Semi-supervised dynamic latent variable modeling: I/O probabilistic slow feature analysis approach

Modeling of high dimensional dynamic data is a challenging task. The high dimensionality problem in process data is usually accounted for using latent variable models. Probabilistic slow feature analysis (PSFA) is an example of such an approach that accounts for high dimensionality while simultaneously capturing the process dynamics. However, PSFA also suffers from a drawback that it cannot use output information when determining the latent slow features. To address this lacunae, extension of the PSFA by incorporating outputs, resulting in Input-Output PSFA (IOPSFA) is proposed. IOPSFA can use both input and output information for extracting latent variables. Hence, inferential models based on IOPSFA are expected to have better predictive ability. The efficacy of the proposed approach with an industrial and a laboratory scale soft sensing case studies that have both complete and incomplete output measurements is evaluated, respectively. © 2018 American Institute of Chemical Engineers AIChE J, 65: 964–979, 2019     

A new algorithm for developing dynamic radial basis function neural network models based on genetic algorithms

A new method for extracting valuable process information from input–output data is presented in this paper. The proposed methodology produces dynamical radial basis function (RBF) neural network models based on a specially designed genetic algorithm (GA), which is used to auto-configure the structure of the networks and obtain the model parameters. The new RBF network training technique formulates a complete optimization problem, which includes the network structure into the set of free variables that are used to minimize the prediction error. This is a different approach compared with the local search methods employed by other structure selection mechanisms, which are often trapped to local minima. Another advantage of the proposed method is that only one run of the algorithm is required to obtain the optimal network structure, in contrast to the standard RBF training techniques, where the produced model is selected by trial and error. The effectiveness of the method is illustrated through the development of dynamical models for two sets of data: simulated data from a Continuous Stirred Tank Reactor (CSTR) and true data collected from a Kamyr digester, which is a rather complicated reactor used in the pulp and paper industry.     

A particle filter approach to identification of nonlinear processes under missing observations

A novel maximum likelihood solution to the problem of identifying parameters of a nonlinear model under missing observations is presented. If the observations are missing, then it is difficult to build a partial likelihood function consisting of only the available observations. Hence, an expectation–maximization (EM) algorithm, which uses the expected value of the complete log‐likelihood function including the missing observations, is developed. The expected value of the complete log‐likelihood (E‐step) in the EM algorithm is approximated using particle filters and smoothers. New expressions for particle filters and smoothers under missing observations are derived. In order to reduce the variance on the smoothed states, a point‐wise (as opposed to path‐based) state estimation procedure is used. The maximization step (M‐step) in the EM algorithm is performed using standard optimization routines. The proposed nonlinear identification approach is illustrated through numerical examples.     

基于分片线性逼近的聚氯乙烯生产计划优化分解算法

聚氯乙烯全流程生产过程计划优化往往描述为复杂MINLP模型,求解难度非常大,为此引入分片线性技术逼近实际生产中的非线性特征,建立基于HH的MILP模型,进一步提出一种基于离线层级模型的分解算法来加速求解过程：第一层在对生产设备以最优能耗点进行层级划分得到离线层级模型的基础上优化一个等价MILP问题,确定表征设备操作状态的二值变量;第二层以HH模型为基础,在二值变量确定的情况下,代入计划优化模型调整设备的工作点,最终确定模型的最优操作决策方案。最后,以一个实际工厂规模的案例来验证模型和算法的有效性,结果表明本算法在基本不损失优化结果性能的前提下可以大大提高求解效率,缩短求解时间达99%以上。     

Optimizing inventory policies in process networks under uncertainty

We address the inventory planning problem in process networks under uncertainty through stochastic programming models. Inventory planning requires the formulation of multiperiod models to represent the time-varying conditions of industrial process, but multistage stochastic programming formulations are often too large to solve. We propose a policy-based approximation of the multistage stochastic model that avoids anticipativity by enforcing the same decision rule for all scenarios. The proposed formulation includes the logic that models inventory policies, and it is used to find the parameters that offer the best expected performance. We propose policies for inventory planning in process networks with arrangements of inventories in parallel and in series. We compare the inventory planning strategies obtained from the policy-based formulation and the analogous two-stage approximation of the multistage stochastic program. Sequential implementation of the planning strategies in receding horizon simulations shows the advantages of the policy-based model, despite the increase in computational complexity.     

Design of responsive supply chains under demand uncertainty

This paper addresses the optimization of supply chain design and planning under responsive criterion and economic criterion with the presence of demand uncertainty. The supply chain consists of multi-site processing facilities and corresponds to a multi-echelon production network with both dedicated and multiproduct plants. The economic criterion is measured in terms of net present value, while the criterion for responsiveness accounts for transportation times, residence times, cyclic schedules in multiproduct plants and inventory management. By using a probabilistic model for stock-out, the expected lead time is proposed as the quantitative measure of supply chain responsiveness. The probabilistic model can also predict the safety stock levels by integrating stock-out probability with demand uncertainty. These are all incorporated into a multi-period mixed-integer nonlinear programming (MINLP) model, which takes into account the selection of manufacturing sites and distribution centers, process technology, production levels, scheduling and inventory levels. The problem is formulated as a bi-criterion optimization model that maximizes the net present value and minimizes the expected lead time. The model is solved with the -constraint method and produces a Pareto-optimal curve that reveals how the optimal net present value, supply chain network structure and safety stock levels change with different values of the expected lead time. A hierarchical algorithm is also proposed based on the decoupling of different decision-making levels (strategic and operational) in the problem. The application of this model and the proposed algorithm are illustrated with two examples of polystyrene supply chains.     

A new approach for scheduling of multipurpose batch processes with unlimited intermediate storage policy

The increasing demand of goods, the high competitiveness in the global marketplace as well as the need to minimize the ecological footprint lead multipurpose batch process industries to seek ways to maximize their productivity with a simultaneous reduction of raw materials and utility consumption and efficient use of processing units. Optimal scheduling of their processes can lead facilities towards this direction. Although a great number of mathematical models have been developed for such scheduling, they may still lead to large model sizes and computational time. In this work, we develop two novel mathematical models using the unit-specific event-based modelling approach in which consumption and production tasks related to the same states are allowed to take place at the same event points. The computational results demonstrate that both proposed mathematical models reduce the number of event points required. The proposed unit-specific event-based model is the most efficient since it both requires a smaller number of event points and significantly less computational time in most cases especially for those examples which are computationally expensive from existing models.     

A new process variable and dynamics selection method based on a genetic algorithm‐based wavelength selection method

Soft sensors have been used in industrial plants to estimate process variables that are difficult to measure online. Soft sensor models predicting an objective variable should be constructed with only important explanatory variables in terms of predictive ability, better interpretation of models and lower measurement costs. Besides, some process variables can affect an objective variable with time‐delays. Therefore, we have proposed the methods for selecting important process variables and optimal time‐delays of each variable simultaneously, by modifying the genetic algorithm‐based wavelength selection method that is one of the wavelength selection methods in spectrum analysis. The proposed methods can select time‐regions of process variables as a unit by using process data that includes process variables that are delayed in the range from zero to a set/given maximum value. The case study with simulation data and real industrial data confirmed that predictive, easy‐to‐interpret, and appropriate models were constructed using the proposed methods. © 2012 American Institute of Chemical Engineers AIChE J, 58: 1829–1840, 2012     

A robust mathematical formulation for multipurpose batch plants

Several scheduling techniques exist in literature based on continuous time representation. The models based on unit specific time points have shown better solution efficiency by reducing the number of time points and problem size. In this paper novel scheduling techniques based on unit specific time point continuous time representation are presented. The proposed models allow nonsimultaneous material transfer into a unit. Nonsimultaneous transfer refers to when a task requires more than one intermediate state it is possible for one state to be transferred and stored in a unit that is processing it for a while and wait for the other intermediates to come together to start the task. This approach gives a better schedule as compared to most published models. The developed MILP scheduling models are based on state sequence network representation that has proven to inherently result in smaller problems in terms of binary variables. The models require a smaller number of time points as compared to single-grid and multi-grid continuous time models. Consequently, they exhibit much better computational performance. Numerical evaluation using literature examples indicate in some of the complex examples that the proposed models give a better objective value as compared to other scheduling models. An added feature of the proposed models is their ability to exactly handle fixed intermediate storage operational philosophy, which has proven to be a subtle drawback in most published scheduling techniques.     

Toward Optimal Operation Conditions of Freeze-Drying Processes via a Multilevel Approach

Freeze drying (lyophilization) offers an attractive dehydration method for valuable food and biological products, because it is capable of preserving product quality and biological activity while extending their shelf life. However, despite these benefits in terms of product quality, freeze drying is also a notoriously energy-intensive and time-consuming process. This requires an expensive operation to construct an efficient optimal decision-making tool able to drive the operation through the most effective paths that minimize time and maximize product quality. Here we propose an integrated approach to operational design and control of the freeze-drying process that combines dynamic modeling with efficient optimized off-line and on-line control. The required mass and energy balance equations still contain inherent nonlinearity, even in their lumped parameter version. This results in a set of complex dynamic, computationally costly optimization problems solved by selected global stochastic optimization algorithms. Real-time disturbances and model uncertainties are addressed via the proposed hierarchical multilevel approach, allowing recalculation of the required control strategies. The framework developed has been revealed as a useful tool to systematically define off-line and on-line optimal operation policies for many food and biological processing units.     

基于模糊PID的退火炉温度控制研究

退火炉的自动控制策略已成为过程控制的研究热点。由于退火炉是一个大惯性、纯滞后、参数时变的非线性控制对象,很难准确建立起炉子的精确数学模型,根据退火炉工业系统的特点,采用了模糊PID参数自调整控制的控制方法,将其应用于退火炉温度控制环节。本文利用Simulink对模糊神经网络PID控制算法进行仿真试验。试验显示,模糊神经网络控制算法能够展现强抗干扰性、强鲁棒性以及良好的控制性能,适用于对退火炉的温度控制中。     

Application of the Microsoft Excel Solver tool in the solution of optimization problems of heat exchanger network systems

Optimization of heat exchanger networks (HEN_S) is a field of interest in Chemical Engineering. The aim is to use the residual heat of process streams to raise the temperature of other process streams which are cold, thus minimizing the use of hot and cold utilities. The usual way to deal with this task is to pose a linear programming problem according to a transshipment model in order to minimize either the general services consumptions or the number of hot and cold stream matchings required. This optimization is subjected to several restrictions given by the energy balances corresponding to each heat exchange.The high number of variables and equations obliges the use of specific software to solve these problems, e.g. GAMS (General Algebraic Modelling System). However, this software is not intuitive and requires a long time of training and a certain level of practice for the students to adequately use it.This work describes the use of Microsoft Excel-Solver for the sequential solving of HEN_S optimization problems according to the transshipment model. Solver is a much simpler, intuitive and easily available tool for the students who, besides, usually learn how to use it in the first years of their degree, in Informatics or similar subjects. Thus, it is not necessary to spend extra time in the teaching of the software, time that may be devoted to other contents.This paper presents an example of HEN_S problem proposed in the subject Simulation and Optimization of Chemical Processes corresponding to the Joint Master in Chemical Engineering of the Rey Juan Carlos University and the Autónoma University of Madrid and solved by the students using Excel-Solver.     

Equation‐oriented flowsheet simulation and optimization using pseudo‐transient models

Tight integration through material and energy recycling is essential to the energy efficiency and economic viability of process and energy systems. Equation‐oriented (EO) steady‐state process simulation and optimization are key enablers in the optimal design of integrated processes. A new process modeling and simulation concept based on pseudo‐transient continuation is introduced. An algorithm for reformulating the steady‐state models of process unit operations as differential‐algebraic equation systems that are statically equivalent with the original model is presented. These pseudo‐transient models improve the convergence of EO process flowsheet simulations by expanding the convergence basin. This concept is used to build a library of pseudo‐transient models for common process unit operations, and this modeling concept seamlessly integrates with a previously developed time‐relaxation optimization algorithm. Two design case studies are presented to validate the proposed framework. © 2014 American Institute of Chemical Engineers AIChE J 60: 4104–4123, 2014     

耦合溴化锂吸收式制冷与有机朗肯循环的合成气深冷分离工艺

从合成气中深冷分离液化天然气(LNG)可以在调峰中发挥重要作用,并显著提升企业的经济效益。然而深冷分离的高能耗是实际工业中的一大问题。本文提出了耦合溴化锂吸收式制冷与有机朗肯循环的甲烷深冷分离工艺。新工艺可以利用原压缩制冷系统的余热从而降低制冷能耗。又因为压缩级数与能耗和可利用余热量成正相关,为使得系统的能耗最低,需同时优化压缩级数与所耦合的余热利用系统。采用自适应遗传算法对新工艺中8种不同压缩级数组合进行优化,通过对比各模型的总能耗、性能系数和单位能耗确定了能耗最低的流程。其结果表明,相比于原工艺总能耗减少了34%;性能系数增加了0.07;单位能耗减少了0.89kW/kg。经济表现为操作费用减少了33%;新增设备投资2550万元,理论上一年即可回收投资成本。     

Diffuse interface model of the freeze-drying process of individually frozen products

Abstract

Vacuum freeze-drying (VFD) is a dehydration method based on the sublimation of the liquid phase contained in a certain product, previously frozen, at low pressure and temperature. Since it is a time and energy consuming process, it is crucial to select the best processing conditions to minimize drying duration, thus reducing the energy requirement. Additionally, product temperature must be monitored since it plays an important role in preserving product quality. The aim of this study was to develop a Diffuse Interface Model (DIM) for in-silico simulation of the freeze-drying process of individually frozen products. Due to the geometrical features of the samples, and to the role of radiation in the heat transfer to the product, the usual one-dimensional approach is inappropriate. Using a DIM, each cell of the computational domain can be described as a porous solid matrix filled by ice and vapor with a time-varying composition, thus allowing the use of a fixed computational grid and making the computation effort less demanding in comparison to moving interface-based models. Drying of eggplant cubic samples was considered as case study: model parameters were estimated by fitting the experimentally measured product temperature and drying time to the calculated ones. The model was proven to be reliable in providing an accurate estimate of both the drying time and the product temperature. Therefore, it can be used for off-line process design and optimization, minimizing the experimental effort required to design and optimize the process.     

一种新型的基于Levenshtein距离层次聚类的时序操作优化方法

现代流程工业过程中，DCS采集并存储了大量的操作时序数据，若能将其中有价值的操作经验和操作信息提取出来，则可大大提高操作系统的性能。然而，操作经验概念较为模糊，无法具体量化。因此，将具有时序特征的操作数据符号化，使操作经验以区块化形式表示，并提出一种基于Levenshtein距离的时序层次凝聚聚类算法，通过对操纵变量的历史时序操作数据进行相似性搜索，进而获得多种相似的操作模式，并将每种类型的操作模式对应的过程变量进行性能分析，从而得到并保存实际工作过程中所需的操作经验，以达到生产过程操作优化的目的。为了验证所提出方法，将其用于连续组分精馏操作过程，实验结果表明所提出的基于Levenshtein距离层次聚类的操作优化方法的有效性。     

Multiperiod Investment Models for the Gradual Reconstruction of Chemical Processes

A multiperiod mixed‐integer nonlinear programming (MINLP) model is presented for the planning of an optimal investment policy for the gradual retrofit of chemical plants over a specified time frame. This approach can be applied when the available funds are insufficient to perform the optimal reconstruction in one step. Through gradual improvements, the process progressively approaches its optimum with considerably lower requirements for new capital, while still exhibiting an attractive increase in economic performance. In the proposed multiperiod problem for gradual retrofit, certain amounts of capital enter into the time periods for the gradual increase of equipment in order to improve the performance of the process, e.g., to lead to an increase in conversion or energy recovery. This capital may arise from internal sources within the company or can be borrowed. The financial benefits resulting from the improvements at the end of the time period can be reinvested for the additional reconstruction over the next development phases. In this way, self‐sustained capital generation for future reconstruction is achieved, and any additional funds required are minimized. The objective function of the proposed problem is formulated as the net present worth of the gradually retrofitted system. This enables several techno‐economic analyses to be performed. The proposed multiperiod investment strategies are illustrated by an investment planning example for the retrofit of a heat‐exchanger network.     

Resource constrained project scheduling problem with setup times after preemptive processes

In this paper, the preemptive resource constrained project scheduling problem with set up times is investigated. In this problem, a fixed setup time is required to restart when an process is preempted. The project contains activities inter-related by finish to start type precedence relations with a time lag of zero, which require a set of renewable resources. The problem formed in this way is an NP-hard. A mixed integer programming model is proposed for the problem and a parameters tuned meta-heuristic namely genetic algorithm is proposed to solve it. To evaluate the validation and performance of the proposed algorithm a set of 100 test problems is used. Comparative statistical results show that the proposed algorithm is efficiently capable to solve the problem.     

优化增量型随机权神经网络及应用

针对传统增量型随机权神经网络(I-RVFLNs)存在网络参数难以优化确定、模型收敛速度慢和结构复杂的问题,提出一种优化增量型随机权神经网络算法,即O-I-RVFLNs。与传统I-RVFLNs不同,所提O-I-RVFLNs算法首先设定了一个期望的建模残差向量,然后在每次新增隐层节点时,选择可以达到或小于此节点期望残差的输入权值和偏置作为该节点的输入参数,进而提高网络的收敛速度。除此之外,考虑到算法在不断迭代更新过程中建模误差越来越小,下降趋势越来越不明显的问题,将各指标参数相邻两次迭代均方根误差的差值考虑在算法终止条件内,并借鉴统计过程控制中的西电规则制定了相应的算法收敛判定准则。最后,基于UCI能效数据和实际高炉工业数据,对所提O-I-RVFLNs算法进行了验证和应用。结果表明,相对于其他RVFLNs算法,所提算法建立的数据模型能够获得更紧凑的网络结构以及更好的泛化性能和预测精度。     

Optimization of esterification and transesterification of esters based on experiment planning

The operation of the optimization algorithm based on planning the experiment via a preliminary analysis of the number of the degrees of freedom has been demonstrated in the preparation of esters. An analysis of the number of degrees of freedom of both individual apparatuses (reactors, reactive distillation columns, and others) and process flow diagrams formed from them has been performed. Mathematical models of processes have been plotted by experiment planning and their optimization has been performed. It has been shown that, when using the principle of combination, substantial gains in energy consumption are achieved for the processes considered. Based on the studies performed, it has been suggested that combined process provides the gain in energy consumption compared to recirculation flowsheet in all cases where the preliminary removal of products from the reaction zone is maintained due to the separating component.