Alireza Fatehi

This paper is concerned with robust identification of processes with time-varying time delays. In reality, the delay values do not simply change randomly, but there is a correlation between consecutive delays. In this paper, the correlation of time delay is modeled by the transition probability of a Markov chain. Furthermore, the measured data are often contaminated by outliers, and therefore, t-distribution is adopted to model the measurement noise. The variational Bayesian (VB) approach is applied to estimate the model parameters along with time delays. Compared with the classical expectation-maximization algorithm, VB approach has the advantage of capturing the uncertainty of the estimated parameter and time delays by providing their full probabilities. The effectiveness of the proposed method is demonstrated by both a numerical example and a pilot-scale hybrid-tank experiment.

State estimation for a system with irregular rate and delayed measurements is studied using fusion Kalman filter. Lab data in process plants is usually more accurate compared to other measurements. However, it is often slow rate and subject to variable delay and irregularity in sampling time. Fast rate state estimation can be conducted using fast rate measurement, while the slow rate lab data can be used to improve the accuracy of estimation whenever it is available. For this purpose, two Kalman filters are used to estimate the states based on each type of measurement. The estimates are fused in the next step by considering the correlation between them. An iterative algorithm to obtain the cross-covariance matrix between the estimation errors of the two Kalman filters is presented and employed in the fusion process. The improvement on the accuracy of estimation and comparison with other optimal fusion state estimation techniques are discussed through a simulation example, a pilot-scale experiment and an industrial case study.

The topic of control performance assessment techniques has drawn lots of attentions and many performance assessment indices have been proposed. These indices are focused on certain malfunctions. Fallacious decision would result in if it is based on individual indices. Therefore, fusion of different indices can improve the assessment accuracy. In this paper, Bayesian and Dempster-Shafer theory are used individually to establish decision fusion strategy and tackle the performance assessment challenge of heavy duty gas turbine. To study the uncertainty effect on these methods, they are applied to the well-known Rowen model of gas turbine. The results illustrate the effectiveness of the proposed performance assessment method of the gas turbine model on the one hand, and the superiority of the Bayesian method when the uncertainty is low and that of Dempster-Shafer theory in the presence of uncertainty, on the other hand.

The main goal of this paper is modelling attention while using it in efficient path planning of mobile robots. The key challenge in concurrently aiming these two goals is how to make an optimal, or near-optimal, decision in spite of time and processing power limitations, which inherently exist in a typical multi-sensor real-world robotic application. To efficiently recognise the environment under these two limitations, attention of an intelligent agent is controlled by employing the reinforcement learning framework. We propose an estimation method using estimated mixture-of-experts task and attention learning in perceptual space. An agent learns how to employ its sensory resources, and when to stop observing, by estimating its perceptual space. In this paper, static estimation of the state space in a learning task problem, which is examined in the WebotsTM simulator, is performed. Simulation results show that a robot learns how to achieve an optimal policy with a controlled cost by estimating the state space instead of continually updating sensory information.

In this contribution, full probability distribution of parameters of ARX model is obtained for on-line problems by means of Bayesian approach and Markov chain Monte Carlo method (MCMC), which provides the ability to be applied on time-varying ARX models as well. Full probability distribution of parameters represent whole available knowledge of parameters. So, decision makers can follow any policies to make decision about point estimation, like dynamic point estimation. Moreover, the Bayesian approach has great potential in combining sources of knowledge much more easier. To decrease the computational efforts, full probability of model parameters are updated based on size-varying partitions. Furthermore, incorporating the posterior probability of previous partition into the jump probability of current partition, in MCMC method, improves the performance of the proposed algorithm from the computation and convergence rate point of view. Simulation results demonstrate the effectiveness and validity of the proposed algorithm.

Gaussian process (GP) regression is a fully probabilistic method for performing non-linear regression. In a Bayesian framework, regression models can be made robust by using heavy-tailed distributions instead of using normal distribution for modeling noise. This work focuses on estimation of parameters for robust GP regression. In literature, these are learned by maximizing the approximate marginal likelihood of data. However, gradient-based optimization algorithms which are used for this purpose can be unstable or may require tuning. In this work, an EM algorithm based approach is derived and implemented to infer the parameters. The pros and cons of the two approaches are analyzed. The advantage of EM algorithm lies in its ease of implementation and theoretical guarantees of numerical stability and convergence while its prediction performance is still comparable to gradient-based approaches. In some cases EM algorithm may be slow to converge. To circumvent this issue a faster EM based approach known as Expectation Conjugate Gradient (ECG) is implemented on robust GP regression. Finally, the proposed EM approach to robust GP regression is validated using an industrial data set.

The main point of this paper is to present an iterative optimization strategy for tuning the parameters of Smith predictor based fractional order PID (SPFOPID) controller. The control scheme considered in this paper is the standard Smith predictor structure. Also, the internal model is considered to be a First Order Plus Dead Time (FOPDT) transfer function. Finally, the proposed method is implemented on a multi input-multi output (MIMO) flow-level plant and the obtained results are compared with the results of applying Smith predictor based PID controller (SPPID) in the similar structure.

A new approach for modeling and monitoring of the multivariate processes in presence of faulty and missing observations is introduced. It is assumed that operating modes of the process can transit to each other following a Markov chain model. Transition probabilities of the Markov chain are time varying as a function of the scheduling variable. Therefore, the transition probabilities will be able to vary adaptively according to different operating modes. In order to handle the problem of missing observations and unknown operating regimes, the expectation maximization algorithm is used to estimate the parameters. The proposed method is tested on two simulations and one industrial case studies. The industrial case study is the abnormal operating condition diagnosis in the primary separation vessel of oil-sand processes. In comparison to the conventional methods, the proposed method shows superior performance in detection of different operating conditions of the process.

This paper presents a neutral system approach to the design of an H∞ controller for input delay systems in presence of uncertain time-invariant delay. It is shown that when proportional derivative (PD) controller is applied to a time-delay system, the resulting closed loop system is generally a time-delay system of neutral type with delay term coefficients depending on the controller parameters. A descriptor model transformation is used to derive an advantageous bounded real lemma representation for the system. Furthermore, new delay-dependent sufficient conditions for the existence of an H∞ PD and PI controller in presence of uncertain delay are derived in terms of matrix inequalities. Some case studies and numerical examples are given in order to illustrate the advantages of the proposed method.

In this paper, a simple method is presented for tuning weighted PIλ + Dμcontroller parameters based on the pole placement controller of pseudo-second-order fractional systems. One of the advantages of this controller is capability of reducing the disturbance effects and improving response to input, simultaneously. In the following sections, the performance of this controller is evaluated experimentally to control the vertical magnetic flux in Damavand tokamak. For this work, at first a fractional order model is identified using output-error technique in time domain. For various practical experiments, having desired time responses for magnetic flux in Damavand tokamak, is vital. To approach this, at first the desired closed loop reference models are obtained based on generalized characteristic ratio assignment method in fractional order systems. After that, for the identified model, a set-point weighting PIλ + Dμcontroller is designed and simulated. Finally, this controller is implemented on digital signal processor control system of the plant to fast/slow control of magnetic flux. The practical results show appropriate performance of this controller.

In this paper, based on the nonmonotonic Lyapunov functions, a new less conservative state feedback controller synthesis method is proposed for a class of discrete time nonlinear systems represented by Takagi-Sugeno (T-S) fuzzy systems. Parallel distributed compensation (PDC) state feedback is employed as the controller structure. Also, a T-S fuzzy observer is designed in a manner similar to state feedback controller design. The observer and the controller can be obtained separately and then combined together to form an output feedback controller by means of the Separation theorem. Both observer and controller are obtained via solving a sequence of linear matrix inequalities. Nonmonotonic Lyapunov method allows the design of controllers for the aforementioned systems where other methods fail. Illustrative examples are presented which show how the proposed method outperforms other methods such as common quadratic, piecewise or non quadratic Lyapunov functions.

A multiple model structure of a prototype industrial gas turbine system is constructed under normal operation using a systematic method that incorporates non-linearity measure and H-gap metric tools with the multiple models technique. First, two new non-linearity indices for multiple input–multiple output systems are introduced and employed for decomposing the operating space of a gas turbine into some linear and non-linear modes. The non-linear modes may be further partitioned into some linear modes. The input and output data in each of the linear modes are used to construct an initial multiple model structure. In order to avoid the increase of the number of linear local models, the H-gap metric is extended to multiple input–multiple output systems and used to measure the similarity between linear local models and to merge the similar models. As a result, an algorithm is proposed for construction of multiple linear local models. The algorithm is employed for the identification of a single-shaft prototype industrial gas turbine.

Control of pH neutralization process has always been one of challenging problem in process control. The method presented here to control this process is the fuzzy identification of systems using Wiener model, and then multiplying the measured signal by the inverse of the nonlinear part of model. Therefore, we can design a linear controller for this new augmented system. This strategy is implemented in a generalized predictive control. One of the advantages of this control structure is consideration of explicit constraint in control of systems which also included in proposed fuzzy predictive control. At the end, the proposed method is tested on the model of a pH neutralization process.

This brief presents a X–Y pedestal using the feedback error learning (FEL) controller with adaptive neural network for low earth orbit (LEO) satellite tracking applications. The aim of the FEL is to derive the inverse dynamic model of the X–Y pedestal. In this brief, the kinematics of X–Y pedestal is obtained. To minimize or eliminate the backlash between gears, an antibacklash gearing system with dual-drive technique is used. The X–Y pedestal is implemented and the experimental results are obtained. They verify the obtained kinematics of the X–Y pedestal, its ability to minimize backlash, and the reduction of the tracking error for LEO satellite tracking in the typical NOAA19 weather satellite. Finally, the experimental results are plotted.

This paper concerns application of data-derived approaches for analyzing and monitoring chemical process instruments, extracting product information, and designing estimation models for primary process variables, or difficult to measure in real-time variables. Modeling of process with an optimized classical neural network, the multi-layer perceptron (MLP) is discussed. Tennessee Eastman Process, a well-known plant wide process benchmark, is applied to validate the proposed approach. Investigations and several algorithms as step response test, Lipschitz number method and forward selection are used. The main advancement introduced here is that a hierarchical level responsible strategy is applied for selection of input variables and respective efficient time delays to attain the highest possible prediction accuracy of the neural network model for industrial process identification.

In this paper input-output pairing is done based on concept of energy. Parseval theorem and cross-covariance samples of input-output are used for estimation of energy. After approximating interaction energy between input and output of the plant, input-output pairing is fulfilled. Through examples, it is illustrated that proposed method is appropriate for input-output pairing. The result is compared with Effective Relative Energy Array (EREA) as another energy based approach for input-output pairing.

In this paper Fault Detection and Isolation (FDI) is shown as a pattern classification problem which can be solved using clustering techniques. Gath-Geva clustering (GGC) is exploited as optimal form by a performance assessment rule for fault detection, while multistage Gath-Geva clustering is employed for the intent of fault isolation. Furthermore since Visbreaker unit is a large scale process, a novel hybrid method on the basis of Principle Component Analysis and Genetic Algorithm optimization was also proposed in order to cope with the curse of dimensionality and complexity of computation problems. There are two main percentile criteria for validation of fault detection namely specificity and sensitivity. Evaluation of fault isolation has been depicted in confusion matrix. For analysis and visualization of the correlated high dimensional data, PCA maps the data point into lower dimensional space. The proposed FDI approaches have been evaluated through experimental Visbreaker process unit data collected in oil refinery.

A combined approach for bumpless transfer multiple model predictive control (Multiple MPC) is proposed based on the Lyapunov function. State-space representation is used to design the controllers and the Lyapunov approach is employed to ensure closed loop stability. The proposed method uses both an intermediate controller and a bumpless mechanism in a unified configuration based on the stability analysis. Previous works on the bumpless multiple MPC design do not ensure closed loop stability, while the mechanism presented in this paper ensures both closed loop stability and applicable control performance for industrial processes. Finally, efficiency of the proposed method is validated by simulation results on non-isothermal continuous stirred-tank reactor (CSTR) system.

In this paper a new approach is proposed to design state feedback controllers for well known TS fuzzy systems. The used controller structure is in familiar parallel distributed compensation (PDC) format and the Lyapunov candidate is the frequently used common quadratic one. The only difference is the stability theorem used which is a revised version of Lyapunov stability theorem that relaxes monotonic condition. It is shown that the proposed method is less conservative than common quadratic method and piecewise method.

Input-Output data modeling using multi layer perceptron networks (MLP) for a laboratory helicopter is presented in this paper. The behavior of the two degree-of-freedom platform exemplifies a high order unstable, nonlinear system with significant cross-coupling between pitch and yaw directional motions. This paper develops a practical algorithm for identifying nonlinear autoregressive model with exogenous inputs (NARX) and nonlinear output error model (NOE) through closed loop identification. In order to collect input-output identifier pairs, a cascade state feedback (CSF) controller is introduced to stabilize the helicopter and after that the procedure of system identification is proposed. The estimated models can be utilized for nonlinear flight simulation and control and fault detection studies.

Expectation formation plays a principal role in economic systems. We examine and revise the standard rational expectations (RE) model, generally taken as the best paradigm for expectations modelling, and suggest a new method to model rational expectations. Conventional conditions that assert the stability and uniqueness of popular solution methods are shown to be insufficient. The agent-based new modelling approach suggested in this paper will be shown to lead to uniquely stable solutions.

Relative gain array (RGA) is the most important and popular method of finding the best pairing in MIMO plants. However, selection of the pairs based on RGA is an offline algorithm with some ambiguity. In this article, the normalised RGA (NRGA) matrix is introduced through the combination of the RGA matrix and pairing rules. The pairing problem can be interpreted as an assignment problem by using NRGA. Therefore, Hungarian algorithm can be applied to pair inputs and outputs of the process. Up to this stage, integrity has not yet been considered. If the determined optimal pairing does not satisfy integrity conditions based on the Niederlinski index, the procedure continues through the suboptimal pairings. The algorithm is fully systematic. It is applied for control structure selection of the well known Tennessee Eastman process plant.

In this paper, we introduce a vector which is able to describe the Niederlinski Index (NI), Relative Gain array (RGA), and the characteristic equation of the relative error matrix. The spectral radius and the structured singular value of the relative error matrix are investigated. The cases where the perfect result of the Relative Gain Array, equal to the identity matrix, coincides with the least interaction in a plant are pointed out. Then, the Jury Algorithm is adopted to get some insight into interaction analysis of multivariable plants. In particular, for interaction analysis of 3×3 plants, simple yet promising conditions in terms of the Relative Gain Array and the NiederLinski Index are derived. Several examples are also discussed to illustrate the main points.

A 3-DOF image stabilizer (periscope) is modeled and controlled, such that target is fixed in the center of camera image. Its nonlinear dynamic equations are extracted by applying the Euler-Lagrange equation of rigid body motion in presence of friction. The extracted equations are second order motion equations. Moreover, fuzzy sliding mode controller (FSMC) for target tracking is applied to this nonlinear system where it is subject to uncertainty and external disturbances. The controller is MIMO and it has knowledge-based structure. This method eliminates chattering that exists in the conventional sliding mode. The effectiveness of the developed algorithm is validated by simulation results.

Extreme nonlinearity and exhibition of severe interaction effects of multivariable pH processes makes it an appropriate test bed for evaluation of advanced controllers. This paper studies different multiple model methods for Generalized Predictive Control using Independent Model approach (GPCI) with adaptive weighting matrices. New method for adaptive determination of weighting matrices, proposed in this paper. Simulation results via typical multivariable pH process demonstrate the effectiveness and validity of the method. Different multiple model methods using adaptive weighting matrices compared with each other.

In this paper, a new approach for input-output pairing for stable and linear time invariant multivariable systems based on inputs-outputs correlation is introduced. Being independent from system's model is the characteristic of the proposed method. It is demonstrated that both static and dynamic properties of the system regarded in the proposed method. Through examples, the accuracy of the proposed approach is investigated. Finally, an example is used to show that in some cases Effective Relative Gain Array (ERGA) leads to improper pairs while the proposed method achieves the appropriate pairs.

In this paper, the relation between Input-output pairing and minimum variance (MV) index as a performance index is studied. Control structure selection or the input-output pairing problem is a key step in designing decentralized controllers for multivariable. The Relative Gain Array (RGA) is an important tool for the control structure selection procedure. In this study, RGA is calculated and decentralized minimum variance controllers are designed for each feasible pairing. The MV performance index will be calculated from the closed loop transfer function using the markov parameters. It is shown that the value of the MV index can propose an input-output pairing that leads to minimum output variance. Several simulation results are provided to show the main points of the paper.

A nonlinear model predictive control (NMPC) algorithm based on a neural network model is proposed for multivariable nonlinear systems. A multi-input multi-output model is developed using multilayer perceptron (MLP) neural network which is trained by Levenberg-Marquardt algorithm and amplitude modulated pseudo random binary (APRBS) signals with noise as data sets. Model predictive control also uses Levenberg-Marquardt algorithm for the control signal optimization. The control performance is improved by using a disturbance model that compensates both model mismatch and external disturbance. The learning rate of disturbance estimation network changes adaptively to treat the model mismatch differently from the external disturbance. Simulation results using the quadruple-tank are employed to show the effectiveness of the method.

In this paper we present a systematic procedure to design robust fuzzy controller for exponentially stabilizing affine nonlinear systems, based on their TS fuzzy model. For robust design we consider modeling error in TS model and as well as perturbation in the original nonlinear system. Minimization of cost function along with mapping closed loop poles to desired poles are considered simultaneously in controller design. As a result, the desired specified performance in transient response can be achieved. Piecewise Discontinues Lyapunov Functions (PDLF) are utilized in our proposed method. To avoid difficulties in boundary conditions in PDLF we opt to design an online controller and check the regions and boundaries continuously. The constraints required to guarantee the exponential stability of the original nonlinear systems and optimal controller design with guaranteeing desired performance are presented in the LMI form. The y well developed. The power of these methods is that searching Lyapunov function and feedback gain can be stated as a convex optimization problem and the task of finding the common Lyapunov function can be readily be formulated into an LMI problem. However this approach is too conservative and there are lots of stable systems that we can not find a common positive definite Lyapunov function for all subsystems. Piecewise quadratic Lyapunov function approach [7],[8] have been considered to avoid conservativeness of quadratic Lyapunov function approaches [4]-[6]. Piecewise quadratic Lyapunov function (PLF) are divided in two categories, one is continuous (PCLF) in boundaries and one of them is discontinuous (PDLF) on boundaries. It was shown that PDLF in contrast with PCLF results in fewer LMIs [9]. To apply all mentioned methods, the system must be presented by a Takagi-Sugeno model and as it was demonstrated TS modeling enables us to deal with high order complicated nonlinear systems. Most of works so far have used PCLF for controller design and stability analysis, but PDLF have been used mainly for stability analysis and there are no reports about using PDLF for controller design. The main reason is difficulties in boundary conditions. effectiveness and applicability of the proposed method is examined on an inverted pendulum system.

Rotary cement kiln is the main part of a cement plant that clinker is produced in it. Clinker is the main ingredient of cement. Continual and prolonged operation of rotary cement kiln is vital in cement factories. However, prolonged operation of the kiln is not possible and periodic repairs of the refractory lining would become necessary, due to non-linear phenomena existing in the kiln, such as sudden falls of coatings in the burning zone and probability of damages to the refractory materials during production. This is the basic reason behind the needs for a comprehensive model which is severely necessary for better control of this process. Such a model can be derived based on the mathematic analysis with consultation of expert operator experiences. In this paper both linear and nonlinear model are identified for rotary kiln of Saveh white cement factory. The linear model is introduced using Box-Jenkins structure. The results of the obtained model were satisfactory compared to some other linear models and can be used for designing adaptive or robust controllers. Also, nonlinear system identification via Neural Network technique is performed and its result was compared to linear models.

The CE 150 made by Humusoft is a laboratory helicopter designed for studying system dynamics and control engineering principles. This helicopter is nonlinear and unstable in vertical direction. In this paper a linear bank of models for modeling a vertical direction of helicopter has been obtained, and model validation is discussed.

In this paper, we use nonlinear system identification method to predict and detect process fault of a cement rotary kiln. After selecting proper inputs and output, an input-output model is identified for the plant. To identify the various operation points in the kiln, Locally Linear Neuro-Fuzzy (LLNF) model is used. This model is trained by LOLIMOT algorithm which is an incremental tree-structure algorithm. Then, by using this method, we obtained 3 distinct models for the normal and faulty situations in the kiln. One of the models is for normal condition of the kiln with 15 minutes prediction horizon. The other two models are for the two faulty situations in the kiln with 7 minutes prediction horizon are presented. At the end, we detect these faults in validation data. The data collected from White Saveh Cement Company is used for in this study.

Rotary cement kiln is the main part of a cement plant that clinker is produced in it. Continual and prolonged operation of rotary cement kiln is vital in cement factories. However, continual operation of the kiln is not possible and periodic repairs of the refractory lining would become necessary, due to non-linear phenomena existing in the kiln, such as sudden falls of coatings in the burning zone and probability of damages to the refractory materials during production. This is the basic reasoning behind the needs for a comprehensive model which is severely necessary for better control of this process. Such a model can be derived based on the mathematical analysis with consultation of expert operator experiences. In this paper linear model is identified for rotary kiln of Saveh white cement factory. The linear model is introduced using Box-Jenkins structure. The results of the obtained model were satisfactory compared to some other models and can be used for designing adaptive or robust controllers.

In this paper, two types of multiple-model adaptive controllers are practically evaluated on a laboratory-scale pH neutralization process. The first one is supervisory switching multiple-model adaptive controller (SMMAC) whose model bank is fixed and selected a priori, and another one is a controller based on multiple models, switching, and tuning strategy (MMST) which uses the possibility of model bank tuning. In addition to investigation of the effect of tuning, the advantage of a disturbance rejection supervisor is studied. Various experiments and exhaustive numerical analyses are provided to assess the abilities of the proposed algorithms.

In this paper, we design a neurofuzzy controller to control several variables of a rotary cement kilns. The variables are back-end temperature, pre-heater temperature, oxygen content and CO2 gas content of the kiln. The fuzzy control system, as an advanced control option for the kilns, is intended to minimize the operator interaction in the control process. The proposed fuzzy controller uses a neural network to optimize TSK-type fuzzy controller. Since there is no generally applicable analytical model for cement kilns, we use the real data derived from Saveh cement factory for the plant identification. A model, which is very similar to the real plant, is identified then; and the identified model is used for control design and simulations. Extensive simulation studies justify the effectiveness and applicability of the proposed control scheme in intelligent control of cement plant.

Inherent nonlinearity of pH processes causes that they are recognized as an appropriate test bench for evaluation of advanced controllers. Because of special characteristics of them, it is evident that adaptive controllers outperform others. This paper presents a comparison between a conventional adaptive controller and a switching multiple-model adaptive one in both regulation and disturbance rejection points of view. A disturbance rejection supervisor is designed to improve the performance of the adaptive controllers in the presence of unmeasured disturbances. A laboratory scale pH process is used as an application example.

Control based on multiple models (MM) is an effective strategy to cope with structural and parametric uncertainty of systems with highly nonlinear dynamics. It relies on a set of local models describing different operating modes of the system. Therefore, the performance is strongly depends on the distribution of the models in the defined operating space. In this paper, the problem of on-line construction of local model set is considered. The necessary specifications of an autonomous learning method are stated, and a high-level supervisor is designed to add an appropriate model to the available model set. The proposed algorithm is evaluated in a simulated pH neutralization process which is a highly nonlinear plant and composed of both abrupt and large continuous changes. The preference of the multiple-model approach with learning ability on a conventional 

Using the assignment technique of operations research, it is possible to obtain optimal controllable and observable pairs based on gramian measure for decentralised control of MIMO plants. The advantages of using assignment method for both open and closed loop performance in a MIMO plant are discussed in this paper.

Neural Network Model Predictive Control (NN-MPC) combines reliable prediction of neural network with excellent performance of model predictive control using nonlinear Levenberg-Marquardt optimization. It is shown that this structure is prone to steady-state error when external disturbances enter or actual system varies from its model. In this paper, these model uncertainties are taken into account using a disturbance model with iterative learning which adaptively change the learning rate to treat gradual effect of the model mismatch differently from the drastic changes of external disturbance. Then, a high-pass filter on error signal is designed to distinguish disturbances from model mismatches. Practical implementation results as well as simulation results demonstrate good performance of the proposed control method.

In this paper two robust controllers are designed for a practical Process Trainer Level plant. The system nonlinearity, time delay and change of parameters are the main problems in design of a desired controller for this plant. To design a controller, the linear models of the system and the disturbance models at different operating points are derived. Then, a parametric uncertainty profile is obtained by system identification strategies which is used in QFT control design. Indeed, for H∞ control design a multiplicative unstructured model is extracted from the parametric uncertainty. All constraints in control design, disturbance rejection and control signal are derived. Based on these constraints, appropriate controllers are determined. To improve robust performance μ-Synthesis with DK iteration is used. Finally all results are compared by applying the different controllers to the plant.

In this paper with reference to analytical results of different well-known relay feedback methods, we illustrate a main deficiency in parameter estimation of processes with a small ratio of time delay to time constant. Then to rectify this problem we introduce a modified relay feedback structure with additional delay to estimate the parameters of the FOPDT transfer function of the system. The significance of this method lies in the fact that many industrial plants perform fairly such as FOPDT systems, and a wide range of processes have negligible dead time versus their long constant time. Also, the estimated FOPDT transfer function from proposed relay feedback test can be used as a priori knowledge in advanced control strategies which need a FOPDT model of the system. The method is straightforward and simulation results illustrate the effectiveness, and simplicity of the proposed method.

A multiple-model adaptive controller is developed using the Self-Organizing Map (SOM) neural network. The considered controller which we name it as Multiple Controller via SOM (MCSOM) is evaluated on the pH neutralization plant. An improved switching algorithm based on excitation level of plant has also been suggested for systems with noisy environments. Identification of pH plant using SOM is discussed and performance of the multiple-model controller is compared to the Self Tuning Regulator (STR) controller.

Closed loop identification of nonlinear model and control of a laboratory helicopter using genetic algorithm is proposed in this paper. The derived model has a nonlinear structure. Using the previous results of the physical modeling of the studied plant, a nonlinear model is considered based on the physical dynamics of the system. However, there is no need to perform numerous physical experiments to estimate the model parameters. Instead, genetic algorithm as a nonlinear optimization technique is used to obtain the parameters of the model. Therefore, the advantage of both modeling and identification methods are employed. In the next step, the parameters of a multi input-multi output (MIMO) PID controller for the derived model will be tuned by GA using the obtained nonlinear model as a simulator of the plant. Applying the controller to both the real plant and the simulation model, the accuracy of the model and the performance of the controller is examined. The results demonstrate that the achieved model accurately fits to the behavior of the real plant and the controller designed based on this model, can control the real system appropriately.

Performance assessment and monitoring of control systems can be used to improve the performance of industrial processes. In this paper, a novel relay feedback based method for monitoring and automatic retuning of a class of proportional-integral (PI) controllers is proposed for the systems with gain nonlinearity. For performance assessment of the closed loop system, a time domain evaluation criteria based on the integral of the absolute value of the error (IAE) and the normalized pick of the error in setpoint (SP) changes are presented. Simulation results on the highly nonlinear pH process have shown the effectiveness and feasibility of this method.

Rotary Kiln is the central and the most complex components of cement production process. The first point at the beginning of the process, which is called back-end, is the calcining zone of the kiln and has a significant role on the quality of the clinker. In this paper to control the back-end temperature of a rotary kiln, we propose a fuzzy controller based on the operator's behaviors. We concentrate on how we can control the back-end temperature of a rotary kiln using a fuzzy controller. The work presented in this paper, is to use the advantages of fuzzy logic techniques to propose and develop a supervisory control system for the cement plant process having complex dynamics. The performance of the proposed controller is investigated by applying it to a simulator of the plant, derived from an MLP neural network model. Finally, the controller's behavior is evaluated in a disturbance environment. Simulation results show that the functionality of the controller causes a smooth behavior in both controller and model outputs.

Input/output pairing is an important task in control of a MIMO process by some SISO controllers. Relative gain array (RGA) is the most important method to find the best pairing. But selection of pairs based on RGA is an offline algorithm which needs some human decision. In this article, Normalized RGA (NRGA) matrix is introduced through the combination of the RGA matrix and its selection rules. Using NRGA, pairing problem can be interpreted as an assignment problem. The well known Hungarian algorithm is applied to the above problem to obtain the optimal pairing.

In this paper, two methods of designing controller for a practical MIMO Flow-Level pilot plant are achieved. Since the open loop plant is unstable, it is stabilized by using inner controller. Then the stabilized plant is identified at different operating points. A nominal model which is non-minimum phase is selected. This system has two outputs which have completely different time constants. Therefore, the identification problem is difficult. The other problem associated with the system is the time delay that is one of the characteristics of the process plants. The transfer function of system is close to triangular and the RGA matrix of plant is close to the identity matrix. Hence, decentralized control can be a good selection for this system which is used in our work. The robustness of the system with decentralized controller is also checked. Two Hinfin, robust controllers based on the knowledge of the disturbance are derived and the results are compared to the results of the decentralized one. To have a good robust performance the mu-synthesis and DK-iteration approach is used.

In this paper, a method for designing a stable fuzzy controller for nonlinear systems with the ability of reference tracking is introduced. First, the nonlinear equations of model helicopter is presented which is modeled by Humusoft Inc. based on linearization around system operating points, a TS fuzzy model is represented. Then, based on state feedback a TS fuzzy controller is developed, which its stability is guaranteed by nonlinear matrix Inequalities (LMI). These inequalities are solved using convex optimization methods by means of Matlab and LMI toolbox. In the following, tracking problem, control signal constraints are discussed.

This paper discusses application of an intelligent system in order to navigate in real-time a small size, four wheeled, indoor mobile robot accurately using ultra-light (160 gr), inexpensive laser range finder without prior information of the environment. A recurrent neural network is used to find the best path to the target of the robot. An accurate grid-based map is generated using a laser range finder scene and location found by a modified dead reckoning system. Finally a motion control method is presented. These approaches are implemented and tested in Resquake mobile robot

In this paper, different approaches to solve the forward kinematics of a three DOF actuator redundant hydraulic parallel manipulator are presented. On the contrary to series manipulators, the forward kinematic map of parallel manipulators involves highly coupled nonlinear equations, which are almost impossible to solve analytically. The proposed methods are using neural networks identification with different structures to solve the problem. The accuracy of the results of each method is analyzed in detail and the advantages and the disadvantages of them in computing the forward kinematic map of the given mechanism is discussed in detail. It is concluded that ANFIS presents the best performance compared to MLP, RBF and PNN networks in this particular application.

In this paper an H∞ controller is designed for a hydraulically actuated active suspension system of a half-modeled vehicle in a cascade feedback structure. Using the proposed structure the nonlinear behavior of actuator is reduced significantly. In the controller synthesis, a proportional controller is used in the inner loop, and a robust H∞ controller forms the outer loop. Two H∞ controllers are designed for this system. First unstructured uncertainty is not considered in the design procedure and secondly, the controller is designed considering uncertainty. Each of these controllers is designed in a decentralized fashion and the vehicle oscillation in the human sensitivity frequency range is reduced to a minimum. Statistical analysis of the simulation result using random input as road roughness, illustrates the effectiveness of the proposed control algorithm for both cases.

Multiple modeling identification using the selforganizing map neural network has been introduced by authors [1]. Two variations of that have been presented; MMSOM and MMISOM. MMSOM is based on using ordinary SOM and MMISOM utilizes the irregular SOM. In MMISOM, the neighborhood between the nodes may change. Therefore, MMISOM has more flexibility to cover concave spaces while SOM is more suitable for convex spaces. In this paper, after a review of both algorithms, some of the properties of MMISOM on the presence of noise are discussed.

…

A method for Multiple identification based on the self-organizing map neural networks is presented and some of its properties is investigated. Inputs to the NN are instantaneous parameters and so the reference vectors of the networks outputs are the parameter estimation of the multiple models.