Search

Scheduling tasks in production facilities are usually hybrid optimization problems of a large combinatorial nature. They involve solving, in near-real time, the integration of the operation of several batch units of continuous dynamics with the discrete manufacture of items in processing lines. Moreover, one has to deal with uncertainty (process delays, unexpected stops) and the management of shared resources (energy, water, etc.) including decisions made by plant operators: still, some tasks in the scheduling layers are done manually. Manufacturing Execution Systems (MESs) are intended to support plant personnel at this level. However, there is still much work to do in terms of performing automatic scheduling, computed in real time, that guides managers to achieve an optimal operation of such complex cyber-physical systems. This work proposes a closed-loop approach to handle the uncertainty arising when facing the online scheduling of supply lines and parallel batch units. These units often share some resources, so effects due to concurrent resource consumption on the system dynamics are explicitly considered in the presented formulation. The proposed decision support system is tested onsite in a tuna cannery, to handle short-term online scheduling of sterilization processes that deal with limited steam, carts, and operators as shared resources

[Resumen] En el presente artículo se aborda el problema de optimización estocástica de la programación de operaciones de crudos en una refinería con terminal marítima. En primer lugar, se pretende evaluar el rendimiento de un modelo de programación estocástica de dos etapas. Para ello calculamos las medidas Valor Esperado de la Información Perfecta (EVPI) y Valor de la Solución Estocástica (VSS), las cuales nos permiten valorar y comparar la solución del modelo estocástico frente a soluciones obtenidas a partir de modelos determinísticos. En segundo lugar, llevamos a cabo un análisis de las soluciones obtenidas al incluir la gestión del riesgo en el modelo estocástico, utilizando la medida Valor en Riesgo Condicional (CVaR)., [Abstract] This paper addresses the optimization of crude oil operations, considering uncertainty, in a marineaccess refinery. First, we evaluate the performance of a two-stage stochastic programming model. For this purpose, we calculate the measures Expected Value of Perfect Information (EVPI) and Value of Stochastic Solution (VSS), which allow us to assess and compare the solution of the stochastic model against solutions obtained from deterministic models. Secondly, we present an analysis of the solutions obtained by including risk management in the stochastic model, using the Conditional Value at Risk (CVaR) measure.

Thermal processing of canned products is subject to variability among the different items in a batch, which results in differences in terms of food quality and safety. Such variability is mainly caused by heterogeneity in the thermal processing unit as well as differences in the quantity of food, its compaction, and the quantity of packing liquid from item to item. Besides, for a proper design of the process operation strategies, different conflicting objectives, such as maximization of product quality and uniformity or minimization of processing time and energy consumption, must be considered. In this paper, we present a suitable formulation to efficiently address the multi-objective optimization problem of thermal processing of canned food under the presence of product variability. Such variability has been characterized from experimental data provided by an industrial cannery. Our formulation exploits the monotonic and convex behavior of the quality/safety indicators and the product kinetics with respect to the uncertain thermal coefficients, avoiding thus the computationally-expensive stochastic approach via Monte-Carlo simulations. In particular, customizable retort temperature profiles are considered as a beneficial alternative to the standard constant retort temperature practice

[Resumen] En este trabajo se presenta el modelado mecanístico de una configuración reactor anóxicofotobiorreactor aerobio de algas y bacterias con recirculación de biomasa para el tratamiento de aguas residuales domésticas. El modelo desarrollado permite estimar la concentración total de sólidos en suspensión en cada uno de los elementos que conforman la instalación. Con el fin de ajustar el modelo a los datos experimentales, se realiza la estimación de los principales parámetros que influyen sobre la concentración total de biomasa en ambos reactores y el sedimentador. El modelo obtenido se valida mediante simulación, comparando los resultados del mismo con los datos experimentales., [Abstract] This work deals with modelling of anoxic-aerobic algal-bacterial photobioreactor with biomass recycling for wastewater treatment. Total suspended solids concentration in elements composing installation is estimated through mathematical model. Parameter estimation in both reactors and settler is realized in order to fit model outputs to experimental data. Model validation is performed via simulation, comparing model outputs and experimental data.

[Resumen] En este trabajo se presenta la implementación de un controlador predictivo basado en un modelo no lineal (NMPC) y un optimizador de consignas en tiempo real (RTO) a una planta piloto híbrida. La planta consta de un reactor CSTR, dos caudalímetros, una bomba, una válvula y cuatro sensores de temperatura. La reacción química se emula a partir de las medidas del proceso y el calor que esta generaría se aplica mediante resistencias eléctricas. Así, la única sustancia involucrada es agua, se conserva la hidrodinámica del proceso y se evitan los inconvenientes típicos del manejo de sustancias químicas. Se desarrolla y ajusta un modelo basado en medidas reales que luego es usado como base para el MPC y el RTO. Se hacen pruebas de seguimiento de consigna del controlador y optimización económica del RTO. El conjunto planta híbrida + MPC + RTO conforma una plataforma flexible, a la vez que realista, para evaluar técnicas de optimización avanzada donde existen discrepancias entre planta-modelo y entre modelos., [Abstract] This work presents the implementation of a real-time optimization (RTO) and a nonlinear model predictive controller (NMPC) to a hybrid pilot plant. The plant consists of a Van de Vusse CSTR, two flowmeters, a pump, a valve, and four temperature sensors. In order to maintain the hydrodynamics and avoid typical problems with chemical substances, all the reactions are simulated. So, the only substance involved in the process is water. A calculation block simulates the chemical reaction using a model and real-time process measurements, calculating the heat generated by the reaction. This heat is applied to the reactor using two heating coils. Another model for the RTO and NMPC layers is formulated, and the parameters are adjusted using experimental measurements. Then, two experiments are carried out: trajectory tracking for the MPC, and economic optimization for the RTO. The present paper shows the advantages of using hybrid pilot plants for the study and research of advanced control strategies.

[Resumen] El presente trabajo se centra en la optimización de la programación de operaciones llevadas a cabo en un sistema compuesto por una refinería y una terminal marítima. El principal desafío radica en coordinar las decisiones que se tomen en ambas instalaciones, y lograr una solución óptima en términos económicos. A partir de la premisa mencionada anteriormente, se presenta la formulación de un modelo de programación matemática mixto entero no lineal con representación continua del tiempo. Dicho modelo es capaz de dar soporte al proceso de toma de decisiones en forma integral., [Abstract] This work focuses on the optimization of short-term crude oil operations scheduling in a system composed of a refinery and a maritime terminal within the middle segment of the oil industry supply chain. The main challenge lies in coordinating the decisions made at both facilities and achieving a solution that is economically competitive. Based on the aforementioned premise, this work presents a continuous-time mixed integer nonlinear programming (MINLP) formulation in order to support the decision-making process in an integral way.

This paper addresses the problem of closed-loop operation of scheduling, together with the interaction of control and scheduling, for a class of processes that appear very often in industry: those that combine continuous production lines with parallel batch units that share some resources. The paper presents a novel approach to this problem, including batching and a new type of precedence in the assignment problem. It also considers the effect of shared resources on the duration of the cycle time of the batch units. The approach is illustrated with a real-life example of a canned tuna factory

Current market trends and business globalization have steered industries towards large production volumes, complex alternative recipes and a larger product portfolio, thus leading to processes of increased complexity. Optimal production scheduling has a direct impact on the overall efficiencyof processing facilities and can lead to increased productivity and minimal costs. Although a plethora of works study the optimal production scheduling problem [1], academic developments have been mostly tested on generic but small problem instances. This is mainly due to the computational complexity of real industrial problems, which easily become intractable. This gap between theory and industrial practice has been identified by the scientific community [2] and some recent works have successfully investigated medium to large problem instances [3]. In this work different solution strategies are proposed for a large real-life food industry

[Resumen] Cuando distintos equipos similares trabajan de manera concurrente compartiendo un recurso, los perfiles de trabajo de cada uno se pueden ver afectados por un consumo simultáneo. De esta manera, surge un problema de programación de las tareas o scheduling, para encontrar la planificación temporal y de asignación óptima que tenga en cuenta estas interdependencias entre los equipos. Este problema se puede complicar aún más si la duración de los procesos depende de los tipos de producto a tratar. En este trabajo se presenta una herramienta que optimiza el scheduling de una seccón de esterilizado de una planta de conservas, donde el vapor se comparte entre distintos autoclaves paralelos. Los autoclaves se emplean en el proceso de esterilizado de las latas de conserva, que tienen distintos perfiles de calentamiento dependiendo del producto que contienen y de la forma de las mismas. Con un enfoque de horizonte deslizante, la herramienta puede ser ejecutada en tiempo real y así poder adaptarse a los cambios que muy probablemente se producirán a lo largo de los turnos de trabajo. El modelo se genera como un problema de programación lineal mixta-entera, y con un planteamiento de slots continuos. Se presenta una mejora de los algoritmos de precedencia que se han venido utilizando en la literatura. Finalmente, se muestran con un ejemplo los resultados obtenidos que demuestran la posibilidad de su implantación real en la planta., [Abstract] When similar pieces of equipment work concurrently sharing one resource, the working profiles can be affected by the consumption of the other equipment. Therefore, a scheduling problem arises in order to find the best temporal planification taking into account these alterations. This problem could get more complex if the profiles are not fixed, but depends on the products introduced in the equipment. In this work the authors present one tool to optimize the scheduling in a sterilization section inside a tuna canning plant, keeping in mind the alterations that the heating procedure of the autoclaves suffer if two or more autoclaves start at the same time. This tool has a moving horizon approach in order to be run iteratively, and to adapt the solution to the actual state of the plant.

[Abstract] This work aims to optimise the operation of an industrial heat-recovery section in a fibre-production factory. The section is formed by a network of heat exchangers able to be supplied with different hot sources. The goal is minimising the resource utilisation in real time (the sources usage) while satisfying a set of operational constrains. Hence, problems of economically optimal hot-sources allocation to heat exchangers arise. The problem is formulated and solved via mixed-integer non-linear programming. Furthermore, concerns about the practical implementation and the degradation of the equipment due to the fouling effect are also taken into account in the optimisation. In this way, the decision-support tool not only provides the optimal allocation, but also suggestions on which heat exchangers are potentially beneficial to be clean.

[Resumen] Este trabajo trata el problema de modelado y distribución óptima de agua en una red de refrigeración asociada a plantas de evaporación. El fin último es desarrollar una herramienta de optimización en tiempo real para balancear el uso de agua con el consumo energético de las plantas (vapor vivo). El problema de optimización incluye modelos experimentales del impacto que tiene la operación del equipo de refrigeración en la eficiencia global de las plantas. Se ha propuesto una metodología eficaz de modelado para optimización, de forma que se asegure la coherencia física de los modelos con el proceso. Además, la herramienta proporciona una estimación periódica de la pérdida de eficiencia en los equipos debido a una operación continuada (ensuciamiento en intercambiadores de calor)., [Abstract] This work deals with the problem of optimal distribution of cooling water through a network of evaporation plants. The end goal is developing an RTO tool to balance the water usage with the energy consumption in the plants (live steam). The optimization problem uses experimental models to assess the impact of the cooling system operation into the global plants efficiency. An effective modeling methodology for optimization purposes is proposed such that models physical coherence is ensured. Moreover, the tool provides a periodic estimation of the equipment performance degradation due to continuous operation (fouling in tube heat exchangers).

[Abstract] This study analyses the usefulness of process network dynamics simulation for decision-making in refinery hydrogen networks. A theoretical hydrogen network of three desulphurisation plants is modelled, and three case scenarios discussed: baseline, high demand, and low demand. Discussion focuses on how the information from the simulation is interpreted and its usefulness for debottlenecking, scheduling and what-if analysis. Stress is put on dynamics of the system and their consequences in process operation throughout the network. Hydrogen purity is highlighted as the most affected variable, and discussed its network wide effect. In addition, the responses of inflows, outflows and headers are analysed. Although the model used is a simplified representation of the actual processes, the simulation analysis showed potential as decision-making support not provided with steady state models. Further researches based on real case-studies should be conducted to better conclude on the efficient usage of simulation in aiding refinery hydrogen networks operational decisions

[Abstract] This work presents a decision-support tool to address the model-based optimization approach for online load allocation and scheduling of cleaning operations in an evaporation network. The aim is improving the resource efficiency by supplying the optimal solution for a given production goal. The approach includes the semi-automatic update of evaporator models, which is based on historical data for minimal modelling effort. The structure of the problem is formulated via mixed-integer programming and integrated into the plant supervision systems. Production constraints, concerns about the practical implementation and visualization preferences are also taken into account in the design of the prototypical tool.

[Resumen] El diseño de modelos apropiados para la simulación es, normalmente, un proceso largo y tedioso; por ello, no es conveniente utilizar el mismo modelo detallado para control y optimización. En la mayoría de los casos, los modelos necesarios para la optimización pueden ser más sencillos siempre y cuando reproduzcan las características principales de los modelos complejos. En este trabajo se ha diseñado un sistema para obtener modelos algebraicos sencillos a partir de un modelado más complejo desarrollado para la simulación.

[Abstract] This work contains the design and testing of a greybox model based on first principles for the drying section of a Medium Density Fiberboard (MDF) manufacturing plant for use in a Model-Predictive Controller (MPC).The fitting of the model to experimental data has been solved with the data reconciliation technique, and the model has then been implemented and tested in simulation for proper behavior. Due to lac k of information in a number of relevant process values, as well as low trustworthiness of some measurements,state and disturbance estimators have been determined as necessary to include in the final MPC control schema.

[Resumen] En este trabajo se describe el desarrollo e implementación de un sistema de control de una planta piloto química. Dicha planta piloto tiene la característica de que el componente principal del proceso, la reacción exotérmica, está siendo simulado. Antes de configurar el sistema SCADA, se ha modelado y simulado el proceso, y definido el sistema de control que se llevará a cabo. Destacar que el sistema SCADA además de proporcionar una interfaz gráfica de visualización del sistema, conecta la tarjeta de adquisición de datos de la planta con diferentes módulos externos accesibles como servidores OPC con diferentes funcionalidades: cuatro capaces de realizar cálculos PID y que de forma conjunta constituyen el sistema de control, y uno que funciona como un bloque calculador.

[Resumen] Se investigan el modelado y la identificación de procesos de fermentación Acetona-Butanol-Etanol (ABE). El enfoque tradicional intenta ajustar los parametros del modelo mediante una optimización para que las simulacines coincidan con los datos experimentales. Estas optimizaciones a menudo tardan mucho y no siempre consiguen un buen ajuste por la necesidad de imponer una estructura de modelo fijo y por la no-convexidad de la función de coste resultante. Se presenta un enfoque que divide el problema en unos sub-problemas que se pueden resolver de forma más efectiva y que elige el modelo del crecimiento de celulas libremente usando ALAMO (Automatic Learning of Algebraic MOdels). Finalmente, se implementa el algoritmo y realiza una identificación con datos reales y se com- prueban los resultados mediante una validación., [Abstract] The modeling and parameter identification of an Acetone-Butanol-Ethanol (ABE) fermentation process is investigated. The traditional approach tries to adjust the model parameters by means of an optimization such that the simulations fit the experimental data. These optimizations often take a lot of time and do not achieve good fits due to the necessity of imposing a xed model structure and due to the non-convexity of the resulting cost function. A different approach, which divides the big problem into smaller and more efficiently solvable subproblems, is presented. Its advantage is that it determines a model of the cellular growth term on the go using ALAMO (Automatic Learning of Algebraic MOdels) and thus others more degrees of freedom. Lastly, the algorithm is implemented, tested and validated with real data.

[Resumen] La operación óptima del proceso usando “real time optimization (RTO)” no está siempre garantizada debido a la presencia de incertidumbre en los modelos que son utilizados para la toma de decisiones y debido a las diferencias que existen entre las capas de la estructura de control que operan en diferente escalas de tiempo y usan distintas clases de modelos. La metodología de adaptación por modificadores, más conocida como “Modifier- adaptation (MA)”, sirve para alcanzar la optimalidad a pesar de la presencia de incertidumbre, utilizando para ello medidas de la planta. Normalmente, los modificadores son actualizados usando información del estado estacionario para calcular los gradientes del proceso, lo que implica una lenta convergencia hacia el óptimo real. Para acelerar la convergencia del método, en este artículo se estiman los gradientes a partir de información del transitorio usando el método de estimación de mínimos cuadrados recursivo. En este artículo se presenta la implementación de esta metodología sobre la simulación del reactor Otto-Williams.