Incremental identification of hybrid process models

Data Reconciliation in Reaction Systems using the Concept of Extents

Recent advances in CFD modeling of bioethanol production processes

Eight single (SGA) and eight multi-population (MGA) genetic algorithms (GA) differing in the sequence of implementation of the main genetic operators’ selection, crossover and mutation, or omitting the mutation operator, have been examined for the purposes of parameter identification of a Saccharomyces cerevisiae fed-batch fermentation process model. The influence of some of the main genetic algorithm parameters, namely number of individuals, maximum number of generations, generation gap, crossover and mutation rates for both SGA and MGA, and insertion and migration probability for MGA only, have been investigated in depth. Almost all applied SGA and MGA led to similar values of the optimization criterion but the SGA with operators’ sequence mutation, crossover and selection, and MGA with operators’ sequence crossover, selection and mutation, are significantly faster than others while keeping the model accuracy. Among the considered GA parameters, generation gap influences most significantly to SGA and MGA convergence time, saving of about 40% of computational time of the algorithms without affecting the model accuracy.

Identification of distillation process model using a transformation from the discrete-time model to continuous-time model

The paper presents modelling and the process of dynamic model identification of the crankshaft system. At the beginning some information about design, operation and test of crankshaft systems is shown. The next section describes the process of the crank-piston mechanism with a vibration damper modelling and the method of obtaining the motion equation. The last but one section demonstrates some aspects of the process of dynamic model identification. Moreover, tests and numerical simulation are presented. At the end, there is a summary of the whole paper. Conclusions about modelling and identification of the dynamic model of crankshaft systems are shown.

Chapter 3 - Rate laws

Hybrid modelling is widely employed in chemical engineering to generate highly accurate predictions. Such an approach merges first-principle modelling with machine learning techniques to identify and model the epistemic uncertainty from experimental data. Despite its advantages, this still requires cross-domain competencies that are difficult to find in the chemical industry and high human involvement. The possibility of automating the identification and training model would be significantly beneficial for the widespread adoption of hybrid modelling methodology within the chemical industry. This work presents a novel algorithm for the automatic identification of hybrid models (HMs) starting from the first-principle representation of the system, described by differential equation sets. The methodology formulates the problem as mixed-integer programming, identifying the equation running under uncertainty, identifying the machine learning model hyperparameters, and training the latter. The Differential Evolution algorithm drives the identification and training tasks. The methodology is validated in three cases, namely a dynamic reaction system, a dynamic bioreactor and a Lotka-Volterra oscillator deviated with polynomial or MRF equation on different levels, generating 14 validation cases. On all of them, the model correctly identifies the position of the uncertainty and the functional form to approximate it. The methodology returns automatically trained HMs with a mean absolute percentage error in the range of 10%, which is in line with the experimental error of the data. The methodology presented in this work presents a step toward the automatic generation of HMs for dynamic systems and the widespread of this technology in the chemical industry.

Mass transfer limiting step determination and a new modified empirical model for Th(ІV) biosorption in a fixed-bed column

<abstract> <p>Mine blasting can have a destructive effect on the environment. Among these effects, air overpressure (AOp) is a major concern. Therefore, a careful assessment of the AOp intensity should be conducted before any blasting operation in order to minimize the associated environmental detriment. Several empirical models have been established to predict and control AOp. However, the current empirical methods have many limitations, including low accuracy, poor generalizability, consideration only of linear relationships among influencing parameters, and investigation of only a few influencing parameters. Thus, the current research presents a hybrid model which combines an extreme gradient boosting algorithm (XGB) with grey wolf optimization (GWO) for accurately predicting AOp. Furthermore, an empirical model and gene expression programming (GEP) were used to assess the validity of the hybrid model (XGB-GWO). An analysis of 66 blastings with their corresponding AOp values and influential parameters was conducted to achieve the goals of this research. The efficiency of AOp prediction methods was evaluated in terms of mean absolute error (MAE), coefficient of determination (R<sup>2</sup>), and root mean square error (RMSE). Based on the calculations, the XGB-GWO model has performed as well as the empirical and GEP models. Next, the most significant parameters for predicting AOp were determined using a sensitivity analysis. Based on the analysis results, stemming length and rock quality designation (RQD) were identified as two variables with the greatest influence. This study showed that the proposed XGB-GWO method was robust and applicable for predicting AOp driven by blasting operations.</p> </abstract>

This chapter begins with an explanation about the importance of modeling and optimization of manufacturing processes not only from the scientific and researching point of view but also for practical industrial applications. Then it introduces the hybrid approach which combines artificial intelligence tools and finite element method for these modeling and optimization tasks. The advantages and shortcomings of each of these techniques are exposed, highlighting the convenience of combining both methods, increasing the robustness and flexibility. Furthermore, the different approaches for combining artificial intelligence and finite element method in modeling and optimization of manufactured processes are outlined and preliminarily evaluated. 1.1 Relevance and Convenience of Hybrid Modeling and Optimization of Manufacturing Processes Modeling of the physical phenomena involved in manufacturing processes (machining, forming, foundry, etc.) has been recognized as one of the most important tasks in manufacturing research. Accurate and realistic mathematical models not only allow understanding how these phenomena take place, but also facilitate the development of new manufacturing processes. Knowledge about the quantitative relationships between the different parameters involved in manufacturing processes also permits the implementation of effective monitoring and control system, which are indispensable in the high automated modern industry. Moreover, the use of highly optimized manufacturing processes is widely accepted as a necessary condition for achieving effectiveness, efficiency and economic competitiveness in manufacturing workshops. R. Quiza et al., Hybrid Modeling and Optimization of Manufacturing, SpringerBriefs in Computational Mechanics, DOI: 10.1007/978-3-642-28085-6_1, The Author(s) 2012 1 Unfortunately, the physical nature of the phenomena underlying the manufacturing processes is not easy to understand, as they involve complex nonlinear relationships which are no completely explained up to date. This situation worsens with the use of modern materials for parts and tools. For example, for many years, the tool life of high speed steels cutting tools, working at relatively low speeds, was described by the well-known Taylor’s law (Childs et al. 2000) with an acceptable error level. However, the introduction of multi-coated carbides and PCBN tools, at high cutting speeds and hard conditions has made the Taylor’s law useless or, at less, extremely limited (Dolinsek et al. 2001). Two main approaches can be used in modeling manufacturing processes (see Fig. 1.1). On one hand, the phenomenological modeling is based on the identification and mathematical description of the physical phenomena. It has the advantage of being more realistic and accurate. Also they help to understand the mechanisms of these phenomena. This mathematical description comes in the form of expressions, usually differential equations which can be solve analytically only for a limited set of simple problems. For most of the real problems, these equations (especially partial differential equations) have not analytical solutions and approximated numerical solutions must be obtained instead. Between the numerical methods used for solving partial differential equation, the finite element method (FEM) has reach the higher application levels, because of its ability for being applied to problems defined over complex spatial domains and the relative simplicity of its computational implementation (Dixit and Dixit 2008). FEM has been successfully applied to a wide variety of manufacturing processes, such as machining (Arrazola and Ozel 2010; Mamalis et al. 2008), forming (Gudur and Dixit 2008; Shahani et al. 2009), welding (Anca et al. 2011) and foundry (Lewis et al. 2005). Sometimes, the phenomenological models are too much idealized for giving results accurate enough for being used in practical applications. In these cases, empirical models, based on correlation of experimental data, play a crucial role. The main drawback of this kind of models is its incapability for identifying or explaining the physical relationship between the involved variables. However, they usually offer accurate outcomes for industrial applications and had been widely used in optimization of manufacturing processes. Traditionally, the mathematical techniques used for correlating data in empirical models were those based on statistics. These tools are quite simple and are Fig. 1.1 Classification of modeling techniques 2

Mass transfer during surfactant-enhanced DNAPL remediation: Pore-scale experiments and new correlation

This work studies the parameter identification of a hybrid model with Particle Swarm Optimization. A hybrid model is based on selection functions that allow the switching between simple mathematical expressions in order to describe a complex behavior. In this work two performance functions are proposed to perform the identification: The former considers a switching between functions on their structure. The latter implements function penalty functions in order to avoid the evaluation of the selection functions. These functions test for the parameter identification of a Shape Memory Alloy model under a numerical simulation. The quality of the computed estimates is tested using statistical tools to assure repeatability and to verify the influence of the performance functions.

Growth and yield modeling has a long history in forestry. The methods of measuring the growth of stand basal area have evolved from those developed in the U.S.A. and Germany during the last century. Stand basal area modeling has progressed rapidly since the first widely used model was published by the U.S. Forest Service. Over the years, a variety of models have been developed for predicting the growth and yield of uneven/even-aged stands using stand-level approaches. The modeling methodology has not only moved from an empirical approach to a more ecological process-based approach but also accommodated a variety of techniques such as: 1) simultaneous equation methods, 2) difference models, 3) artificial neural network techniques, 4) linear/nonlinear regression models, and 5) matrix models. Empirical models using statistical methods were developed to reproduce accurately and precisely field observations. In contrast, process models have a shorter history, developed originally as research and education tools with the aim of increasing the understanding of cause and effect relationships. Empirical and process models can be married into hybrid models in which the shortcomings of both component approaches can, to some extent, be overcome. Algebraic difference forms of stand basal area models which consist of stand age, stand density and site quality can fully describe stand growth dynamics. This paper reviews the current literature regarding stand basal area models, discusses the basic types of models and their merits and outlines recent progress in modeling growth and dynamics of stand basal area. Future trends involving algebraic difference forms, good fitting variables and model types into stand basal area modeling strategies are discussed.

Multiscale analysis and modeling of multiphase chemical reactors

Chapter 17 - Mass transfer effects: immobilized and heterogeneous reaction systems