{"id":650,"date":"2019-04-13T01:11:18","date_gmt":"2019-04-13T04:11:18","guid":{"rendered":"https:\/\/www.ufrgs.br\/psebr\/?page_id=650"},"modified":"2019-05-12T15:09:35","modified_gmt":"2019-05-12T18:09:35","slug":"palestras-tematicas","status":"publish","type":"page","link":"https:\/\/www.ufrgs.br\/psebr\/palestras-tematicas\/","title":{"rendered":"Palestras Tem\u00e1ticas"},"content":{"rendered":"\n

Modelagem e Simula\u00e7\u00e3o de Processos<\/strong><\/h2>\n\n\n\n

Towards a biomass\nmatrix for fuels and chemicals: how Process Systems Engineering can help bridging\nthe gap for the low carbon economy?<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Roberto de Campos Giordano<\/strong>
Senior Full Professor
Chemical Engineering Department, Universidade Federal de S\u00e3o Carlos (UFSCar) <\/p>\n\n\n\n


It is a consensus nowadays that a sustainable economy will rely on a multiplicity of energy sources, with biomass playing an important role. Furthermore, biorefineries will have to yield molecules and building blocks to replace fossil oil-derived products. Nevertheless, an important gap still remains to be bridged: how to make feasible this transition in the real economy? Within this framework, techno-economic-environmental analyses of new industrial processes with the smallest possible carbon and water footprints will have increasing relevance for decision making in the near future. And PSE approaches for modeling, simulation, analysis, synthesis, energy and water integration, optimization and debottlenecking will certainly be extensively applied. Still, uncertainties concerning environmental impact and economic indices, which will be applied to decide between new, competing technical solutions, must be considered to reduce the risks of non-optimal decisions by stakeholders and governments.<\/p>\n\n\n\n

Biografia<\/strong>: Senior\nFull Professor of the Chemical Engineering Department of the Federal University\nof S\u00e3o Carlos (DEQ\/UFSCar), Brazil, where he works ever since 1980. His PhD in\nChemical Engineering was at the Polytechnic School of the S\u00e3o Paulo University\n(USP), and the pos-doc at MIT (USA), working with Biochemical Engineering. His\nresearch activities focus on the development, modeling, simulation,\noptimization and control of several industrial biotechnology processes,\nincluding production of biofuels, human and animal vaccines, hydrolysates from\nresidues of dairy and soybean industries, antibiotics via green chemistry and\nenzymes for industrial applications. The implementation of computational tools\nfor techno-economic-environmental analysis of bioprocesses\/biorefineries\n“from the cradle”, at early stages of development, is presently one\nof his main research lines.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Otimiza\u00e7\u00e3o Processos<\/strong><\/h2>\n\n\n\n

New Techniques for Real Time Optimization: Asynchronous and Persistent Adaptation<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Galo Antonio Carrillo Le Roux<\/strong>
Full Professor
Chemical Engineering Department, Polytechnic School, Universidade de S\u00e3o Paulo (USP)<\/p>\n\n\n\n



Real Time Optimization dates back to the 80\u2019s. It is a technique that has attracted the attention of the process industry because of its simplicity and of the promised earnings. However, some initial promises ran out to be much harder to get than expected. There are some intrinsic issues that take place in the application of RTO to real industrial processes that were underestimated. Those issues concern mainly difficulties in parameter estimation and on the conditions for reaching an optimal solution. There have been many advances in the recent years in those fields. However, something that is still underestimated is the problem of steady state detection. Most RTO techniques rely on steady-state models. The adaptation of such models needs explicitly that the process is in steady-state, and, in practice, because of perturbations, processes rarely meet the conditions to be considered in steady-state rigorously. In order to use the impressive arsenal developed by commercial process software suppliers for RTO the rigor to test this condition ends up being relaxed. In this presentation we discuss the consequences of relaxing SS conditions and a method that allows models of different nature (dynamic or steady-state) to be adapted to different parts of a process, asynchronously, is presented. This allows that the process be modeled, even if not strictly in steady-state, without much loss of precision. Finally, comparisons are presented based on a process involving distillation and reaction.<\/p>\n\n\n\n

Biografia<\/strong>: B.Sc. (1987) and M.Sc. (1991) in Chemical Engineering from the Department of Chemical Engineering of the Polytechnic School of the University of S\u00e3o Paulo and Ph.D. in Process Engineering from Institut National Polytechnique de Toulouse, France (1995). Research focus on modelling, simulation, optimization and process control, with emphasis in parameter estimation and system identification.<\/p>\n\n\n\n

<\/p>\n\n\n\n

S\u00edntese e Projeto de Processos<\/h2>\n\n\n\n

Application of Modeling and Simulation in Process Synthesis and Optimization<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Ivan Guilherme Bessa<\/strong>
Technology Specialist Engineer
BRASKEM<\/p>\n\n\n\n




The ways and purposes in which modeling and simulation resources are applied within BRASKEM are outlined. Emphasis is given in the path that has been followed to acquire the present experience, and in the different applications of steady-state models and dynamic models in process synthesis, process development, and process optimization. Computational fluid dynamics models deserve a special approach due to some of its features. Steady-state simulation models have been applied within BRASKEM for a long time, either for process synthesis or for off-line optimization, as a support for the work of the engineering teams involved in process follow up. Starting from in-house modeling programs developed in cooperation with universities, usage of commercial packages became the standard way for representing petrochemical processes with its typical unit operations. Even though, some preparatory steps are always necessary to ensure that models built with commercial packages will provide reliable results, useful for process synthesis or optimization. Some pitfalls that could hamper this type of work are depicted. Some studies like a flare system evaluation or a chemical reaction equipment design require usage of dynamic models. Although commercial packages could be used in these applications, sometimes \u2013 mainly for chemical reaction systems \u2013 special models developed outside the general simulation framework are more useful. BRASKEM has a fruitful experience of cooperation with universities in this field, and the development of a model for the dehydration of ethanol, with UFRJ, is a good example. Computational fluid dynamics models are growing in importance with applications in the study of special flow conditions that can affect performance of some equipment items. Examples are identification of erosion conditions in piping, and evaluation of temperature limiting conditions into furnace chambers. Such studies were not so common in the past, but internal expertise is being developed so that these models will become more widely applied. <\/p>\n\n\n\n

Biography<\/strong>: Technology Specialist Engineer at BRASKEM. He is a Chemical Engineer (from IME, Rio de Janeiro), and he holds a M.Sc. degree in Che. Engng. (from COPPE\/UFRJ). He has over 30 years of experience in COPENE and BRASKEM (Cama\u00e7ari) in petrochemical processes, where he was assigned to several engineering teams and leadership positions. He worked in the implementation of large projects, as well as in process improvement activities. During his career, he has been involved with modeling and simulation, process engineering, process design and development, and technology improvement and selection.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Controle e Instrumenta\u00e7\u00e3o de Processos<\/strong><\/h2>\n\n\n\n

A Nonlinear MPC with Optimizing Targets<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Darci Odloak<\/strong>
Full Professor
Chemical Engineering Department, Polytechnic School, Universidade de S\u00e3o Paulo (USP)<\/p>\n\n\n\n


In the presentation, it is discussed the integration of the conventional RTO with the Nonlinear Model Predictive Control in a two layer structure where the process optimization is performed in the upper layer, which solves a static problem. In the lower layer a dynamic control problem based on the Nonlinear Model Predictive Control is solved. The controller includes the optimum input targets defined by the upper layer and it is assumed that the outputs are controlled inside control zones. Different formulations that guarantee the closed-loop stability are discussed and exemplified for a small nonlinear reactor system.<\/p>\n\n\n\n

Biografia<\/strong>: Full Professor at the Department of Chemical Engineering of the Polytechnic School of the Universidade de S\u00e3o Paulo since 1987. He received a M.Sc. from the Universidade Federal do Rio de Janeiro (COPPE) and a Ph.D. from the University of Leeds-UK in 1980. He worked for Petrobras from 1991 to 1996 as the head of the Advanced Control Group that developed and implemented an in-house advanced control packaged in the main oil refineries of Brazil. His present research interest is in integration of control and real time optimization, nonlinear model predictive control and distributed model predictive control.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Assessment,\nDiagnosis, and  Maintenance  of Industrial MPCs in the Digital\nTransformation Age<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Jorge O. Trierweiler<\/strong>
Full Professor
Chemical Engineering Department, Engineering School, Universidade Federal do Rio Grande do Sul (UFRGS)<\/p>\n\n\n\n

It is hardly surprising that Big Data, i.e. the management of large and complex data volumes, is perceived as important and expectations are high. This data tsunami creates impatience since companies have invested heavily in data infrastructure for collecting, storing, and retrieving data. However, this data is not used anywhere close to its potential. Especially in the petrochemical and refinery process industries, the data infrastructure has long been a reality, mainly due to their safety needs. For these industries, Model Predictive Controllers (MPCs) are the most widely used advanced control strategy. The predictive capacity of this kind of controllers is one of its greatest advantages, since it allows that the MPC take into account the process constraints to compute the control actions. The use of MPCs can improve the process operation reducing the process variability, which directly makes the operation safer and more cost-effective. With the course of time, the operational conditions of the process naturally change, and the performance of the controller decreases if the maintenance is not performed. Among the sources of performance degradation, the low quality of the process model is one of the most frequent. Thus, precise model assessment is essential for MPC performance, diagnosis, and maintenance. Because a model is an abstraction of the real system behavior, a model-plant mismatch (MPM) will always be present. However, sometimes these MPMs are so significant that they lead the controller to poor performance. Thus, it is necessary to quantify the MPM, which cannot be compensated by the feedback controller and, therefore, will deteriorate the corresponding closed-loop behavior.
Another point is that MPC with control ranges is a common practice in industrial applications. This approach is used when the controller does not have enough degrees of freedom to control all its outputs. It means that the number of manipulated variables is smaller than the number of controlled variables and it is not possible to maintain all of them in a fixed setpoint. The alternative is to control the outputs by range, where the MPC works to maintain all the controlled variables inside a given band, called soft constraints. In the literature, the available techniques for MPC tuning usually consider a specific operating point (OP), while in real plants, controllers should be robust in a wide operating region facing different plant behaviors that arise due to disturbances, saturations, and nonlinearities. In this work, a method for MPC tuning proposed in our previous work is extended for a robust tuning for classical (square) MPCs. This technique applies to any predictive control algorithm, and it considers multi-scenarios based on the closed-loop attainable performance of the system. The sequential procedure is applied, where initially the attainable performance for each scenario (herein, different OPs are used) is determined, and an estimate of the closed-loop potential is computed. At the end, the optimum scaling for the model and the MPC tuning parameters are calculated, solving an optimization problem that uses the attainable trajectories for each scenario as a reference. The selection of the controller\u2019s process model is also determined according to constraints of the attainable performance determination.
This talk will cover all the above issues related to industrial MPCs. A holistic methodology already published in more than 10 qualified papers will be unifinly presented. The vision of GIMSCOP\/UFRGS in this area was published in several unique papers as described in [1]\u2013[11]. These papers reconcile the correct combination of the new techniques of machine learning and data mining with the basic principles (such as feedback control, optimization, phenomenological modeling, etc.). These new tools are implemented in the called BRToolKits, which are available for industrial applications through the PGA (\u201cAsset Management Platform\u201d) developed in partnership with PETROBRAS and TriSolutions.
References:<\/strong>
[1] V. Botelho, J. O. Trierweiler, M. Farenzena, and R. Duraiski, \u201cMethodology for Detecting Model-Plant Mismatches Affecting Model Predictive Control Performance,\u201d Ind. Eng. Chem. Res.<\/em>, vol. 54, no. 48, 2015.
[2] V. Botelho, J. O. Trierweiler, M. Farenzena, and R. Duraiski, \u201cPerspectives and challenges in performance assessment of model predictive control,\u201d Can. J. Chem. Eng.<\/em>, vol. 94, no. 7, 2016.
[3] V. Botelho, J. O. Trierweiler, and M. Farenzena, \u201cDiagnosis of poor performance in model predictive controllers: unmeasured disturbance versus model\u2212plant mismatch,\u201d Ind. Eng. Chem. Res.<\/em>, vol. 55, no. 44, 2016.
[4] V. R. Botelho, J. O. Trierweiler, M. Farenzena, and R. Duraiski, \u201cAssessment of model-plant mismatch by the nominal sensitivity function for unconstrained MPC,\u201d IFAC-PapersOnLine<\/em>, vol. 28, no. 8, 2015.
[5] V. Botelho, J. O. Trierweiler, and M. Farenzena, \u201cMPC Model assessment of highly coupled systems,\u201d Ind. Eng. Chem. Res.<\/em>, vol. 55, no. 50, 2016.
[6] V. R. BOTELHO, J. O. TRIERWEILER, M. FARENZENA, and G. H. M\u00dcLLER, \u201cDesafios E Perspectivas Na Auditoria De Mpcs,\u201d no. February, pp. 11422\u201311429, 2015.
[7] V. R. Botelho et al.<\/em>, \u201cModel assessment of MPCs with control ranges: An industrial application in a delayed coking unit,\u201d Control Eng. Pract.<\/em>, vol. 84, pp. 261\u2013273, 2019.
[8] J. E. W. Santos, J. O. Trierweiler, and M. Farenzena, \u201cRobust Tuning for Classical MPC through the Multi-scenarios Approach,\u201d Ind. Eng. Chem. Res.<\/em>, vol. 58, no. 8, pp. 3146\u20133158, 2019.
[9] J. E. W. Santos, J. O. Trierweiler, and M. Farenzena, \u201cModel Predictive Control Tuning Strategy for Non-Square Systems and Range Controlled Variables Based on Multi-Scenarios Approach,\u201d Ind. Eng. Chem. Res.<\/em>, vol. 56, no. 40, 2017.
[10] \u00c9. R. P. Claro, V. Botelho, J. O. Trierweiler, and M. Farenzena, \u201cModel Performance Assessment of a Predictive Controller for Propylene\/Propane Separation,\u201d IFAC-PapersOnLine<\/em>, vol. 49, no. 7, 2016.
[11] J. O. Trierweiler and L. A. Farina, \u201cRPN tuning strategy for model predictive control,\u201d J. Process Control<\/em>, vol. 13, no. 7, 2003.<\/p>\n\n\n\n

Biografia: <\/strong>Full professor at the Chemical Engineering Department of the Universidade Federal do Rio Grande do Sul (UFRGS), Brazil. He is graduated in chemical engineer in 1988 at UFRGS, with M.Sc. Degree obtained at PEQ\/COPPE\/UFRJ (Brazil) in 1991 and Dr.-Ing. at Dortmund University (Germany) in 1996 both in chemical engineering in process control and optimization. In 2001 he held a postdoctoral internship in biosensors at Leibniz University Hannover (Germany) and in 2008, in biorefineries and bioprocesses control at University RWTH Aachen (Germany). Additionally, in 2005 he completed an MBA in business management at UFRGS. Since 1997, he is the Head of the research group called GIMSCOP (Group of Intensification, Modeling, Simulation, Control, and Optimization of Processes) at UFRGS. He has supervised more than 70 M.Sc. and Ph.D. theses and has more than 326 articles published in scientific journals, events, and international conferences. In 2010, as result of his outstanding contributions to education and training in the science, engineering, and technology of instrumentation, systems, and automation, he received the \u201cDonald P. Eckman Education Award\u201d given by the ISA \u201cthe International Society of Automation\u201d. In  2017, GIMSCOP\/UFRGS + TriSolutions + PETROBRAS received the technological innovation award from ANP (Brazilian National Oil Agency). In 2019, the GIMSCOP won the GERDAU’s international challenge. <\/p>\n\n\n\n

<\/p>\n\n\n\n

Modelagem Molecular e Projeto de Pro<\/strong>duto<\/h2>\n\n\n\n

Getting closer to a fully predictive equation of state for rigorous simulation of chemical process: COSMO-SAC-Phi<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Rafael de Pelegrini Soares<\/strong>
Associate Professor
Chemical Engineering Department, Engineering School, Universidade Federal do Rio Grande do Sul (UFRGS)<\/p>\n\n\n\n



Rigorous (first principles) chemical process simulation usually rely on equations of state for the computation of phase equilibria, enthalpies, molar volumes, etc. Typically, this task is accomplished by cubic equations of state, requiring at least critical properties and the acentric factor of all substances involved. Further, for mixtures including polar compounds, binary interaction parameters are also usually required. For many industrially relevant processes these information are unavailable, hindering the use of advanced process engineering tools available today. The situation can be even worse in the design of new processes, when less known substances are involved. This issue can be addressed by the recently developed COSMO-SAC-Phi (CSP) equation of state [Fluid Phase Equilib. V. 488. Pg. 13, 2019<\/em>]. CSP relies on quantum chemistry (QC) computations, requiring only the molecular formula of each compound. Although the QC computations are very costly, they can be executed beforehand (only once for every molecule) and stored in a database. At its current state, CSP still requires some pure compound data (vapor pressure and liquid molar volume) to calibrate 3 parameters per compound. When calibrated with these pure compound data, the method was able to predict phase equilibria data with an average error of 10% for several mixtures comprising molecules with up to 10 carbons and pressures up to 100 bar. This is superior to the Soave-Redlich-Kwong performance for the same mixtures (average of 26 % error). Currently, the CSP method is being improved to become fully predictive. This would enable its use even when absolutely no experimental data is available, only compound molecular formulas would be needed.<\/p>\n\n\n\n

Biografia<\/strong>: Associate Professor at the Chemical Engineering Department of the Universidade Federal do Rio Grande do Sul (UFRGS), Brazil. He has degrees in Chemical Engineering from this same University (B.S. in 2001, M.S. in 2003, and Ph.D. in 2007). In his graduate work, Dr. Soares did research on efficient numerical methods for the dynamic simulation of complex chemical processes. This research has led to the development of an equation-oriented chemical process simulator called EMSO, which uses sophisticated schemes for automatic and symbolic differentiation combined with modern solvers for differential-algebraic systems of equations. EMSO is continuously extended and updated and today has more than 20’000 registered users worldwide. In addition to his research on dynamic simulations, Dr. Soares also collaborates with the development for the iiSE steady-state simulator. Both simulators adopt the equation-oriented approach and are very flexible for simulation and optimization studies. He received the Best M.S. Thesis in Brazil Award of Brazilian Association of Chemical Engineers in 2005. He has also received the prestigious Petrobras Technology Award twice, in 2006 and 2013, in the area of Refining and Petrochemicals. In addition to his interest on process simulation, Dr. Soares actively works on the development of thermodynamic models and computational codes related to thermo-physical properties prediction for industrial applications. In this line, he has contributed with the development of the F-SAC group-contribution activity coefficient model; the Self-Consistent Mixing Rule (SCMR); and the COSMO-SAC-Phi approach.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Detec\u00e7\u00e3o e Diagn\u00f3stico de Falhas<\/strong><\/h2>\n\n\n\n

Fault\nDetection and Diagnosis in Controlled Systems<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Lu\u00eds Cl\u00e1udio Oliveira-Lopes<\/strong>
Associate Professor
Chemical Engineering School, Universidade Federal Uberl\u00e2ndia (UFU)<\/p>\n\n\n\n




Fault detection and diagnosis (isolation and identification) in controlled chemical processes have great importance for process monitoring and fault tolerant control. Early fault detection is significant in avoiding process\/product\/performance degradation. Furthermore, an early and correct diagnostic of faulty components eases the decision process on repairing operating conditions and mitigates effects of failures in the overall plant behavior. The objective of this paper is fourfold: it presents the main used technologies for fault detection, it evaluates fault diagnosis techniques; it values the effects of abnormal behavior propagation in controlled systems and it discusses controlled embedded inherent safety measures for detection and diagnosis for control tolerance. Among several approaches discussed herein, model-based methodologies using residual generation, online and offline data-driven anomaly detection, artificial methodologies based on intelligent machines such as Support Vector Machines based on Vapnik’s statistical learning theory, Artificial Neural Networks with supervised learning and unsupervised learning and fuzzy logic based fault detection and classification in controlled chemical process are discussed. Chemical plants are used as benchmark for all studied methods. Approaches are evaluated and their strength and weakness are valued through the introduction of a detection\/diagnosis performance index, data necessary characteristics, data preparation effort and training speed of each technique are also addressed.<\/p>\n\n\n\n

Biografia<\/strong>: Professor at Universidade Federal de Uberl\u00e2ndia (UFU), Brazil, in the School of Chemical Engineering. He earned his Ph. D. in Chemical Engineering at Lehigh University (USA), M.Sc. in Chemical Engineering at COPPE\/UFRJ (Brazil), and B.S. in Chemical Engineering at the Universidade Federal da Bahia (UFBA), Brazil, he studied Petrochemical engineering at University of Bologna (Italy) and was a visiting researcher in the Chemical and Biomolecular Engineering Department at UCLA (USA). He has interest and research contributions in the areas of energy production, process control, hybrid systems, optimization, industrial safety, and fault detection and diagnosis.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Aplica\u00e7\u00f5es Industriais de Ferramentas de PSE<\/strong><\/h2>\n\n\n\n

The Digital\nTwin in the Chemical Industry<\/strong><\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Bruna Bertoni Giron<\/strong>
Senior Engineer
AVEVA<\/p>\n\n\n\n






Today the Chemical Industry major trends are related to sustainability, digitalization and innovation. Under this scenario, the Digital Twin rises as a solution for many of the chemical industry objectives, supporting the entire plant lifecycle. Traditional process simulators are single purpose architecture, old programming code with improved graphical interface and limited in capabilities. So, in the conventional approach several simulations are built for the plant project and operation, leading to high costs in engineering efforts, different software licenses and, the worst of all, data inconsistency across the lifecycle. The Digital Twin for process simulation uses a platform approach that evolves the simulation model from the conceptual engineering to the operation optimization. The same platform is used for process simulation and process utilities (cooling water, flare, steam and others), allowing engineers to evaluate how each system impacts the other. Same simulation is used for equipment and pipeline sizing. Then, the model is switched to dynamic mode for control strategy validation. Dynamic studies are performed in earlier stages in the project lifecycle, promoting savings in equipment acquisition. Simulation can also be switched back to static mode, for re-evaluation of heat and material balance and re-sizing at any time. The process simulation Digital Twin can address several issues identified under 3 mega trends presented by Marker Research Blog for the Chemical industry: <\/p>\n\n\n\n

1.Sustainability & the Circular Economy: <\/p>\n\n\n\n