Zeljko Durovic

Fundamentals of Stochastic Signals, Systems and Estimation Theory: With worked Examples

The main theme of this book deals with fundamental concepts underlying stochastic signal or linear stochastic systems, their modelling and analysis as well as model-based signal processing. Two popular stochastic models, the polynomial (or transfer function) model and the state space model are employed in schemes that lead to the estimation of unknown system parameters or states. The book is written for undergraduate and graduate students as well as practising engineers, specializing the the areas of electrical communications, signal processing and automatic control. Many examples illustrate the concepts of this book and the reader learns how to write software implementations of estimators on computers.

Recognition and classification of geometric shapes using neural networks

The research presented in this paper refers to classification of geometric shapes (cubes, pyramids and cylinders) using multilayer neural network. The input data of the algorithm are the images of shapes placed in different positions and distances from the camera. The classification is based on feature vectors that are obtained using methods of digital image processing. Feature vectors are inputs of neural network. Supervised training of neural network is performed. Reduction algorithm was used in aim of dimension reduction of feature vectors, so the classification results can be displayed graphically. Recognition and classification of geometric shapes may be of interest for realization of many robotic tasks, especially those related to catching of objects with robotic arm or movement of a robot with a set of obstacles.

Fault diagnosis for steam separators based on parameter identification and CUSUM classification

A method for diagnosing faults in steam separators is presented. Faults in the water level, water flow and steam flow sensors are analyzed. Precise models of the steam separator system are difficult to obtain, which makes the most common model-based fault detection and isolation approaches unapplicable. An identification-based method is used instead: parameters of the process are identified in real time, and the resulting data samples, which we denote as residuals, are used as inputs to a CUSUM-type classification scheme. It then decides if a fault is present, and if so, which one. In other words, residuals are first generated by parameter identification, and then evaluated by a modification of the CUSUM test. The choice of the CUSUM algorithm was motivated by its optimality with respect to detection delay. The identified parameters are assumed to be normally distributed. This assumption is experimentally verified: the true probability density functions (PDF) are estimated, and the performance of the detector based on these estimated PDFs is compared to that of the previous detector, based on the Gaussian PDF. The proposed method was tested on real-world data, obtained from the TEKO B1 Unit of the Kostolac Thermal Power Plant in Serbia. The results suggest extremely low probabilities of false alarm, missed detection and false isolation. As for detection delay, just one residual sample is needed for proper fault diagnosis in some cases, while 83 samples are needed in the worst-case scenario.

Control of Thermal Power Plant Combustion Distribution Using Extremum Seeking

High demands for increasing robustness, safety, and efficiency in thermal power plants are the main motivation behind ongoing attempts to optimize combustion. This paper presents a study of modeling and control of the combustion process in a tangentially fired pulverized-coal boiler. It proposes an approach to flame geometry and position control by means of reallocation of firing. Such control ensures flame focus maintenance away from the walls of the boiler, and thus prevents many unwanted by-products of combustion. In addition, uniform heat dissipation over mills enhances the energy efficiency and reliability of the boiler. First, experimental data obtained from the 350-MW boiler of the Nikola Tesla power plant, Serbia, are analyzed in detail. This results in a model identification procedure using an adaptive parameter estimation method. Second, constrained multivariate extremum seeking (ES) is proposed in this paper, to optimally tune boiler operation in order to maintain the desired flame configuration in the furnace. Finally, the effectiveness of the ES adaptive controller in the presence of disturbances is demonstrated through simulations performed on the experimentally identified model of the boiler.

Expert system based on hidden Markov models for recognition of radar targets

Design of an expert system based on Hidden Markov Models for recognition of radar targets in a zone of ground surveillance radar is presented in the paper. Parameters of the real radar echo signal represented in a form of autoregressive models are used as an input of the designed expert system. The real radar echoes have been collected for the purpose of this research. Obtained results show that designed system has some certain advantages, but there are also some limitations in recognition of the analyzed sequences.

Vessel Detection Algorithm Used in a Laser Monitoring System of the Lock Gate Zone

In this paper, we propose a vessel detection algorithm used in an online laser monitoring system in the lock of a hydropower plant. The system has to ensure the strict detection of the position of a vessel in order to prevent the manipulation of pound lock doors while the vessel is in the door zone. This paper describes in detail the monitoring concept, i.e., the detection algorithm implemented in the computer program that performs object detection in the view field of laser scanners. The detection algorithm has been developed in accordance with the modular principle and contains a number of functional partitions based on pattern recognition, i.e., the partition for the recognition of interference, the partition for water surface recognition in the conditions of debris floating on the water surface, the partition for the recognition of interference caused by the overflight of a single bird or a flock of birds, the partition for the recognition of interference caused by meteorological conditions, the partition for the recognition of high waves, and the partition for the recognition and detection of vessels. The main parts of the monitoring system are as follows: infrared laser scanners, controllers, an industrial PC, developed software with the implemented detection algorithm, a database, and a SCADA interface. This monitoring system has a vital role in keeping water transport operations safe and in the preventive maintenance and avoidance of vessel damaging in the area of gates at each end that controls the level of water in the lock chambers. The implementation of the detection algorithm has significantly improved the characteristics of the monitoring system. The system successfully detects all vessels, whereas the number of false detections remains neglectable.

Fault diagnosis in nonlinear stochastic systems via particle filtering

We consider the problem of detecting malfunctions in the actuators or sensors of systems which can be described by nonlinear/non-Gaussian stochastic state-space models. The basic idea is to estimate the state vector of such models using a sequential Monte Carlo technique known as particle filtering. We present several approaches to detecting faults and pinpointing their location within the system, using either one or a bank of particle filters.

Recognition and classification of deaf signs using neural networks

One approach for deaf signs recognition and classification is presented in the paper. It is assumed that the signs are presented in digital images. Recognition algorithm is consisted of several stages. At the beginning it is necessary to perform appropriate image processing in sense of segmentation and filtration of the input images. Aim is to detect arm position, i.e. sign of interest. For this purpose classifier for skin detection is used. Next stage has to generate feature vectors, which are used as inputs in neural network. Supervised training of neural network is performed. Reduction algorithm was used for purpose of dimension reduction of feature vectors, so the classification results can be displayed graphically.

Mobile robot control based on principles of Electrostatics

This paper presents the procedure of designing a control algorithm for differential driven mobile robot that is a typical example of a system with non-holonomic constraints. Mobile robot is controlled to reach an endpoint starting from random initial position in space. While moving to the endpoint robot is supposed to avoid any obstacle that is in its way. The designing procedure is based on Electrostatics theory. Robot and obstacles are supposed to be positively charged, while the endpoint is negatively charged. With this approach, mobile robot avoids detected obstacles and moves to the endpoint. Efficiency of proposed procedure is illustrated by detailed simulations.

Comparison of identification procedures in the frame of fault detection and isolation

Paper presents a model-based fault detection and isolation technique which relies on system identification procedure. Several open-loop identification methods are discussed since they are usually less complex and thus easier for implementation in large-scale systems. However, since the system analysis is usually conducted in a closed-loop structure, prediction error based direct closed-loop method was also taken into consideration. The efficiency of these four identification methods is demonstrated keeping in mind the final goal of the algorithm, which is the fault detection and isolation. Therefore, the performance of these methods in the light of FDI is emphasized, while the accuracy of parameter estimates is considered less important. The comparative analysis of suggested methods was done using data obtained from the steam separator in TEKO B1 unit of the Kostolac thermal power plant, Republic of Serbia.