Academic Positions

  • 2018 2017

    Post. Doctoral Researcher

    Aalto University Department of Chemical and Metallurgical Engineering

    • Present 2013

      Assistant Professor

      Institute of Information and Communication Technologies, Bulgaria, Department of Intelligent Systems

    • 2012 2009

      Assistant Professor

      Institute of Cryobiology and Food Technologies, Agricultural Academy, Bulgaria, Department of Cryobiology and Lyophilization

    • 2016 2015

      Visiting Lecturer

      Technical Univeristy-Sofia, branch Plovdiv, Faculty of Electronics and Automation

      2014 2013

      Visiting Lecturer

      Technical Univeristy-Sofia, Faculty of Computer Systems and Control

Education & Training

  • CERT 2017

    Fault Detection for Complex Systems

    Universidad de Vallodolid, Spain

  • CERT 2017

    Project Managment 1 & 2

    Aalto Univeristy, Espoo, Finland

  • Ph.D. 2011

    Ph.D. in Engineering Physics/Control Systems

    Institute of Cryobiology and Food Technologies, Agricultural Academy, Bulgaria

  • M.Sc.2005

    Master of Science in Control Engineering

    Technical University-Sofia, branch Plovdiv, Bulgaria

  • B.Sc.2003

    Bachelor of Science in Control Engineering

    Technical University-Sofia, branch Plovdiv, Bulgaria

Honors, Awards and Grants

  • 2011
    1-st Award for Best Young Researcher Paper
    image
    on 4-th Scientific Session for Young Researchers, University of Chemical Technology and Metallurgy, Bulgaria.
  • 2008
    2-nd Award for Best Paper
    image
    on 4-th International IEEE conference on "Intelligent Systems", Varna, Bulgaria
  • 2005
    Master of Science with Honors
    image
    of the Technical University-Sofia, branch Plovdiv, Bulgaria

Research Projects

  • image

    InterCriteria Analysis – A New Approach to Decision Making

    Very short description of the project.

    The InterCriteria Analysis can be successfully applied to problems, where measuring according to some of the criteria is slower or more expensive, which results in delaying or raising the cost of the overall process of decision making. When solving such problems it is necessary to adopt an approach for reasonable elimination of these criteria, in order to achieve economy and efficiency.

    The approach has already demonstrated first evidences of its potential, when applied to economic data, and there have been outlined specific areas of its future application, for which we can obtain appropriate test data. Hence, one of the project objectives is specifying the general framework of the problems addressed by the approach. There will be investigated the connections between the new approach and the multicriteria decision making methods, as well as its relation with the cognitive maps. Another project objective is the development of a software application, implementing the InterCriteria Analysis approach.

  • image

    FP7 - Advanced Computing for Innovation

    Very short description of the project.

    The general objective of the projectis to strengthen the ICT research and innovation capacity by increasing the knowledge and skills of IICT researchers in emerging areas as well as by purchasing modern research infrastructure.

Filter by type:

Sort by year:

Supervisory Tuning of Nonlinear Model Predictive Controller

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers Proceedings of the International Conference on "Intelligent Control Systems", Brno, Czech Republic, 29 August - 11 September, 2005, pp. 128 - 133, ISBN 80-214-2976-3

Abstract

It is proposed in this work a fuzzy supervisory algorithm for adaptive tuning of a weighting factor in Nonlinear Model Predictive Control (NMPC). A nonlinear predictive controller is designed on the basis of the Takagi-Sugeno neuro fuzzy predictive model. A high control performance can be obtained, by on-line adaptation of the tuned parameter. The effectiveness of the proposed approach is demonstrated by experimental simulations in MATLAB & SIMULINK environment to level control of a three cascaded water tanks. The simulation results prove the efficiency of the proposed algorithm and show improvements in some quality control criteria.

Multivariable Nonlinear Fuzzy–Neural Predictive Controller

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers Scientific Works on Scientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2005’, University of Food Technologies, Plovdiv, Bulgaria, Vol. LII, Issue 3, pp. 208 - 213, 13 – 14 October, 2005. ISSN 0477-0250, Bulgaria

Abstract

In this paper is presented a method for designing a multivariable nonlinear predictive controller based on a Takagi-Sugeno fuzzy-neural model. It is used a simplified gradient technique in optimization task to calculate predictions on the future control actions and for on-line adaptation of the fuzzy-neural model. The proposed nonlinear predictive algorithm is used to control level in a system with quadruple water tanks.

Improving performance of Nonlinear Model Predictive controller by incorporating a digital filter

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers Scientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2006’, University of Food Technologies, 28 October, 2006, Plovdiv, Bulgaria. Vol. LIII, Issue 2, pp. 111-116, ISSN 0477-0250

Abstract

It is presented in this paper a possibility to improve the performance of nonlinear model predictive controller by using a digital predictive filter. The predictive controller is based on a Takagi-Sugeno fuzzy-neural model and it is simplified by a gradient optimization algorithm. The proposed modified algorithm is used to control level in a system with triple water tanks.

Comparative study of different Intelligent Control Algorithms

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers Scientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2006’, University of Food Technologies, 28 October, 2006, Plovdiv, Bulgaria, Vol. LIII, Issue 2, pp. 117-122, ISSN 0477-0250

Abstract

In this paper it is presented a comparative study of different intelligent control algorithms. The study includes comparaison between hybrid fuzzy PID controllers and Generalized Predictive Controller (GPC). Analysis is made in MATLAB/SIMULINK environement to contol a temperature in a heat exchanger.

Nonlinear Model Predictive Controller with Adaptive Learning rate Scheduling of an internal model

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers In Proceedings of the international conference ‘Modern Trends in Control’, Technical University of Kosice, Slovakia, 15-30 June 2006, pp. 289-298, ISBN 80-969224-6-7

Abstract

It is presented in this paper a method for adaptive learning rate scheduling of the internal model in nonlinear model predictive controller. The controller is based on a Takagi-Sugeno fuzzy-neural model and a simplified gradient optimization algorithm. The proposed approach is used to control the level in a system with triple water tanks.

Nonlinear Model Based Predictive Controller using a Fuzzy-Neural Hammerstein model

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers In Proceedings of the international conference ‘Modern Trends in Control’, Technical University of Kosice, Slovakia,15-30 June 2006, pp. 299-308, ISBN 80-969224-6-7

Abstract

It is presented in this paper a method for designing a nonlinear model predictive controller. The controller is based on a Hammerstein fuzzy-neural predictive model and а simplified gradient optimization algorithm. The proposed approach is used to control the level in a system with triple water tanks.

Adaptive Supervisory tuning of Nonlinear Model Predictive controller for a heat exchanger

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers In Proceedings of the international IFAC conference ‘Energy saving control in plants and buildings’, Bansko, Bulgaria, 2-5 October 2006, Pages 93-98, ISBN-10:954-9641-47-3

Abstract

It is presented in this paper an adaptive predictive supervisory algorithm to the temperature control of a heating system with a heat exchanger. The nonlinear predictive control strategy is designed on the basis of a Takagi-Sugeno fuzzy-neural model and a simple optimization procedure. An additional supervisory level in the control system is introduced for adaptive tuning of a weighting factor in the predefined optimization criterion. Using the proposed algorithm a higher system performance can be achieved which leads to reduction of the energy consumption into the heating system. The proposed approach is studied by experimental simulations to control a temperature in the heating system. Copyright © 2006 IFAC

Fuzzy-Neural Model Predictive Control of a building heating system

Margarita Terziyska,Yancho Todorov, Michail Petrov
Conference Papers In Proceedings of the international IFAC conference ‘Energy saving control in plants and buildings’. Bansko, Bulgaria, 2-5 October 2006, Pages 69-74, ISBN-10:954-9641-47-3

Abstract

This paper describes the development of a Model Predictive Controller with supervision control of a building heating system. A fuzzy–neural model and optimizing procedure as a part of a nonlinear predictive controller are utilized on-line to determine the future values of control actions based on dependence between outdoor and indoor temperatures. A learning algorithm for parameters in fuzzy-neural implementation of the predictive model is additionally applied. Simulation results with a model of a single room heating system demonstrate that a better system performance can be achieved in comparison to classical PID control. Copyright  2006 IFAC

Nonlinear model based predictive controller using a Fuzzy-Neural Wiener-Hammerstein model

Yancho Todorov, Margarita Terziyska, Michail Petrov
Conference Papers Proceedings of the international conference “Process Control’07”, Stibske Pleso, Slovak Republic, Slovakia, Pages 216-1 - 216-6, June, 2007.

Abstract

It is presented in this paper a method for designing a nonlinear model predictive controller. The controller is based on a hybrid Wiener-Hammerstein fuzzy-neural predictive model and а simplified gradient optimization algorithm. The proposed approach is used to control the product temperature in a Lyophlization plant. The controller efficiency is tested and proved by simulation experiments in Matlab & Simulink.

Modeling of a Lyophilization plant by means of hybrid Fuzzy-Neural Wiener-Hammerstein model

Yancho Todorov
Conference PapersScientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2007’, University of Food Technologies, Plovdiv, Bulgaria, Volume LIX, Issue 3, Pages 89 - 94, 14 October, 2007.

Abstract

It is presented in this paper a method for designing a nonlinear model predictive controller. The controller is based on a hybrid Wiener-Hammerstein fuzzy-neural predictive model and а simplified gradient optimization algorithm. The proposed approach is used to control the product temperature in a Lyophlization plant. The controller efficiency is tested and proved by simulation experiments in Matlab & Simulink.

Volterra Model Predictive Control of a Lyophilization plant model

Yancho Todorov, Tsvetan Tsvetkov
Conference PapersProceedings of the 4-th International IEEE conference "Intelligent Systems", Varna, Bulgaira, 6-8, September 2008, Volume III, Pages 20-13/20-18.

Abstract

Lyophilization plants are widely used by pharmaceutical industries to produce stable dried medications and important preparations. Since, a Lyophilization cycle involves a high energy demands it is needed to be used an improved control strategy in order to minimize the operating costs. This paper describes a method for designing a nonlinear model predictive controller to be used in a Lyophilization plant. The controller is based on a truncated fuzzy-neural Volterra predictive model and a simplified gradient optimization algorithm. The proposed approach is studied to control the product temperature in a Lyophilization plant. The efficiency of the proposed approach is tested and proved by simulation experiments.

Modeling of a Lyophilization plant by means of a fuzzy-neural Volterra model

Yancho Todorov, Sylvia Ivanova
Conference Papers Scientific Works on Scientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2008’, University of Food Technologies, Plovdiv, Bulgaria, 16 October 2008, Volume LV, Issue 2, Pages 147-152.

Abstract

It is presented in this paper a design methodology for a truncated Volterra Fuzzy-Neural process model. The structure of the model is implemented by means of a simple fuzzy inference system with a learning procedure based on the minimization of an instant error. It is made simulation experiment in modeling of a nonlinear relation in a Lyophilization plant.

Adaptive tuning of a fuzzy PID controller for lyophilization plant

Yancho Todorov, Iliana Nacheva, Daniela Miteva
Conference Papers Scientific Works on Scientific Conference with International Participation ‘FOOD SCIENCE, ENGINEERING AND TECHNOLOGIES 2009’, University of Food Technologies, Plovdiv, Bulgaria, Volume LVI, Issue 3, Pages 151-159, October, 2009.

Abstract

А design methodology for adaptive tuning of а fuzzy PID type controller is presented. The adaptation of the fuzzy PID gain coefficients is done by implementing a simple internal model scheme by Hammerstein fuzzy-neural predictive model and a gradient optimization procedure. Simulation experiments to control a nonlinear relation in Lyophilization plant are made.

Model Predictive Control of a Lyophilization plant: A simplified approach using Wiener and Hammerstein systems

Yancho Todorov, Michail Petrov
Journal Paper Journal of Control and Intelligent Systems, Volume 39, Issue 1, January 2011, Pages 23-32

Abstract

Lyophilization process is widely used in pharmaceutical industries, preparing stable dried medications and important biopreparations, so they remain stable and easier to store at room temperature. Since a lyophilization cycle involves high energy demands, an improved control strategy has to be used in order to minimize the operating costs. This paper deals with the design methodology of nonlinear model predictive controllers for lyophilization plant. The controllers are based on fuzzy-neural predictive models and simplified gradient optimization algorithm. As predictive models, fuzzy-neural implementations of Hammerstein and Wiener-Hammerstein systems are used. Such structures provide fast and reliable system identification using small number of parameters which reduces the computational burden during the optimization procedure. The potential benefits of the proposed approaches are demonstrated by simulation experiments.

Fuzzy Model Predictive Control of a MIMO system

Michail Petrov, Sevil Ahmed, Albena Taneva,Yancho Todorov
Conference PaperProceedings of the 40-th ANNIVERSARY CONFERENCE OF DEPARTMENT "AUTOMATION", University of Chemical Technilogy and Metallurgy, Bulgaria, 18 March 2011, Pages 45 – 48, ISBN 978-954-465-043-8.

Abstract

In this paper Nonlinear Model Predictive Control (NMPC) is studied as a more applicable approach for optimal control of multivariable processes. A state-space representation of a Takagi-Sugeno type fuzzy-neural model is proposed as a predictive model. This type of model ensures easier description and direct computation of the gradient control vector during the predictive optimization task. The identification procedure relies on a two-step training algorithm, which is known in field of artificial neural networks. The proposed Fuzzy NMPC approach is studied by experimental simulations in Matlab/Simulink® environment in order to control the liquid levels in a multi tank system. The simulation results demonstrate that the main process variables have a good performance and the process control quality is satisfied.

Model Predictive Control of a Lyophilization plant: A Newton method approach

Yancho Todorov, Michail Petrov, Sevil Ahmed
Conference PaperProceedings of Intrnational Conference "Automatics and Informatics' 2011", Society for Automatics and Informatics, Bulgaria, Pages B-89 - B-92.

Abstract

This paper describes a method for designing a nonlinear model predictive controller to be used in a lyophilization plant. The controller is based on a truncated fuzzy-neural Volterra predictive model and а simplified Newton method as optimization algorithm. The proposed approach is studied to control the product temperature in a lyophilization plant. The efficiency of the proposed approach is tested and proved by simulation experiments.

Volterra Model Predictive Control of a Lyophilization plant: A Newton Optimization approach

Yancho Todorov, Sevil Ahmed, Michail Petrov
Journal PaperJournal of Information Technologies and Control, Submitted 28/10/2011, Year IX, Volume 4, Pages 9 - 15, Printed in 2012, ISSN:1312-2622.

Abstract

Lyophilization process is widely used by pharmaceutical and food industries preparing stable dried medications and important biopreparations. Recent advances in lyophilization technology impose the application of innovative strategies for reliable determination of the current process conditions and control of the drying cycles. This paper describes a method for designing a nonlinear model predictive controller to be used in a lyophilization plant. The controller is based on a truncated fuzzy-neural Volterra predictive model and а simplified Newton method as optimization algorithm. The proposed approach is studied to control the product temperature in a lyophilization plant. Several simulation experiments have been performed in order to demonstrate the efficiency of the proposed approach. The obtained results are compared with the classical Gradient optimization procedure.

State-Space Predictive Control of a Lyophilization plant: A fuzzy-neural Hammerstein model approach

Yancho Todorov, Sevil Ahmed, Michail Petrov
Conference PaperProceedings of the 1st IFAC Workshop "Dynamics and Control in Agriculture and Food Processing", 13-16 June, 2012, Plovdiv, Bulgaria, Pages 181 - 186.

Abstract

This paper describes a design methodology for model predictive controller based on Hammerstein State-Space model.The model nonlinearity is approximated by using a simple Takagi-Sugeno inference,while the linear part is flexibly introduced. The antecedent part of the fuzzy rules representsa general State-Space systemand its parameters are scheduled at each sampling period by minimization of an instant errormeasurement function.As optimization procedure itis used an implementation of Hildreth Quadratic Programming algorithm. The effectiveness of the presented approach is demonstrated by experiments for control of a lyophilization plant.

Implementations of a Hammerstein Fuzzy-Neural Model for Predictive Control of a Lyophilization Plant

Yancho Todorov, Sevil Ahmed, Michail Petrov, Vasiliiy Chitanov
Conference PaperProceedings of the 6-th IEEE Conference on "Intelligent Systems", Volume II, Pages 315 - 319, 6 - 8 September, 2012, Sofia, Bulgaria.

Abstract

In this paper it is proposed a nonlinear approach to model predictive control that is based on a Takagi–Sugeno (TS) fuzzy model. An industrial evaporator system is taken as an exemplary process and its prediction model is used in the controller. Evaporators are widely used in the food processing industry to remove a portion of the water from food products. This reduces bulk and weight for subsequent processing, increases solids content (as for jams and molasses), helps preserve the product, provides convenience to the end consumer and concentrates color or flavor. Accurate nonlinear models of the evaporator system components are described. The final model of the evaporator system in state space implementation is used in model predictive control (MPC) scheme. Optimization objectives in MPC include minimization of the difference between the predicted and desired response trajectories, and the control effort subjected to prescribed constraints. The case study is implemented using MATLAB/Simulink. The simulation results show that the main process variables have good performance and process quality is satisfied.

State-Space Predictive Control of a Lyophilization plant: A fuzzy-neural Hammerstein model approach

Sevil Ahmed, Michail Petrov, Alebena Taneva,Yancho Todorov
Conference PaperProceedings of the 1st IFAC Workshop "Dynamics and Control in Agriculture and Food Processing", 13-16 June, 2012, Plovdiv, Bulgaria, Pages 187 - 192.

Abstract

In this paper it is proposed a nonlinear approach to model predictive control that is based on a Takagi–Sugeno (TS) fuzzy model. An industrial evaporator system is taken as an exemplary process and its prediction model is used in the controller. Evaporators are widely used in the food processing industry to remove a portion of the water from food products. This reduces bulk and weight for subsequent processing, increases solids content (as for jams and molasses), helps preserve the product, provides convenience to the end consumer and concentrates color or flavor. Accurate nonlinear models of the evaporator system components are described. The final model of the evaporator system in state space implementation is used in model predictive control (MPC) scheme. Optimization objectives in MPC include minimization of the difference between the predicted and desired response trajectories, and the control effort subjected to prescribed constraints. The case study is implemented using MATLAB/Simulink. The simulation results show that the main process variables have good performance and process quality is satisfied.

Application of Fuzzy-Neural Modeling and Quadratic Programming for Model Predictive Control

Yancho Todorov, Sevil Ahmed, Michail Petrov
Conference Paper Proceedings of the 6-th Annual Meeting of Bulgarian Section of the Sosiety of Industrial and Applied Mathematics, Pages 146-155, 19-20 December, 2012, Sofia, Bulgaria.

Abstract

In this paper it is proposed a nonlinear approach to model predictive control that is based on a Takagi–Sugeno (TS) fuzzy model. An industrial evaporator system is taken as an exemplary process and its prediction model is used in the controller. Evaporators are widely used in the food processing industry to remove a portion of the water from food products. This reduces bulk and weight for subsequent processing, increases solids content (as for jams and molasses), helps preserve the product, provides convenience to the end consumer and concentrates color or flavor. Accurate nonlinear models of the evaporator system components are described. The final model of the evaporator system in state space implementation is used in model predictive control (MPC) scheme. Optimization objectives in MPC include minimization of the difference between the predicted and desired response trajectories, and the control effort subjected to prescribed constraints. The case study is implemented using MATLAB/Simulink. The simulation results show that the main process variables have good performance and process quality is satisfied.

Real-time Supervisory tuning of Predictive Controller

Margarita Terziyska, Yancho Todorov, Michail Petrov
Journal Paper Journal of the Technical University-Sofia, branch Plovdiv, "Fundamental Sciences and Applications", Volume 19, Pages 155-160, 2013.

Abstract

One of the main problems in predictive controllers is their tuning. The values of the prediction and control horizons and weighting factors in the optimization criterion are usually determined heuristically. This paper presents an approach of supervisory tuning of predictive controller. The supervisor carries out the adaptive adjustment of the weighting factor ρ that directly affects the output value of control in constant horizons of prediction. The proposed algorithm for neural-fuzzy predictive control has been tested in real conditions on a laboratory heating system, which is located in the Faculty of Mechanical Engineering of the Technical University in Prague.

Fuzzy-Neural Predictive Control using Levenberg-Marquardt optimization approach

Yancho Todorov, Margarita Terziyska, Sevil Ahmed, Michail Petrov
Conference Paper Proceedings of International IEEE conference on Innovations in Intelligent Systems and Applications, INISTA'2013, Albena, Bulgaria, 2013, Pages. 1 - 5, ISBN 978-1-4799-0659-8.

Abstract

It is proposed in this paper an investigation of the influence of different optimization policies in Nonlinear Model Predictive Control (NMPC). The nonlinear predictive controller is designed on the basis of Takagi-Sugeno fuzzy-neural predictive model, while the proper computation of the control actions is done by applying the Gradient descent, Newton-Raphson approach and Levenberg-Marquardt optimization procedures. The efficiency of the proposed optimization strategies is demonstrated by experiments in MATLAB environment to control a Continuous Stirred Tank Reactor (CSTR).

Soft computing applications in food technology

Yancho Todorov, Iliana Nacheva, Petia Metodieva, Maria Doneva, Tsvetan Tsvetkov
Journal Paper Bulgarian Journal of Agricultural Science, Volume 19, Issue 3, Pages 503-507, ISSN 1310 - 0351, 2013.

Abstract

This paper describes the potentials of the application of modern soft computing techniques into development stage of contemporary food products. Recently, soft computing has been extensively studied and applied for scientific research and engineering purposes. In biological and food engineering, researchers have developed methods of fuzzy logic, artificial neural networks, genetic algorithms, decision trees, and support vector machines to study complex characteristics of many products in order to be adopted cost effective measures satisfying the production constraints and consumer expectations

Recurrent Fuzzy-Neural Network with Fast Learning Algorithm for Predictive Control

Yancho Todorov, Margarita Terziyska, Michail Petrov
Book Chapter Springer | September, 2013 | ISBN: 978-3-642-40727-7
image

The book constitutes the proceedings of the 23rd International Conference on Artificial Neural Networks, ICANN 2013, held in Sofia, Bulgaria, in September 2013. The 78 papers included in the proceedings were carefully reviewed and selected from 128 submissions. The focus of the papers is on following topics: neurofinance graphical network models, brain machine interfaces, evolutionary neural networks, neurodynamics, complex systems, neuroinformatics, neuroengineering, hybrid systems, computational biology, neural hardware, bioinspired embedded systems, and collective intelligence.

Levenberg-Marcquadt training approach for Recurrent Fuzzy-Neural Network

Yancho Todorov, Margarita Terziyska, Michail Petrov
Conference Paper Proceedings of International Conference "Automatics and Informatics'2013,Sofia, Bulgaria, 3 - 6 October, 2013, Pages I-139-I-141, ISSN 1313-1850.

Abstract

This paper describes the development of a Levenberg-Marcquardt learning approach for the consequent part of the fuzzy rules in recurrent Takagi-Sugeno type inference. The recurrent relation in the proposed fuzzy-neural network represents a global feedback from the fuzzy-neural network output to its relevant inputs, being fuzzified in the next training sample. To prove the efficiency of the proposed fuzzy-neural structure, simulation experiments for prediction of Mackey-Glass chaotic time series are performed. A comparison with classical Gradient descent method is also studied.

Nonlinear model predictive control of an evaporator system using fuzzy-neural observer

Sevil Ahmed, Michail Petrov, Albena Taneva, Yancho Todorov
Conference Paper Proceedings of the International conference "Food Science, Engineering and Technologies'2013", University of Food Technologies-Plovdiv, Bulgaria,2013, Volume LX, Pages 57 - 62, ISSN 1314- 7102.

Abstract

In this paper it is proposed a nonlinear approach to model predictive control that is based on a Takagi– Sugeno (TS)fuzzy model representation of a state observer. An industrial evaporator system is taken as an exemplary process and its prediction model is used in the controller. Accurate nonlinear models of the evaporator system components are described. The final model of the evaporator system in state space implementation is used in model based control. The MPC scheme is based on an explicit use of the predictive model of the system response to obtain the control actions by minimizing a cost function. Optimization objectives in MPC include minimization of the difference between the predicted and desired response trajectories, and the control effort subjected to prescribed constraints. The case study is implemented using MATLAB/Simulink. The simulation results show that the main process variables have good performance and process quality is satisfied.

Nonlinear model predictive control of an evaporator system using fuzzy-neural observer

Margarita Terziyska, Yancho Todorov, Michail Petrov
Conference Paper Proceedings of the International conference "Food Science, Engineering and Technologies'2013" University of Food Technologies-Plovdiv, Bulgaria, 2013, Volume LX, Pages 63 - 68, ISSN 1314- 7102.

Abstract

A nonlinear predictive controller based on a recurrent neuro-fuzzy model is presented in this paper. Neuro-fuzzy model is realized with the T-S inference mechanism and includes global and local (after the rules layer) feedbacks. The proposed model is coupled with an optimization approach for computation of the control actions into a model based predictive controller. The efficiency of the proposed controл policy is proved by simulation experiments to control a Continuous Stirred Tank Reactor (CSTR).

Modeling of Chaotic Time Series by Interval Type-2 NEO-Fuzzy Neural Network

Yancho Todorov, Margarita Terziyska
Book Chapter Springer | September, 2014 | ISBN: 978-3-319-11179-7
image

The book constitutes the proceedings of the 24th International Conference on Artificial Neural Networks, ICANN 2014, held in Hamburg, Germany, in September 2014. The 107 papers included in the proceedings were carefully reviewed and selected from 173 submissions. The focus of the papers is on following topics: recurrent networks; competitive learning and self-organisation; clustering and classification; trees and graphs; human-machine interaction; deep networks; theory; reinforcement learning and action; vision; supervised learning; dynamical models and time series; neuroscience; and applications.

Nonlinear model predictive control of an evaporator system using fuzzy-neural observer

Margarita Terziyska, Yancho Todorov
Journal Paper International Journal of Reasoning Based Intelligent Systems, Volume 6, Issue 3/4, Pages 136-144, 2014, ISSN online: 1755-0564, ISSN print: 1755-0556.

Abstract

This paper describes the development of fast optimisation polices based on Newtonian approaches, as effective algorithms to solve the on-line optimisation task, during the operation of a predictive controller. To simplify the calculation of the control actions, an iterative solutions based on Newton-Raphson and Levenberg-Marquardt approaches, are proposed. To avoid the computational load related to Hessian inversion, a simple Gaussian elimination in a form of matrix decomposition is applied. As plant response predictor, a Takagi-Sugeno fuzzy-neural network, with global and local (after the rules layer) recurrent nodes, is used. The efficiency of the proposed optimisation strategies is demonstrated by simulation experiments in MATLAB environment to control a continuous stirred tank reactor.

Innovative technologies for creation of probiotic foods

Iliana Nacheva, Maria Doneva,Yancho Todorov, Petia Metodieva, Daniela Miteva, Krasimir Dimov, Tsvetan Tsvetkov
Journal Paper Bulgarian Journal of Agricultural Science, Volume 20, Issue 4, Pages 830-833, 2014, ISSN 1310 - 0351.

Abstract

Lyophilized probiotic products have been developed by high technology methods for nutrition prophylaxis and healthy nutrition to ensure a maximum working capacity and well- being of each individual. In their composition are included natural sources of essential bioelements and physiologically active substances – a complex of probiotic lactic acid bacteria, oligosaccharides, antioxidants, vitamins and others. Their fine consistency and chemical composition create a possibility for rhythmical introduction in the organism of nutritious mixtures with adequate content of plastic substances and energy. Their proved healthy effect makes them suitable to be included in combined nutrition diets for ensuring of better quality of life in terms of the effective prevention and improvement of the health status of the population.

State-Space Fuzzy-Neural Network for Modeling of Nonlinear Dynamics

Yancho Todorov, Margarita Terziyska
Conference Paper Proceedings of International IEEE symposium Innovations in Intelligent Systems and Applications, INISTA'2014, Alberobello, Italy. Pages 212 - 217, ISBN: 978-1-4799-3019-7.

Abstract

This paper describes a novel idea for designing a fuzzy-neural network for modeling of nonlinear system dynamics. The presented approach assumes a state-space representation in order to obtain a more compact form of the model, without statement of a great number of parameters needed to represent a nonlinear behavior. To increase the flexibility of the network, simple Takagi-Sugeno inferences are used to estimate the current system states, by a set of a multiple local linear state estimators. Afterwards, the output of the network is defined, as function of the current and estimated system parameters. A simple learning algorithm based on two step Gradient descent procedure to adjust the network parameters, is applied. The potentials of the proposed modeling network are demonstrated by simulation experiments to model an oscillating pendulum and a nonlinear drying plant.

An Intelligent Approach to Formulate the Contents of Novel Functional Food

Yancho Todorov, Maria Doneva, Petia Metodieva, Iliana Nacheva
Conference Paper Proceedings of International IEEE symposium Innovations in Intelligent Systems and Applications, INISTA'2014, Alberobello, Italy. Pages 98 - 103, ISBN: 978-1-4799-3019-7.

Abstract

This paper describes an applied approach using an Adaptive Neuro-Fuzzy Inference System to formulate the contents of novel diary functional food. In the development stage for a new functional food, it is required a careful balancing in the product ingredients in order to be achieved not only a healthily effect but an acceptable sensory properties. This imposes the solving of multiparametric task, how to select an optimal product composition in order to obtain a products with a great percent of consumer acceptability. Since, the main sensory characteristics of the products can be assessed by trained panelists and encoded by a numerical estimates, the task can be solved by a simple fuzzy input-output mapping, in order to conclude how each component of the product affects a selected sensory characteristic.

Neo-fuzzy Neural Network for modeling of nonlinear MIMO dynamics

MArgarita Terziyska,Yancho Todorov, Luybka Doukovska
Journal Paper Journal of Technical University-Sofia,branch Plovdiv, "Fundamental Sicences and Applications", 2015, Volume 21, Pages 65 - 71.

Abstract

This paper presents the structure and the learning algorithm of a multi-input multi-output (MIMO) Neo-fuzzy neural network for nonlinear system modeling. The applied approach lies on the idea of Neo-fuzzy neuron whose dynamics depend on its own temporal behavior, while his output is generated as a singleton function. To demonstrate efficiency of the proposed modeling structure, a simulation experiments in Matlab environment modeling a nonlinear MIMO process dynamics are performed.

Analysis of the particle distributuon in granular functional food

Margarita Terziyska,Yancho Todorov, Iliana Nacheva, Maria Doneva, Petia Metodieva
Journal Paper Journal of Technical University-Sofia,branch Plovdiv, "Fundamental Sicences and Applications", 2015, Volume 21, Pages 361 - 366.

Abstract

In this paper an analysis of the particles distribution in novel granular functional food by using a laser of particle analyzer ANALYSETTE 22 NanoTec plus is studied. The main objective of the investigation is to evaluate the influence of a varying ingredient in the product on its granular distribution. A future work on the basis of the obtained results will be the assessment of the influence of the particle distribution on various physical parameters of the product composition using intelligent modeling techniques.

Simple heuristic approach for training of Type-2 NEO-Fuzzy Neural Network

Yancho Todorov, Margarita Terziyska
Conference Paper Proceedings of IEEE International Conference "Process Control'2015", Stribske Pleso, Slovak Republic, Pages 278 - 283.

Abstract

This paper describes the development of Interval Type-2 NEO-Fuzzy Neural Network for modeling of complex dynamics. The proposed network represents a parallel set of multiple zero order Sugeno type approximations, related only to their own input argument. As learning procedure a simple heuristic backpropagation approach, where the sign of the gradient is taken into account, is adopted. To improve the robustness of the network and the possibilities for handling uncertainties, Interval Type-2 Gaussian fuzzy sets are introduced into the network topology. The potentials of the proposed approach in modeling of Mackey-Glass and Rossler Chaotic time series are studied. A comparison is made with the classical Gradient Descent learning approach.

Distributed Fuzzy-Neural State-Space Predictive Control

Yancho Todorov, Margarita Terziyska, Luybka Doukovska
Conference Paper Proceedings of IEEE International Conference "Process Control'2015", Stribske Pleso, Slovak Republic, Pages 31 - 36.

Abstract

This paper describes the development of nonlinear state-space predictive controller based on distributed fuzzy-neural model. The presented approach assumes a state-space representation in order to obtain more compact form of the model, without statement of a great number of parameters needed to represent nonlinear relations. To increase the flexibility of the network, a set of fuzzy inferences is used to estimate the current system states, as well as to construct a simple predictor needed to update the future system behavior along the defined horizons. At each sampling period an optimization task performing Quadratic Programming minimization assuming the imposed constraints on the system parameters is solved. The performance of the proposed controller is assessed by simulation experiments in modeling and control of nonlinear systems with complicated dynamics.

NEO-Fuzzy State-Space Predictive Control

Yancho Todorov, Margarita Terziyska, Michail Petrov
Conference Paper Proceedings of IFAC International Conference "Technology, Culture and International Stability, TECIS'15", 23 - 27 September, 2015, Sozopol, Bulgaria, Published in IFACPapersOn-line, Volume 48, Issue 24, Pages 99-104.

Abstract

This paper describes the development of a novel state-space model predictive controller. The proposed modelling structure used to capture and predict the nonlinear process dynamics lies on the concept for a neo-fuzzy neuron, deployed in state-space. The introduced approach represents a set of simple fuzzy inferences along the temporal behaviour of each input node, whose dynamics is expressed as a singleton function. The learning algorithm for the proposed modelling structure is realized as a gradient descent procedure. On the basis of the obtained neo-fuzzy state-space model, a fuzzy predictor for the purpose of predictive control is developed. The achieved predictions are used to optimize the future system response by implementing a quadratic programming optimization procedure along the stated controller horizons. The potentials of the proposed approach are studied by simulation experiments to modelling and control of a nonlinear drying plant.

Comparative Study on the Efficiency of Hybrid Learning Procedures Used for Training of Fuzzy-Neural Networks

Yancho Todorov, Margarita Terziyska
Book Chapter Cambridge Scholars Publishing | June, 2015 | ISBN-13: 978-1-4438-6401-5
image

In this book, a wide range of problems concerning recent achievements in the field of industrial and applied mathematics are presented. It provides new ideas and research for scientists developing and studying mathematical methods and algorithms, and researchers applying them for solving real-life problems. The importance of the computing infrastructure is unquestionable for the development of modern science.

The main focus of the book is the application of mathematics to industry and science. It promotes basic research in mathematics leading to new methods and techniques useful to industry and science. The volume also considers strategy-making integration between scientists of applied mathematics and those working in applied informatics, which has potential for long-lasting integration and co-operation. The integration role is regarded here as a tool for consolidation and reinforcement of the research, education and training, and for the transfer of scientific and management knowledge. This volume operates as a medium for the exchange of information and ideas between mathematicians and other technical and scientific personnel. The book will be essential for the promotion of interdisciplinary collaboration between applied mathematics and science, engineering and technology.

The main topics examined in this volume are: numerical methods and algorithms; control systems and applications; partial differential equations and real-life applications; the high performance of scientific computing; linear algebra applications; neurosciences; algorithms in industrial mathematics; equations of mathematical physics; and industrial applications of mechanics.

Intuitionistic Neo-Fuzzy predictive control

Margarita Terziyska,Yancho Todorov
Conference Paper Proceedings of IEEE Conference on "Intelligent Systems'16", 4 - 6 September, 2016, Sofia, Bulgaria, Published in IEEE Xplore, Pages 635-640.

Abstract

This paper presents an implicit predictive control strategy based on Intuitionistic Neo-Fuzzy predictor, as a first attempt to investigate the potentials of the intuitionistic fuzzy logic for the purpose of control applications. The proposed predictor represents a simple fuzzy-neural network as fusion from the concepts of the intuitionistic fuzzy logic, the neo-fuzzy neuron theory and the classical Takagi-Sugeno inference mechanism. The predictions are then coupled into generalized predictive control scheme where a standard quadratic control cost function is minimized over a set of predefined horizons. For simplicity, the considered process variables and the calculated output control sequence are iteratively bounded instead of explicitly constrained, in order to investigate the computational procedures related to implementation of an intuitionistic fuzzy logic. To investigate the potentials of the proposed predictive control approach, numerical experiments to control a Continuous Stirred Tank Reactor (CSTR) under uncertain conditions are studied.

Intuitionistic Neo-Fuzzy Network for modeling of nonlinear systems dynamics

Margarita Terziyska,Yancho Todorov
Conference Paper Proceedings of IEEE Conference on "Intelligent Systems'16", 4 - 6 September, 2016, Sofia, Bulgaria, Published in IEEE Xplore, Pages 616-621.

Abstract

In this paper, an approach to design an Intuitionistic Neo-Fuzzy Network (INFN) is presented. The proposed architecture combines the advantages of the Intuitionistic Fuzzy Logic (IFL) to deal with uncertainties and the Neo-Fuzzy Neural Network approach to represent nonlinear systems with topologies including small number of parameters. As a learning approach for the consequent fuzzy rules parameters, the gradient optimization procedure is proposed. The investigate the potentials of the generated INF structure, the modeling of a three benchmark chaotic time series - Mackey-Glass, Lorenz and Rossler under uncertain conditions is investigated. The obtained results prove the flexibility of the approach and its further extension to Model Predictive Control is investigated too.

Input space selective fuzzification in intuitionistic semi fuzzy-neural network

Margarita Terziyska,Yancho Todorov, Marius Olteanu
Conference Paper Proceedings of IEEE Conference on "Electronics, Computer and Artificial Intelligence'16", 30 June - 2 July, 2016, Ploesti, Romania, Published in IEEE Xplore

Abstract

In this paper, the influence of the selective fuzzification of the input space in Intuitionistic Semi-Fuzzy Neural Network (ISFNN) is investigated. The ISFNN represents a structure modification of the classical fuzzy-neural approach where selective fuzzification as a means to reduce the number of the generated fuzzy rules is proposed, thus expected to reduce the number of the associated learning parameters and to achieve a degree of computational simplicity. On the other hand, the potentials of the network are supplemented by intuitionistic fuzzy logic, in order to handle uncertain data variations. As a learning procedure for the proposed structure, a two-step gradient descent algorithm is employed. To investigate the influence of input space fuzzificaton, several test experiments in modeling of a two benchmark chaotic systems — Mackey-Glass and Rossler chaotic time series are made.

Reduced Rule-Base Fuzzy-Neural Network

Margarita Terziyska,Yancho Todorov
Book Chapter Springer | February, 2017 | ISBN: 978-3-319-41438-6
image

In this paper two different fuzzy-neural systems with reduced fuzzy rules bases, namely Distributed Adaptive Neuro Fuzzy Architecture (DANFA) and Semi Fuzzy Neural Network (SFNN), are presented. Both structures are realized with Takagi-Sugeno fuzzy inference mechanism and they posses reduced number of parameters for update during the learning procedure. Thus, the computational time for algorithm execution is additionally reduced, which make the modeling structures a promising solution for real time applications. As a learning approach for the designed structures a simplified two-step gradient descent approach is implemented. To demonstrate the potentials of both models, simulation experiments with two benchmark chaotic time systems—Mackey-Glass and Rossler are studied. The obtained results show accurate models performance with minimal prediction error.

State-Space Fuzzy-Neural Predictive Control

Yancho Todorov, Margarita Terziyska, Michail Petrov
Book Chapter Springer | January, 2017 | ISBN: 978-3-319-41438-6
image

The purpose of this work is to give an idea about the available potentials of state-space predictive control methodology based on fuzzy-neural modeling technique and different optimization procedures for process control. The proposed controller methodologies are based on Fuzzy-Neural State-Space Hammerstein model and variants of Quadratic Programming optimization algorithms. The effects of the proposed approaches are studied by simulation experiments to control a primary drying cycle in small-scale freeze-drying plant. The obtained results show a well-driven drying process without violation of the system constraints and accurate minimum error model prediction of the considered system states and output.

Teaching History

  • 2018 2017

    Process Control, I/II

    The courses cover topics from the basics of contol systems design and implementation.

    2017 2016

    Nonlinear Systems and Neural Networks

    The course covers topics from the theory of Nonlinear Systems with emphasis to modeling of nonlinear systems using neural networks.

  • 2017 2016

    Artificial Intelligence

    The course covers topics from the classics of Artificial Intelligence with emphasis to relevant concepts and paradigms.

  • 2014 2013

    Foundations of Neural Networks

    The course covers topics from the field of Neural Networks, giving insigt to the basic structures of simple neuron, the available network topologies and thier respective leargning algoritgms. An attention is given to the classical paradigms and structures.

  • 2015 2013

    Foundations of Fuzzy and Neuro-Fuzzy systems.

    The course cover topics of the cassical fuzzy systems theory and mathematics with emphasis to fuzzy Mamdani and Takagi-Sugeno approaches. A special attention is given to application of the proposed methodologies in purpose to modeling and control of dynamical systems.

List of the availabe courses to provide

  • FNN Data

    Neuro-Fuzzy Networks for dynamical data analysis

    The purpose of the course is to provide a specialized insight of the potentials of different neuro-fuzzy topologies for dynamical data estimation. An emphasis is given to approaches for building of reduced rule-based networks for for capturing the bahaviour of fast dynamical systems.

  • FNN MPC

    Fuuzy-Neural Model Predictive Control

    MPC is andvanced control technique widely used in practise during the past years which uses an availabe proceess model and an optimization procedure to compute a set of optimal controls for a specific plant process. An attention is given to approaches of using fuzzy-neural mdeling techniqes for the purpose of dynamical plant identification and cosntruction of simple fuzzy predictors for providing a set of prediction horizons used to callculate the optimal controls. Different time domain representation of the MPC are proposed using either regresion or state-space representation.

  • Bio OPT

    Bioinspired optimisation

    The purpose of the course is to provide lectures in the fields of bioinspired and collective intelligence with emphasis to available algorithms for heuristic optimisation. The covered topics include swarm intelligence and artificial ant and bee colonies.

At My Office

You can find me at my office located at Aalto University, School of Chemical Engineering, Department of Chemical and Metallurgical Engineering (3rd floor), Room E301, Kemistintie 1, 02150 Espoo, Finland

I am at my office every day from 9:00 until 17:00, but you may consider a call to fix an appointment.

Citations

You can find the complete list of my citations following my profiles:

Google Scholar

Scopus

How to find my office?

My office is reachable by bus (102,102T, 103, 195) from the Helsinki city center