeolites can no longer be regarded as continuous, but discrete, which leads to a series of phenomena such as deactivation with pseudo-phase transitions on the macroscale. Based on the discrete transfer phenomena in zeolites at the nanoscale, this paper reviews the research and discovery of a series of new phenomena using graph theory, small-world networks, and advanced characterization methods to analyze the deactivation and transfer phenomena. The atomic-level precise structural modification of the zeolite and the innovation it brings to the macro coal chemical process is introduced. Towards the future, the paradigm for analyzing the discrete behavior in zeolites established in this article will provide a new way of thinking for precisely adjusting the zeolite structure and developing the next generation coal chemical process.     

人工神经网络在化工过程中的应用进展

人工神经网络作为人工智能的重要组成部分,以其超强的鲁棒性、容错性、可充分逼近任何复杂的非线性关系、并行处理、可学习和自适应等优点在改善化工过程传统生产技术诊断滞后、难以优化控制、物性估算误差较大以及不能处理非线性复杂情况等问题上有着广阔的发展空间。本文概述了人工神经网络的原理和发展历程,综述并分析了人工神经网络在故障诊断、过程控制和优化、质量控制、定量结构-活性/性质相关性分析、物性估算、专家系统以及聚类分析等化工过程中的应用原理以及研究进展和现状。最后指出卷积神经网络等深度学习算法的性能高、速度快,在化工过程中发展和应用深度学习算法将成为其发展方向和研究热点。     

神经网络遗传算法及其在乙烯裂解过程预测中的应用研究

针对复杂神经网络模型的结构特性和ＢＰ网络梯度下降法的缺点,将遗传算法与神经网络相结合,提出了基于改进连续化遗传算法的神经网络学方法,并将其应用于大型乙烯裂解产品分布预测,获得了较好的过于模拟结果。     

基于学习速率自适应的脉冲神经网络快速学习方(英文)

For accelerating the supervised learning by the SpikeProp algorithm with the temporal coding paradigm in spiking neural networks (SNNs), three learning rate adaptation methods (heuristic rule, delta-delta rule, and delta-bar-delta rule), which are used to speed up training in artificial neural networks, are used to develop the training algorithms for feedforward SNN. The performance of these algorithms is investigated by four experiments: classical XOR (exclusive or) problem, Iris dataset, fault diagnosis in the Tennessee Eastman process, and Poisson trains of discrete spikes. The results demonstrate that all the three learning rate adaptation methods are able to speed up convergence of SNN compared with the original SpikeProp algorithm. Furthermore, if the adaptive learning rate is used in combination with the momentum term, the two modifications will balance each other in a beneficial way to accomplish rapid and steady convergence. In the three learning rate adaptation methods, delta-bar-delta rule performs the best. The delta-bar-delta method with momentum has the fastest convergence rate, the greatest stability of training process, and the maximum accuracy of network learning. The proposed algorithms in this paper are simple and efficient, and consequently valuable for practical applications of SNN.     

Automatic design of neural network structures

Artificial neural networks are being extensively applied in many fields of science and engineering. Despite their wide range of applications and their flexibility, there is still no general framework or procedure through which the appropriate neural network for a specific task can be designed. The design of neural networks is still very dependent upon the designer's experience. This is an obvious barrier to the wider application of neural networks. To mitigate this barrier methods have been developed to automate the design of neural networks. A new method for the auto-design of neural networks was developed, which is based on genetic algorithms (GA) and Lindenmayer Systems. The method is less computationally intensive than existing iterative design procedures, hence it can be applied to the automatic design of neural networks for complex processes. To evaluate the performance of the new design procedure, it was tested for the design of industry standard neural networks. The method was also applied to design neural networks to model the dynamics of a pH neutralization process and a CSTR reactor in which a set of nonlinear reactions takes place. The networks obtained by the new algorithm for these typical chemical processes was much simpler, yet more accurate than those designed by traditional methods.     

R－X图解法计算多级串联理想混合反应器

论述Ｒ－Ｘ图解法计算多级串联理想混合反应器的基本原理，并结合实例，对Ｒ－Ｘ法的计算步骤予以详细介绍，为多级串联反应器的设计与计算，提供一种新颖适用的图解计算方法。     

R－X图解法计算多级串联理想混合反应器

论述Ｒ－Ｘ图解法计算多级串联理想混合反应器的基本原理，并结合实例，对Ｒ－Ｘ法的计算步骤予以详细介绍，为多级串联反应器的设计与计算，提供一种新颖适用的图解计算方法。     

Extractive Distillation Systems Involving Complex Columns with Partially Coupled Heat and Material Flows

A method based on the representation of flowsheets as graphs is proposed to synthesize flowsheets of extractive distillation of multicomponent azeotropic mixtures in complex columns with partially coupled heat and material flows. It is shown that the flowsheets constructed can provide a significant decrease in the energy consumption for separation because the process becomes structurally closer to thermodynamically reversible distillation.     

Pattern recognition in chemical process flowsheets

By recognizing similarities in flowsheets, engineers can understand ways in which to improve the design and efficiency of chemical processes. However, there is no prior literature on how to compare flowsheets and mine them for common patterns. To fill this gap, we propose the first systematic methodology to mine patterns in chemical process flowsheets. The proposed methodology consists of three major steps, each of which has a polynomial time complexity. We apply our methodology to several case studies that involve comparisons of up to 18 different flowsheets. The patterns identified by our methodology are consistent with engineering practice and heuristic rules in the process synthesis literature. © 2018 American Institute of Chemical Engineers AIChE J, 65: 592–603, 2019     

Application of Graphs of Distillation Trajectories to Synthesis of Separation Flowsheets

A graph method and a matrix method are proposed for synthesizing distillation flowsheets by analyzing process trajectories.     

定性/定量集成优化控制器的形式化设计方法

在化工过程中广泛存在着混杂系统的特性,即系统的信息结构由系统事件和系统状态构成。本文以时段演算为工具提出面向混合信息结构的混杂系统在定性/定量双重指标约束下的形式化设计方法,它将时段演算应用于混杂系统的需求刻划,得出控制系统的定性/定量行为和语义的定义,给出控制系统的行为描述形式化、优化描述形式化和基于时段演算的形式化设计步骤。最后结合一个应用实例,说明设计方法的有效性。     

Development of a heuristic methodology for precise sensor network design

This paper develops a heuristic methodology for designing precise sensor/measurement networks for linear material flow circuits based on the general principle of variance reduction through data reconciliation. Firstly, the paper develops a new ‘flowsheet-independent’ formula for estimating the adjusted variance of process streams on linear material flow circuits. Here, the weighted least squares problem is symbolically optimised to obtain a generalized ‘flowsheet-independent’ formula, which depends only on the measured variances of process streams. Secondly, this formula is used to develop a set of generic design principles/heuristics for maximising the variance reduction of process streams. These design principles/heuristics can be used for making a priori design decisions in selecting optimal flowsheets and/or measurement schemes for reducing the variance of process streams on linear material flow circuits. A precise knowledge of material flow rates is critical to areas such as process accounting and quality control which is important to, for example, large process industries such as the petrochemical and mining industries. Process streams are divided into two categories: (i) terminal and (ii) interacting (interconnecting) streams. The design principles/heuristics illustrate the importance of precise measurement of interacting streams and the benefits of highly interconnected flowsheets and/or measurement schemes in maximising variance reduction. The generalized formula and the design principles/heuristics are tested numerically using an industrial case study.     

Sensor network design for contaminant detection and identification in water distribution networks

Water distribution networks (WDN) are vulnerable to either intentional or accidental contamination. In order to protect against such intrusions, effective and efficient online monitoring systems are needed. Due to cost and maintenance reasons, it is not possible to locate sensors at each and every potential intrusion point. In this work, we design minimal sensor networks which satisfy the two important properties of observability (ability to detect an intrusion) and identifiability (ability to identify the point of intrusion). Based on the hydraulic analysis of the network, a bipartite graph is constructed between intrusion points and the corresponding nodes that can potentially be affected by the contaminant. The problem of sensor network design is converted to a minimum set cover problem on the bipartite graph, and is solved using a greedy heuristic algorithm. The proposed method is illustrated using a medium scale urban WDN.     

Identification of mixing parameters in agitated pulp chests: A continuous time domain approach

Parameter identification is quite challenging in mixing, which is extensively employed in chemical process industry. Agitated pulp chests are more difficult to characterize because they handle non-Newtonian pulp suspensions and non-ideal flows such as short circuiting, recirculation and channeling. In the present study, we characterize the agitated pulp chests in the continuous time domain, which obviates the restrictions imposed by the discrete time approach. For this purpose, a robust and efficient hybrid genetic algorithm is utilized along with a differential-algebraic model of mixing. Necessary derivatives including auxiliary differential equations are obtained for gradient search. Using realistic large sets of mixing data, both the algorithm and the model are successfully validated before characterizing laboratory-scale agitated pulp chests. Superior characterizations are obtained compared to those yielded by the discrete time domain approach. This outcome highlights the benefit of the continuous time domain approach developed in this work.