Includes bibliography. / Successful multivariable control system design demands knowledge, skill and creativity of the designer. The goal of the research described in this dissertation was to investigate, implement, and evaluate methods by which artificial intelligence techniques, in a broad sense, may be used in a design system to assist the user. An intelligent, interactive, control system design tool has been developed to fulfil this aim. The design tool comprises two main components; an expert system on the upper level, and a powerful CACSD package on the lower level. The expert system has been constructed to assist and guide the designer in using the facilities provided by the underlying CACSD package. Unlike other expert systems, the user is also aided in formulating and refining a comprehensive and achievable design specification, and in dealing with conflicts which may arise within this specification. The assistance is aimed at both novice and experienced designers. The CACSD package includes a synthesis program which attempts to find a controller that satisfies the design specification. The synthesis program is based upon a recent factorization theory approach, where the linear multivariable control system design problem is translated into, and techniques efficiency solved as, a quadratic programming problem, which significantly improve the time and space of this method have been developed, making it practical to solve substantial multivariable design problems using only a microcomputer. The design system has been used by students at the University of Cape Town. Designs produced using the expert system tool are compared against those produced using classical design methods.

No / A multivariable model of an automotive gas turbine, obtained from the linearized system equations is investigated. To facilitate vehicle speed changes, whilst protecting the system against thermal damage, control of the power turbine inlet gas temperature and gas generator speed is proposed by feedback regulation. Fuel flow and the power turbine nozzle area variations are the selected, manipulatable inputs. Owing to the limited control energy available for regulation purposes a multivariable, optimum, minimum control effort strategy is employed in the inner loop controller design study. Simulated, open and closed loop system responses are presented for purposes of comparison. Significant improvements in the transient response interaction reaction times and low steady state output interaction achieved using passive compensation and output feedback alone. Simplification of the closed loop configuration is proposed in the final implementation without performance penalties.

Made available in DSpace on 2016-08-31T13:33:39Z (GMT). No. of bitstreams: 1 ClaytonAMO_DISSERT.pdf: 1420692 bytes, checksum: 6fff19d912e4ebade2d582644a6d1364 (MD5) Previous issue date: 2014-02-21 / In this paper, we propose an appropriate genetic algorithm to design state observers for multivariable control systems, which response time meets the specifications previously provided by the user of the system, in terms of the maximum overshoot, and in terms of the settling time. For this, will be designed and evaluated different forms of initialization and different forms of setup the genetic algorithm. Furthermore, we intend to investigate the performance results of this technique in control systems utilizing state observers full order or reduced order. Through this analysis, intend to contribute to the specification of an appropriate genetic algorithm to design state observers for multivariable control systems, by making this an efficient alternative to traditional methods in the literature / Neste trabalho, será proposto um algoritmo genético adequado ao projeto de observadores de estado para sistemas de controle multivariável, cuja resposta temporal atenda às especificações previamente fornecidas pelo usuário do sistema, tanto em termos do máximo sobressinal, quanto em termos do tempo de acomodação. Para isso, serão concebidas e avaliadas diferentes formas de inicialização e diferentes formas de configuração de algoritmo genético. Por outro lado, pretende-se investigar os resultados de desempenho desta técnica em sistemas de controle que utilizem os observadores de estado de ordem plena, ou de ordem reduzida. Por meio desta análise, pretendemos contribuir com a especificação de um algoritmo genético adequado ao projeto de observadores de estado para sistemas de controle multivariável, fazendo-se desta uma alternativa eficiente aos métodos tradicionais presentes na literatura

[ES] El control descentralizado de sistemas multivariables es una tarea compleja y su eficiencia depende principalmente de la selección adecuada de sus lazos de control. Por lo general, para seleccionar estos lazos de control se calculan medidas de interacción entre sus variables. Las metodologías clásicas que se han desarrollado para este propósito pueden dar resultados divergentes (en cuanto a los lazos de control a establecer). Esto es debido, entre otras cosas, a que miden las interacciones entre las variables del sistema de diferentes maneras. Además, normalmente no incorporan en el proceso de selección de lazos de control la sintonización de sus controladores. En esta tesis se ha desarrollado una metodología para seleccionar lazos de control óptimos en sistemas multivariables usando un enfoque de optimización multiobjetivo. La metodología analiza el problema de selección óptima de lazos de control y sintonización óptima de las estructuras de control en un marco de trabajo unificado. La metodología permite analizar las características de cada combinación de lazos de control de manera detallada comparando sus desempeños de forma global, lo cual permite a un diseñador tener información relevante para tomar decisiones adecuadas para controlar eficientemente un proceso multivariable. En la metodología propuesta se muestra como las preferencias del diseñador juegan un papel muy importante en la selección de los lazos de control en un sistema multivariable. En esta tesis se aplica la nueva metodología propuesta a varios problemas de ingeniería de control tanto lineales como no lineales. En estos ejemplos se compara la metodología propuesta con las metodologías clásicas de selección de lazos de control más usadas. Esto ha permitido revelar información valiosa sobre el control descentralizado de sistemas multivariables que no hubiese sido factible obtener con las metodologías tradicionales. / [CA] El control descentralitzat de sistemes multivariables és una tasca complexa i la seua eficiència depén principalment de la selecció adequada dels seus llaços de control. En general per a seleccionar aquests llaços de control es calculen mesures d'interacció entre les seues variables. Les metodologies clàssiques que s'han desenvolupat per a aquest propòsit poden donar resultats divergents (quant als llaços de control a establir). Això és degut, entre altres coses, al fet que mesuren les interaccions entre les variables del sistema de diferents maneres. A més normalment no incorporen en el procés de selecció de llaços de control la sintonització dels seus controladors. En aquesta tesi s'ha desenvolupat una metodologia per a seleccionar llaços de control òptims en sistemes multivariables usant un enfocament d'optimització multi-objectiu. La metodologia analitza el problema de selecció òptima de llaços de control i sintonització òptima de les estructures de control en un marc de treball unificat. La metodologia permet analitzar les característiques de cada combinació de llaços de control de manera detallada comparant els seus acompliments de manera global, la qual cosa permet a un dissenyador tindre informació rellevant per a prendre decisions adequades per a controlar eficientment un procés multivariable. En la metodologia proposada es mostra com les preferències del dissenyador tenen un rol molt important en la selecció dels llaços de control en un sistema multivariable. En aquesta tesi s'aplica la nova metodologia proposada a diversos problemes d'enginyeria de control tant lineals com no lineals. En aquests exemples es compara la metodologia proposada amb les metodologies clàssiques de selecció de llaços de control més usades. Això ha permés revelar informació valuosa sobre el control descentralitzat de sistemes multivariables que no haguera sigut factible obtindre amb les metodologies tradicionals. / [EN] Decentralized control of multivariable systems is a complex problem and its efficiency depends mainly on the suitable selection of its control loops (inputoutput pairings). In general, to select these control loops, measures of interaction between their variables are calculated. The classical methodologies that have been developed for this purpose can give divergent results (in terms of the type of loop pairing to choose). This is because they generally analyze the loop pairing selection and controller tuning independently and optimize a single objective. In this thesis a methodology to select optimal input-output pairings in multivariable systems using a multiobjective optimization approach has been developed. The methodology analyzes the problem of optimal selection of control loops and optimal tuning of control structures in a unified framework. The methodology allows a detailed analysis of the characteristics of each control loop, globally comparing their performance, which allows a designer to have relevant information to make adequate decisions to efficiently control a multivariable process. The proposed methodology shows how the designer's preferences have a very important role in the selection of an input-output pairing in a multivariate system. In this thesis, the new proposed methodology is applied to various control engineering problems, both linear and non-linear. In these examples, the proposed methodology is compared with the classical methodologies of selection of input-output pairings most used for the control of multivariable systems. This has revealed valuable information on the decentralized control of multivariate systems that would not have been feasible to obtain with traditional methodologies. / Este trabajo ha sido subvencionado por la Universidad Politécnica Salesiana (UPS) a través del convenio CB-755-2015 / Huilcapi Subia, V. (2021). Desarrollo de una metodología para la selección de lazos de control en sistemas multivariables mediante técnicas de optimización multiobjetivo [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/165014 / TESIS