This document discusses using neural networks to control the lateral skew of a railway wheel set on an experimental railway stand (roller rig). It first describes the roller rig setup and issues with previous control methods using state feedback and cascade PID control. It then discusses using neural networks for adaptive identification and control of the lateral skew. Specifically, it examines using linear neural units and quadratic neural units trained with real-time recurrent learning or backpropagation through time. The results of applying these various neural network approaches to identify and control the lateral skew of the roller rig are analyzed.
Identification and Real Time Control of a DC MotorIOSR Journals
This document presents a method for identifying and controlling a DC motor. It describes an experimental setup using sensors to measure the motor's speed. A parametric estimation method called PEM is used to identify the dynamic model of the motor based on input-output data. The identified model is validated via simulation. Then, an optimal linear quadratic regulator (LQR) controller is designed using the identified model to control the motor's speed and position. Simulations and experiments show the controller is able to track step and square wave position and velocity commands.
IRJET- Control of Induction Motor using Neural NetworkIRJET Journal
This document describes research into using neural networks to control induction motors. It begins by introducing the topic and noting limitations of traditional PI controllers for induction motor control. It then provides details on the experimental setup, which uses an artificial neural network (ANN) to mimic a PI controller for speed control of an induction motor drive system. Simulation results are presented comparing the performance of the ANN controller to a traditional PI controller under different dynamic operating conditions. The document concludes that the ANN mapping controller provides superior performance to the PI controller.
Mechatronics design of ball and beam system education and researchAlexander Decker
This document describes the proposed design of a mechatronics ball and beam system for educational and research purposes. The system aims to test and analyze various control strategies by using a servo motor to continuously adjust the angle of a beam to stabilize a rolling ball. It discusses the requirements, conceptual design, and concurrent selection and integration of subsystems including the mechanical design, sensors, actuators, and control algorithm. Preliminary block diagrams and component layouts are presented and the mechanical system parts are listed. The design is intended to help users visualize and experiment with different control methods on a precise and cost-effective experimental unit.
This document discusses different control methods for vehicle lateral control, including classical control theory, modern control theory, fuzzy logic, sliding mode control, and neural networks. It develops a 2DOF bicycle model of a vehicle and uses pole placement control to design a lateral controller. Simulation results show the vehicle can track a reference input but with large overshoots in yaw rate and velocity. An improved controller is designed with slower response but smaller state variable fluctuations. Future work involves implementing the controller with an observer and designing longitudinal control.
This document summarizes a paper presented at the International Conference on Mechatronics in Kumamoto, Japan in May 2007. The paper presents analysis and implementation of exact model knowledge and direct adaptive control schemes for a 4th order ball and beam system. Two controllers are designed - one using the exact model and one using direct adaptive control. Experimental results show that both controllers can track constant and sinusoidal references for the ball position asymptotically on a physical ball and beam system.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology.
Hardware-in-the-loop based comparative analysis of speed controllers for a tw...journalBEEI
A comparative study of speed control performance of an induction motor drive system connecting to a load via a non-rigid shaft. The nonrigidity of the coupling is represented by stiffness and damping coefficients deteriorating speed regulating operations of the system and can be regarded as a two-mass system. In the paper, the ability of flatness based and backstepping controls in control the two-mass system is verified through comprehensive hardware-in-the-loop experiments and with the assumption of ideal stator current loop performance. Step-by-step control design procedures are given, in addition, system responses with classical PID control are also provided for parallel comparisons.
Estimating the parameters of a geared DC motor is crucial in terms of its non-linear features. In this paper, parameters of a geared DC motor are estimated genetically. Mathematical model of the DC motor is determined by Kirchhoff’s law and dynamic model of its shafts and gearbox. Parameters of the geared DC motor are initially estimated by MATLAB/SIMULINK. The estimated parameters are defined as initial values for Genetic Algorithm (GA) to minimize the error of the simulated and actual angular trajectory captured by an encoder. The optimal estimated model of the geared DC motor is validated by different voltages as the input of the actual DC motor and its mathematical model. The results and numerical analysis illustrate it can be ascertained that GA is appropriate to estimate the parameters of platforms with non linear characteristics.
Identification and Real Time Control of a DC MotorIOSR Journals
This document presents a method for identifying and controlling a DC motor. It describes an experimental setup using sensors to measure the motor's speed. A parametric estimation method called PEM is used to identify the dynamic model of the motor based on input-output data. The identified model is validated via simulation. Then, an optimal linear quadratic regulator (LQR) controller is designed using the identified model to control the motor's speed and position. Simulations and experiments show the controller is able to track step and square wave position and velocity commands.
IRJET- Control of Induction Motor using Neural NetworkIRJET Journal
This document describes research into using neural networks to control induction motors. It begins by introducing the topic and noting limitations of traditional PI controllers for induction motor control. It then provides details on the experimental setup, which uses an artificial neural network (ANN) to mimic a PI controller for speed control of an induction motor drive system. Simulation results are presented comparing the performance of the ANN controller to a traditional PI controller under different dynamic operating conditions. The document concludes that the ANN mapping controller provides superior performance to the PI controller.
Mechatronics design of ball and beam system education and researchAlexander Decker
This document describes the proposed design of a mechatronics ball and beam system for educational and research purposes. The system aims to test and analyze various control strategies by using a servo motor to continuously adjust the angle of a beam to stabilize a rolling ball. It discusses the requirements, conceptual design, and concurrent selection and integration of subsystems including the mechanical design, sensors, actuators, and control algorithm. Preliminary block diagrams and component layouts are presented and the mechanical system parts are listed. The design is intended to help users visualize and experiment with different control methods on a precise and cost-effective experimental unit.
This document discusses different control methods for vehicle lateral control, including classical control theory, modern control theory, fuzzy logic, sliding mode control, and neural networks. It develops a 2DOF bicycle model of a vehicle and uses pole placement control to design a lateral controller. Simulation results show the vehicle can track a reference input but with large overshoots in yaw rate and velocity. An improved controller is designed with slower response but smaller state variable fluctuations. Future work involves implementing the controller with an observer and designing longitudinal control.
This document summarizes a paper presented at the International Conference on Mechatronics in Kumamoto, Japan in May 2007. The paper presents analysis and implementation of exact model knowledge and direct adaptive control schemes for a 4th order ball and beam system. Two controllers are designed - one using the exact model and one using direct adaptive control. Experimental results show that both controllers can track constant and sinusoidal references for the ball position asymptotically on a physical ball and beam system.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology.
Hardware-in-the-loop based comparative analysis of speed controllers for a tw...journalBEEI
A comparative study of speed control performance of an induction motor drive system connecting to a load via a non-rigid shaft. The nonrigidity of the coupling is represented by stiffness and damping coefficients deteriorating speed regulating operations of the system and can be regarded as a two-mass system. In the paper, the ability of flatness based and backstepping controls in control the two-mass system is verified through comprehensive hardware-in-the-loop experiments and with the assumption of ideal stator current loop performance. Step-by-step control design procedures are given, in addition, system responses with classical PID control are also provided for parallel comparisons.
Estimating the parameters of a geared DC motor is crucial in terms of its non-linear features. In this paper, parameters of a geared DC motor are estimated genetically. Mathematical model of the DC motor is determined by Kirchhoff’s law and dynamic model of its shafts and gearbox. Parameters of the geared DC motor are initially estimated by MATLAB/SIMULINK. The estimated parameters are defined as initial values for Genetic Algorithm (GA) to minimize the error of the simulated and actual angular trajectory captured by an encoder. The optimal estimated model of the geared DC motor is validated by different voltages as the input of the actual DC motor and its mathematical model. The results and numerical analysis illustrate it can be ascertained that GA is appropriate to estimate the parameters of platforms with non linear characteristics.
Impact analysis of actuator torque degradation on the IRB 120 robot performan...IJECEIAES
Actuators in a robot system may become faulty during their life cycle. Locked joints, free-moving joints, and the loss of actuator torque are common faulty types of robot joints where the actuators fail. Locked and free-moving joint issues are addressed by many published articles, whereas the actuator torque loss still opens attractive investigation challenges. The objectives of this study are to classify the loss of robot actuator torque, named actuator torque degradation, into three different cases: Boundary degradation of torque, boundary degradation of torque rate, and proportional degradation of torque, and to analyze their impact on the performance of a typical 6-DOF robot (i.e., the IRB 120 robot). Typically, controllers of robots are not pre-designed specifically for anticipating these faults. To isolate and focus on the impact of only actuator torque degradation faults, all robot parameters are assumed to be known precisely, and a popular closed-loop controller is used to investigate the robot’s responses under these faults. By exploiting MATLAB-the reliable simulation environment, a simscape-based quasi-physical model of the robot is built and utilized instead of an actual expensive prototype. The simulation results indicate that the robot responses cannot follow the desired path properly in most fault cases.
A NEW FUZZY LOGIC BASED SPACE VECTOR MODULATION APPROACH ON DIRECT TORQUE CON...csandit
The induction motors are indispensable motor types for industrial applications due to its wellknown
advantages. Therefore, many kind of control scheme are proposed for induction motors
over the past years and direct torque control has gained great importance inside of them due to
fast dynamic torque response behavior and simple control structure. This paper suggests a new
approach on the direct torque controlled induction motors, Fuzzy logic based space vector
modulation, to overcome disadvantages of conventional direct torque control like high torque
ripple. In the proposed approach, optimum switching states are calculated by fuzzy logic
controller and applied by space vector pulse width modulator to voltage source inverter. In
order to test and compare the proposed DTC scheme with conventional DTC scheme
simulations, in Matlab/Simulink, have been carried out in different speed and load conditions.
The simulation results showed that a significant improvement in the dynamic torque and speed
responses when compared to the conventional DTC scheme.
Adaptive Control of a Robotic Arm Using Neural Networks Based ApproachWaqas Tariq
A new neural networks and time series prediction based method has been discussed to control the complex nonlinear multi variable robotic arm motion system in 3d environment without engaging the complicated and voluminous dynamic equations of robotic arms in controller design stage, the proposed method gives such compatibility to the manipulator that it could have significant changes in its dynamic properties, like getting mechanical loads, without need to change designs of the controller.
Output feedback trajectory stabilization of the uncertainty DC servomechanism...ISA Interchange
This work proposes a solution for the output feedback trajectory-tracking problem in the case of an uncertain DC servomechanism system. The system consists of a pendulum actuated by a DC motor and subject to a time-varying bounded disturbance. The control law consists of a Proportional Derivative controller and an uncertain estimator that allows compensating the effects of the unknown bounded perturbation. Because the motor velocity state is not available from measurements, a second-order sliding-mode observer permits the estimation of this variable in finite time. This last feature allows applying the Separation Principle. The convergence analysis is carried out by means of the Lyapunov method. Results obtained from numerical simulations and experiments in a laboratory prototype show the performance of the closed loop system.
A NEW TOOL FOR LARGE SCALE POWER SYSTEM TRANSIENT SECURITY ASSESSMENTPower System Operation
This document proposes a new method for fast assessment of transient security power limits in large power systems using neural networks. It establishes a nonlinear mapping between transient energy margin and generator power at different fault clearing times and load levels using a self-organizing map. The transient security power limits of generators can then be estimated very quickly by inputting fault clearing time and load level data. Testing on a sample power system shows the proposed method can accurately estimate security limits without needing to calculate analytical sensitivities, providing faster results than traditional methods.
Velocity control of a two-wheeled inverted pendulum mobile robot: a fuzzy mod...journalBEEI
This paper presents the design of a fuzzy tracking controller for balancing and velocity control of a Two-Wheeled Inverted Pendulum (TWIP) mobile robot based on its Takagi-Sugino (T-S) fuzzy model, fuzzy Lyapunov function and non-parallel distributed compensation (non-PDC) control law. The T-S fuzzy model of the TWIP mobile robot was developed from its nonlinear dynamical equations of motion. Stabilization conditions in a form of linear matrix inequalities (LMIs) were derived based on the T-S fuzzy model of the TWIP mobile robot, a fuzzy Lyapunov function and a non-PDC control law. Based on the derived stabilization conditions and the T-S fuzzy model of the TWIP mobile robot, a state feedback velocity tracking controller was then proposed for the TWIP mobile robot. The balancing and velocity tracking performance of the proposed controller was investigated via simulations. The simulation result shows the effectiveness of the proposed control scheme.
Model predictive control of magnetic levitation system IJECEIAES
In this work, we suggest a technique of controller design that applied to systems based on nonlinear. We inform the sufficient conditions for the stability of closed loop system. The asymptotic stability of equilibrium and the nonlinear controller can be applied to improvement the stability of magnetic levitation system (MagLev). The MAgLev nonlinear nodel can be obtained by state equation based on Lagrange function and model predictive control has been used for MagLev system.
Comparison of different controllers for the improvement of Dynamic response o...IJERA Editor
This document compares different fuzzy logic controllers for improving the dynamic response of an indirect vector controlled induction motor drive. It presents a new fuzzy PI controller with scaling factors and evaluates its performance against fuzzy PI and fuzzy MRAC (model reference adaptive control) controllers. Simulation results show that the proposed fuzzy PI with scaling factors has a faster settling time than fuzzy PI, and is less complex than fuzzy MRAC while still providing good parameter insensitivity. The proposed controller provides a compromise between complexity, accuracy and settling time for induction motor applications.
The Neural Network-Combined Optimal Control System of Induction MotorIJECEIAES
This research aims to propose the optimal control method combined with the neuron network for an induction motor. In the proposed system, the induction motor is a nonlinear object which is controlled at each working point. At these working-points, the state equation of the induction motor is linear, so it is possible to apply the linear quadratic regular algorithm for the induction motor. Therefore, the parameters of the state feedback controller are the functions. The output-input relationships of these functions are set through the neural network. The numerical simulation results show that the quality of the control system of the induction motor is very high: The response speed always follows the desired speed with the short transition time and the small overshoot. Furthermore, the system is robust in the case of changing the load torque, and the parameters of the induction motor are incorrectly defined
PID Controller Design for a Real Time Ball and Beam System – A Double Integra...idescitation
In this paper, the authors have discussed and shown
how to tune the PID controller in closed loop with time-delay
for the double integrator systems for a particular stability
margins. In math model it is assumed that time delay (ô) of
the plant is known. As a case study the authors have consid-
ered the mathematical model of the real-time beam and ball
system and analyzed the simulation and real time response.
This document presents an output-based input shaping and proportional integral derivative (PID) control for load hoisting control of a 3D crane system. Unlike conventional input shaping that uses model parameters, output-based filtering is designed using the target system's output signal to avoid issues from model uncertainties. Simulation results show precise payload positioning with negligible sway is achieved. The proposed hybrid control is robust and can be easily implemented on higher order systems. Logarithmic decrement techniques are used to determine the crane system's damping ratio and natural frequency, simplifying the design. An output-based filter is designed by minimizing the difference between the target system and reference system outputs. Filter gains are then calculated to shape the input and reduce residual vibrations
This document discusses the theoretical and experimental study of dynamics and control of a two-link flexible robot arm. It presents the following:
1) A linear dynamic model of the robot arm is constructed using Ritz's approach and validated experimentally.
2) Three active control schemes are designed using pole placement technique for path tracking and vibration suppression, and tested via simulation.
3) A discrete-time state variable feedback control scheme is presented and designed to control the robot arm's motion using three sensors for position and curvature feedback.
Model Validation and Control of an In-Wheel DC Motor Prototype for Hybrid El...Scientific Review SR
In this paper, a mathematical model and a controller for a DC motor are developed for the
construction of an in-wheel motor. In-wheel motors can be used in hybrid electric vehicles to provide traction
force of front or rear wheels. The model identification is achieved using a simple and low cost data acquisition
system. An Arduino Uno embedded board system is used to collect data from sensors to a computer and for
control purposes. Data processing is performed using Matlab/Simulink. Validations of the devel oped
mathematical model and controller performance are carried out by comparing simulation and experimental results.
The results obtained show that the mathematical model is accurate enough to assist in speed controller design and
implementation.
This document presents a method for tuning the parameters of a PID controller for a brushless DC motor using particle swarm optimization. PSO is used to find the optimal proportional, integral and derivative gains to minimize error metrics for the motor's step response. The BLDC motor is modeled in Simulink. PSO searches through potential solutions in a multi-dimensional space to determine PID parameters that produce the best step response with minimal overshoot, rise time, settling time and steady state error. The results show the PSO-tuned PID controller achieves better dynamic performance than other methods.
—This paper presents a new image based visual servoing (IBVS) control scheme for omnidirectional wheeled mobile robots with four swedish wheels. The contribution is the proposal of a scheme that consider the overall dynamic of the system; this means, we put together mechanical and electrical dynamics. The actuators are direct current (DC) motors, which imply that the system input signals are armature voltage applied to DC motors. In our control scheme the PD control law and eye-to-hand camera configuration are used to compute the armature voltages and to measure system states, respectively. Stability proof is performed via Lypunov direct method and LaSalle's invariance principle. Simulation and experimental results were performed in order to validate the theoretical proposal and to show the good performance of the posture errors. Keywords—IBVS, posture control, omnidirectional wheeled mobile robot, dynamic actuator, Lyapunov direct method.
Speed and Torque Control of Mechanically Coupled Permanent Magnet Direct Curr...IDES Editor
A new controller is designed for speed and torque
control of a Permanent Magnet DC motor based on
measurements of speed and current. This research work
focuses on investigating the effects of control of the speed and
torque of two brushless dc motors that are mechanically
coupled. Two controller design methods: the Root Locus
method and Bode Plot method as well as two controllers:
Proportional-Integral-Derivative (PID) and Proportional-
Integral (PI) are used to obtain the control objectives of speed
control and torque control. The simulation is performed using
MATLAB/SIMULINK software. The effects of varying the
controller gains on the system performance is studied and
quantified. The simulation results show that the speed control
objectives of the motor are satisfied even in the case of torque
disturbance from the other motor.
In this paper, the tracking control scheme is presented using the framework of finite-time sliding mode control (SMC) law and high-gain observer for disturbed/uncertain multi-motor driving systems under the consideration multi-output systems. The convergence time of sliding mode control is estimated in connection with linear matrix inequalities (LMIs). The input state stability (ISS) of proposed controller was analyzed by Lyapunov stability theory. Finally, the extensive simulation results are given to validate the advantages of proposed control design.
An Unmanned Rotorcraft System with Embedded DesignIOSR Journals
This document describes the design and implementation of an embedded control system for an unmanned rotorcraft. It uses a backstepping control law with state estimation provided by an extended Kalman filter using data from a marker-based vision system and inertial measurement unit. A self-organizing map is used to select local operating regimes for modeling the nonlinear dynamics. The control algorithm was tested in a hardware-in-the-loop simulation and able to track spiral paths in 3D space with the embedded controller implemented on a Freescale MPC555 processor.
In this paper, the artificial neural network (ANN) has been utilized for rotating machinery faults detection and classification. First, experiments were performed to measure the lateral vibration signals of laboratory test rigs for rotor-disk-blade when the blades are defective. A rotor-disk-blade system with 6 regular blades and 5 blades with various defects was constructed. Second, the ANN was applied to classify the different x- and y-axis lateral vibrations due to different blade faults. The results based on training and testing with different data samples of the fault types indicate that the ANN is robust and can effectively identify and distinguish different blade faults caused by lateral vibrations in a rotor. As compared to the literature, the present paper presents a novel work of identifying and classifying various rotating blade faults commonly encountered in rotating machines using ANN. Experimental data of lateral vibrations of the rotor-disk-blade system in both x- and y-directions are used for the training and testing of the network.
Power Optimization of Battery Charging System Using FPGA Based Neural Network...IJERA Editor
This paper involves designing a small scale battery charging system which is powered via a photovoltaic panel. This work aims at the usage of solar energy for charging the battery and optimizing the power of the system. Implementation is done using Artificial Neural Network (ANN) on FPGA. To develop this system an Artificial Neural Network is trained and its result is further used for the PWM technique. PWM pulse generation has been done using Papilio board which is based on XILINX Spartan 3E FPGA. The ANN with PWM technique is ported on FPGA which is programmed using VHDL. This able to automatically control the whole charging system operation without requirement of external sensory unit. The simulation results are achieved by using MATLAB and XILINX. These results allowed demonstrating the charging of the battery using proposed ANN and PWM technique.
Application of artificial neural network for blast performance evaluationeSAT Publishing House
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
Language Identification: A neural network approachAlberto Simões
This document describes a neural network approach for language identification. It discusses extracting features from text such as alphabet characters and character sequences (unigrams, bigrams, trigrams) that are common in different languages. Training data is prepared from texts of over 105 languages on the TED website, with out-of-vocabulary words removed. The neural network architecture has an input layer for features, hidden layers, and an output layer for language predictions. Alphabet features count Unicode character classes and are binarized. Trigrams are used as sequence features to aid comparisons between languages.
Impact analysis of actuator torque degradation on the IRB 120 robot performan...IJECEIAES
Actuators in a robot system may become faulty during their life cycle. Locked joints, free-moving joints, and the loss of actuator torque are common faulty types of robot joints where the actuators fail. Locked and free-moving joint issues are addressed by many published articles, whereas the actuator torque loss still opens attractive investigation challenges. The objectives of this study are to classify the loss of robot actuator torque, named actuator torque degradation, into three different cases: Boundary degradation of torque, boundary degradation of torque rate, and proportional degradation of torque, and to analyze their impact on the performance of a typical 6-DOF robot (i.e., the IRB 120 robot). Typically, controllers of robots are not pre-designed specifically for anticipating these faults. To isolate and focus on the impact of only actuator torque degradation faults, all robot parameters are assumed to be known precisely, and a popular closed-loop controller is used to investigate the robot’s responses under these faults. By exploiting MATLAB-the reliable simulation environment, a simscape-based quasi-physical model of the robot is built and utilized instead of an actual expensive prototype. The simulation results indicate that the robot responses cannot follow the desired path properly in most fault cases.
A NEW FUZZY LOGIC BASED SPACE VECTOR MODULATION APPROACH ON DIRECT TORQUE CON...csandit
The induction motors are indispensable motor types for industrial applications due to its wellknown
advantages. Therefore, many kind of control scheme are proposed for induction motors
over the past years and direct torque control has gained great importance inside of them due to
fast dynamic torque response behavior and simple control structure. This paper suggests a new
approach on the direct torque controlled induction motors, Fuzzy logic based space vector
modulation, to overcome disadvantages of conventional direct torque control like high torque
ripple. In the proposed approach, optimum switching states are calculated by fuzzy logic
controller and applied by space vector pulse width modulator to voltage source inverter. In
order to test and compare the proposed DTC scheme with conventional DTC scheme
simulations, in Matlab/Simulink, have been carried out in different speed and load conditions.
The simulation results showed that a significant improvement in the dynamic torque and speed
responses when compared to the conventional DTC scheme.
Adaptive Control of a Robotic Arm Using Neural Networks Based ApproachWaqas Tariq
A new neural networks and time series prediction based method has been discussed to control the complex nonlinear multi variable robotic arm motion system in 3d environment without engaging the complicated and voluminous dynamic equations of robotic arms in controller design stage, the proposed method gives such compatibility to the manipulator that it could have significant changes in its dynamic properties, like getting mechanical loads, without need to change designs of the controller.
Output feedback trajectory stabilization of the uncertainty DC servomechanism...ISA Interchange
This work proposes a solution for the output feedback trajectory-tracking problem in the case of an uncertain DC servomechanism system. The system consists of a pendulum actuated by a DC motor and subject to a time-varying bounded disturbance. The control law consists of a Proportional Derivative controller and an uncertain estimator that allows compensating the effects of the unknown bounded perturbation. Because the motor velocity state is not available from measurements, a second-order sliding-mode observer permits the estimation of this variable in finite time. This last feature allows applying the Separation Principle. The convergence analysis is carried out by means of the Lyapunov method. Results obtained from numerical simulations and experiments in a laboratory prototype show the performance of the closed loop system.
A NEW TOOL FOR LARGE SCALE POWER SYSTEM TRANSIENT SECURITY ASSESSMENTPower System Operation
This document proposes a new method for fast assessment of transient security power limits in large power systems using neural networks. It establishes a nonlinear mapping between transient energy margin and generator power at different fault clearing times and load levels using a self-organizing map. The transient security power limits of generators can then be estimated very quickly by inputting fault clearing time and load level data. Testing on a sample power system shows the proposed method can accurately estimate security limits without needing to calculate analytical sensitivities, providing faster results than traditional methods.
Velocity control of a two-wheeled inverted pendulum mobile robot: a fuzzy mod...journalBEEI
This paper presents the design of a fuzzy tracking controller for balancing and velocity control of a Two-Wheeled Inverted Pendulum (TWIP) mobile robot based on its Takagi-Sugino (T-S) fuzzy model, fuzzy Lyapunov function and non-parallel distributed compensation (non-PDC) control law. The T-S fuzzy model of the TWIP mobile robot was developed from its nonlinear dynamical equations of motion. Stabilization conditions in a form of linear matrix inequalities (LMIs) were derived based on the T-S fuzzy model of the TWIP mobile robot, a fuzzy Lyapunov function and a non-PDC control law. Based on the derived stabilization conditions and the T-S fuzzy model of the TWIP mobile robot, a state feedback velocity tracking controller was then proposed for the TWIP mobile robot. The balancing and velocity tracking performance of the proposed controller was investigated via simulations. The simulation result shows the effectiveness of the proposed control scheme.
Model predictive control of magnetic levitation system IJECEIAES
In this work, we suggest a technique of controller design that applied to systems based on nonlinear. We inform the sufficient conditions for the stability of closed loop system. The asymptotic stability of equilibrium and the nonlinear controller can be applied to improvement the stability of magnetic levitation system (MagLev). The MAgLev nonlinear nodel can be obtained by state equation based on Lagrange function and model predictive control has been used for MagLev system.
Comparison of different controllers for the improvement of Dynamic response o...IJERA Editor
This document compares different fuzzy logic controllers for improving the dynamic response of an indirect vector controlled induction motor drive. It presents a new fuzzy PI controller with scaling factors and evaluates its performance against fuzzy PI and fuzzy MRAC (model reference adaptive control) controllers. Simulation results show that the proposed fuzzy PI with scaling factors has a faster settling time than fuzzy PI, and is less complex than fuzzy MRAC while still providing good parameter insensitivity. The proposed controller provides a compromise between complexity, accuracy and settling time for induction motor applications.
The Neural Network-Combined Optimal Control System of Induction MotorIJECEIAES
This research aims to propose the optimal control method combined with the neuron network for an induction motor. In the proposed system, the induction motor is a nonlinear object which is controlled at each working point. At these working-points, the state equation of the induction motor is linear, so it is possible to apply the linear quadratic regular algorithm for the induction motor. Therefore, the parameters of the state feedback controller are the functions. The output-input relationships of these functions are set through the neural network. The numerical simulation results show that the quality of the control system of the induction motor is very high: The response speed always follows the desired speed with the short transition time and the small overshoot. Furthermore, the system is robust in the case of changing the load torque, and the parameters of the induction motor are incorrectly defined
PID Controller Design for a Real Time Ball and Beam System – A Double Integra...idescitation
In this paper, the authors have discussed and shown
how to tune the PID controller in closed loop with time-delay
for the double integrator systems for a particular stability
margins. In math model it is assumed that time delay (ô) of
the plant is known. As a case study the authors have consid-
ered the mathematical model of the real-time beam and ball
system and analyzed the simulation and real time response.
This document presents an output-based input shaping and proportional integral derivative (PID) control for load hoisting control of a 3D crane system. Unlike conventional input shaping that uses model parameters, output-based filtering is designed using the target system's output signal to avoid issues from model uncertainties. Simulation results show precise payload positioning with negligible sway is achieved. The proposed hybrid control is robust and can be easily implemented on higher order systems. Logarithmic decrement techniques are used to determine the crane system's damping ratio and natural frequency, simplifying the design. An output-based filter is designed by minimizing the difference between the target system and reference system outputs. Filter gains are then calculated to shape the input and reduce residual vibrations
This document discusses the theoretical and experimental study of dynamics and control of a two-link flexible robot arm. It presents the following:
1) A linear dynamic model of the robot arm is constructed using Ritz's approach and validated experimentally.
2) Three active control schemes are designed using pole placement technique for path tracking and vibration suppression, and tested via simulation.
3) A discrete-time state variable feedback control scheme is presented and designed to control the robot arm's motion using three sensors for position and curvature feedback.
Model Validation and Control of an In-Wheel DC Motor Prototype for Hybrid El...Scientific Review SR
In this paper, a mathematical model and a controller for a DC motor are developed for the
construction of an in-wheel motor. In-wheel motors can be used in hybrid electric vehicles to provide traction
force of front or rear wheels. The model identification is achieved using a simple and low cost data acquisition
system. An Arduino Uno embedded board system is used to collect data from sensors to a computer and for
control purposes. Data processing is performed using Matlab/Simulink. Validations of the devel oped
mathematical model and controller performance are carried out by comparing simulation and experimental results.
The results obtained show that the mathematical model is accurate enough to assist in speed controller design and
implementation.
This document presents a method for tuning the parameters of a PID controller for a brushless DC motor using particle swarm optimization. PSO is used to find the optimal proportional, integral and derivative gains to minimize error metrics for the motor's step response. The BLDC motor is modeled in Simulink. PSO searches through potential solutions in a multi-dimensional space to determine PID parameters that produce the best step response with minimal overshoot, rise time, settling time and steady state error. The results show the PSO-tuned PID controller achieves better dynamic performance than other methods.
—This paper presents a new image based visual servoing (IBVS) control scheme for omnidirectional wheeled mobile robots with four swedish wheels. The contribution is the proposal of a scheme that consider the overall dynamic of the system; this means, we put together mechanical and electrical dynamics. The actuators are direct current (DC) motors, which imply that the system input signals are armature voltage applied to DC motors. In our control scheme the PD control law and eye-to-hand camera configuration are used to compute the armature voltages and to measure system states, respectively. Stability proof is performed via Lypunov direct method and LaSalle's invariance principle. Simulation and experimental results were performed in order to validate the theoretical proposal and to show the good performance of the posture errors. Keywords—IBVS, posture control, omnidirectional wheeled mobile robot, dynamic actuator, Lyapunov direct method.
Speed and Torque Control of Mechanically Coupled Permanent Magnet Direct Curr...IDES Editor
A new controller is designed for speed and torque
control of a Permanent Magnet DC motor based on
measurements of speed and current. This research work
focuses on investigating the effects of control of the speed and
torque of two brushless dc motors that are mechanically
coupled. Two controller design methods: the Root Locus
method and Bode Plot method as well as two controllers:
Proportional-Integral-Derivative (PID) and Proportional-
Integral (PI) are used to obtain the control objectives of speed
control and torque control. The simulation is performed using
MATLAB/SIMULINK software. The effects of varying the
controller gains on the system performance is studied and
quantified. The simulation results show that the speed control
objectives of the motor are satisfied even in the case of torque
disturbance from the other motor.
In this paper, the tracking control scheme is presented using the framework of finite-time sliding mode control (SMC) law and high-gain observer for disturbed/uncertain multi-motor driving systems under the consideration multi-output systems. The convergence time of sliding mode control is estimated in connection with linear matrix inequalities (LMIs). The input state stability (ISS) of proposed controller was analyzed by Lyapunov stability theory. Finally, the extensive simulation results are given to validate the advantages of proposed control design.
An Unmanned Rotorcraft System with Embedded DesignIOSR Journals
This document describes the design and implementation of an embedded control system for an unmanned rotorcraft. It uses a backstepping control law with state estimation provided by an extended Kalman filter using data from a marker-based vision system and inertial measurement unit. A self-organizing map is used to select local operating regimes for modeling the nonlinear dynamics. The control algorithm was tested in a hardware-in-the-loop simulation and able to track spiral paths in 3D space with the embedded controller implemented on a Freescale MPC555 processor.
In this paper, the artificial neural network (ANN) has been utilized for rotating machinery faults detection and classification. First, experiments were performed to measure the lateral vibration signals of laboratory test rigs for rotor-disk-blade when the blades are defective. A rotor-disk-blade system with 6 regular blades and 5 blades with various defects was constructed. Second, the ANN was applied to classify the different x- and y-axis lateral vibrations due to different blade faults. The results based on training and testing with different data samples of the fault types indicate that the ANN is robust and can effectively identify and distinguish different blade faults caused by lateral vibrations in a rotor. As compared to the literature, the present paper presents a novel work of identifying and classifying various rotating blade faults commonly encountered in rotating machines using ANN. Experimental data of lateral vibrations of the rotor-disk-blade system in both x- and y-directions are used for the training and testing of the network.
Power Optimization of Battery Charging System Using FPGA Based Neural Network...IJERA Editor
This paper involves designing a small scale battery charging system which is powered via a photovoltaic panel. This work aims at the usage of solar energy for charging the battery and optimizing the power of the system. Implementation is done using Artificial Neural Network (ANN) on FPGA. To develop this system an Artificial Neural Network is trained and its result is further used for the PWM technique. PWM pulse generation has been done using Papilio board which is based on XILINX Spartan 3E FPGA. The ANN with PWM technique is ported on FPGA which is programmed using VHDL. This able to automatically control the whole charging system operation without requirement of external sensory unit. The simulation results are achieved by using MATLAB and XILINX. These results allowed demonstrating the charging of the battery using proposed ANN and PWM technique.
Application of artificial neural network for blast performance evaluationeSAT Publishing House
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
Language Identification: A neural network approachAlberto Simões
This document describes a neural network approach for language identification. It discusses extracting features from text such as alphabet characters and character sequences (unigrams, bigrams, trigrams) that are common in different languages. Training data is prepared from texts of over 105 languages on the TED website, with out-of-vocabulary words removed. The neural network architecture has an input layer for features, hidden layers, and an output layer for language predictions. Alphabet features count Unicode character classes and are binarized. Trigrams are used as sequence features to aid comparisons between languages.
The document discusses anti-lock braking systems (ABS) and how neural networks can be used to control ABS. It first defines ABS and explains how it works to prevent wheel lockup during braking by monitoring wheel speed and modulating brake pressure. It then discusses types of ABS and the components of an ABS system. The document proposes using a neural network-based hybrid controller to improve ABS performance over different road conditions by adapting to changes in friction. It concludes that ABS improves vehicle control and reduces stopping distances compared to non-ABS systems.
Prediction of Exchange Rate Using Deep Neural Network Tomoki Hayashi
This document proposes using a deep neural network to predict currency exchange rates. It discusses using DNN to directly predict future exchange rates or to perform binary classification to predict if the rate will increase or decrease. The model takes in features like past exchange rates, moving averages and volatility indicators. Experiments show the model can predict trend transitions with over 75% accuracy on closed tests and over 60% on open tests by classifying trend direction changes. Pre-training is done using restricted Boltzmann machines to initialize weights before fine-tuning with backpropagation.
This document describes a robust learning control scheme for tank gun control servo systems. The control scheme uses iterative learning control and adaptive control techniques to achieve high-precision velocity tracking for the tank gun despite nonlinear uncertainties and external disturbances. Lyapunov stability analysis is used to design the learning controller. The unknown parameters are estimated using a full saturation difference learning strategy. Simulation results show that as the number of iterations increases, the system state is able to accurately track the reference signal over the entire time interval while all signals remain bounded.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
Towards a good abs design for more Reliable vehicles on the roadsijcsit
This document summarizes a research paper that proposes a design approach for antilock braking systems (ABS) using Stopwatch Petri Nets (SWPN) to model the functional and dysfunctional behavior. SWPN allow modeling of interruptible systems by representing the suspension and resumption of tasks. The proposed approach extracts feared scenarios, which could lead to dangerous states, directly from the SWPN model without generating the reachability graph. The method involves identifying normal states, target states like feared states, then performing backward and forward reasoning on the SWPN model to determine event sequences and contexts that could result in feared states. This helps identify failures during the ABS design phase to improve reliability.
Neural Network Control Based on Adaptive Observer for Quadrotor HelicopterIJITCA Journal
A neural network control scheme with an adaptive observer is proposed in this paper to Quadrotor helicopter stabilization. The unknown part in Quadrotor dynamical model was estimated on line by a
Single Hidden Layer network. To solve the non measurable states problem a new adaptive observer was proposed. The main purpose here is to reduce the measurement noise amplification caused by conventional
high gain observer by introducing some changes in observer’s original structure that can minimize the variance and the amplitude of the noisy signal without increasing tracking error. The stability analysis of the overall closed-loop system/ observer is performed using the Lyapunov direct method. Simulation results are given to highlight the performances of the proposed scheme.
This work treats the modeling and simulation of non-linear system behavior of an induction motor using backstepping sliding mode control (BACK- SMC). First, the direct field oriented control IM is derived. Then, a sliding for direct field oriented control is proposed to compensate the uncertainties, which occur in the control. Finally, the study of Backstepping sliding controls strategy of the induction motor drive. Our non linear system is simulated in MATLAB SIMULINK environment, the results obtained illustrate the efficiency of the proposed control with no overshoot, and the rising time is improved with good disturbances rejections comparing with the classical control law.
This paper describes the design and the simulation of a non-linear controller for two-mass system using induction motor basing on the backstepping method. The aim is to control the speed actual value of load motor matching with the speed reference load motor, moreover, electrical drive’s respone ensuring the “fast, accurate and small overshoot” and reducing the resonance oscillations for two-mass system using induction motor fed by voltage source inveter with ideally control performance of stator current. Backstepping controller uses the non-linear equations of an induction motor and the linear dynamical equations of two-mass system, the Lyapunov analysis and the errors between the real and the desired values. The controller has been implemented in both simulation and hardware-in-the-loop (HIL) real-time experiments using Typhoon HIL 402 system, when the drive system operates at a stable speed (rotor flux is constant) and greater than rated speed (field weakening area). The simulation and HIL results presented the correctness and effectiveness of the controller is proposed; furthermore, compared to PI method to see the response of the system clearly.
Abstract - This paper addresses some of the potential benefits of
using fuzzy logic controllers to control an inverted pendulum
system. The stages of the development of a fuzzy logic controller
using a four input Takagi-Sugeno fuzzy model were presented.
The main idea of this paper is to implement and optimize fuzzy
logic control algorithms in order to balance the inverted
pendulum and at the same time reducing the computational time
of the controller. In this work, the inverted pendulum system was
modeled and constructed using Simulink and the performance of
the proposed fuzzy logic controller is compared to the more
commonly used PID controller through simulations using Matlab.
Simulation results show that the Fuzzy Logic Controllers are far
more superior compared to PID controllers in terms of overshoot,
settling time and response to parameter changes.
In this paper, DTC is applied for two-level inverter fed IM drives based on neuronal hysteresis comparators and The Direct Torque Control (DTC) is known to produce quick and robust response in AC drive system. However, during steady state, torque, flux and current ripple. An improvement of electric drive system can be obtained using a DTC method based on ANNs which reduces the torque and flux ripples, the estimated the rotor speed using the KUBOTA observer method based on measurements of electrical quantities of the motor. The validity of the proposed methods is confirmed by the simulation results.The THD (Total Harmonic Distortion) of stator current, torque ripple and stator flux ripple are determined and compared with conventional DTC control scheme using Matlab/Simulink environment.
This document introduces the fuzzy model reference learning control (FMRLC) method. FMRLC uses a reference model to provide feedback to modify the membership functions of a fuzzy controller. This allows the closed-loop system to behave like the reference model and achieve the desired performance. The effectiveness of FMRLC is demonstrated through its application to rocket velocity control and robot manipulator control. FMRLC can achieve high performance learning control for nonlinear, time-varying systems.
Experimental verification of SMC with moving switching lines applied to hoisti...ISA Interchange
In this paper we propose sliding mode control strategies for the point-to-point motion control of a hoisting crane. The strategies employ time-varying switching lines (characterized by a constant angle of inclination) which move either with a constant deceleration or a constant velocity to the origin of the error state space. An appropriate design of these switching lines results in non-oscillatory convergence of the regulation error in the closed-loop system. Parameters of the lines are selected optimally in the sense of two criteria, i.e. integral absolute error (IAE) and integral of the time multiplied by the absolute error (ITAE). Furthermore, the velocity and acceleration constraints are explicitly taken into account in the optimization process. Theoretical considerations are verified by experimental tests conducted on a laboratory scale hoisting crane.
Iaetsd estimation of frequency for a single link-flexibleIaetsd Iaetsd
This document proposes an adaptive control method for an uncertain flexible robotic arm. It uses a fast online closed-loop identification method combined with an output feedback controller called a Generalized Proportional Integral (GPI) controller. An algebraic identification method is used to identify unknown system parameters and update the GPI controller in real-time. Simulations show the robustness of the adaptive controller. The document describes the flexible manipulator model, GPI controller design, algebraic estimator development, adaptive control procedure, and simulation results showing the effectiveness of the adaptive control system.
Trajectory Control With MPC For A Robot Manipülatör Using ANN ModelIJMER
In this study, in a computer the dynamic motion modelling of manipulator and control of
trajectory with an algorithm this has been tested. First after dynamic motion simulation of manipulator
has been made MPC Control. The result in this study we can observe that computed torque method gives
better results than MPC methods. So in trajectory control it is approved of using computed torque
method. In last part of this study the results are estimated forward development are exemined and
suggested. The model predictive control (MPC) technique for an articulated robot with n joints is
introduced in this paper. The proposed MPC control action is conceptually different with the trajectory
robot control methods in that the control action is determined by optimising a performance index over
the time horizon. A neural network (NN) is used in this paper as the predictive model.
Iaetsd design of a robust fuzzy logic controller for a single-link flexible m...Iaetsd Iaetsd
This document describes the design of a fuzzy logic controller for a single-link flexible manipulator. A fuzzy-PID controller is used to control an uncertain flexible robotic arm and its internal motor dynamics parameters. The controller is tested against conventional integral and PID controllers in simulations. The results show the proposed fuzzy PID controller has better robustness under variations in motor dynamics compared to the other controllers.
Neural Network Control Based on Adaptive Observer for Quadrotor HelicopterIJITCA Journal
A neural network control scheme with an adaptive observer is proposed in this paper to Quadrotor helicopter stabilization. The unknown part in Quadrotor dynamical model was estimated on line by a Single Hidden Layer network. To solve the non measurable states problem a new adaptive observer was proposed. The main purpose here is to reduce the measurement noise amplification caused by conventional high gain observer by introducing some changes in observer’s original structure that can minimize the variance and the amplitude of the noisy signal without increasing tracking error. The stability analysis of the overall closed-loop system/ observer is performed using the Lyapunov direct method. Simulation results are given to highlight the performances of the proposed scheme
This document presents an adaptive sliding-mode (ASM) controller for a vehicle steer-by-wire (SbW) system. It models the SbW system as a second-order system and regards self-aligning torque and friction as external disturbances. An ASM controller is designed that can estimate the coefficient of self-aligning torque and handle parametric uncertainties. Experiments show the ASM controller achieves better tracking of a slalom path and circular path than a sliding-mode controller or H-infinity controller under various road conditions.
A two-wheeled self-balancing robot (TWSBR) is an underactuated system that
is inherently nonlinear and unstable. While many control methods have been
introduced to enhance the performance, there is no unique solution when it comes to hardware implementation as the robot’s stability is highly dependent on accuracy of sensors and robustness of the electronic control systems. In
this study, a TWSBR that is controlled by an embedded NI myRIO-1900 board with LabVIEW-based control scheme is developed. We compare the performance between proportional-integral-derivative (PID) and linear quadratic regulator (LQR) schemes which are designed based on the TWSBR’s model that
is constructed from Newtonian principles. A hybrid PID-LQR scheme is then proposed to compensate for the individual components’ limitations. Experimental results demonstrate the PID is more effective at regulating the tilt angle of the robot in the presence of external disturbances, but it necessitates a higher velocity to sustain its equilibrium. The LQR on the other hand outperforms PID
in terms of maximum initial tilt angle. By combining both schemes, significant
improvements can be observed, such as an increase in maximum initial tilt angle
and a reduction in settling time.
Comparative analysis of observer-based LQR and LMI controllers of an inverted...journalBEEI
An inverted pendulum is a multivariable, unstable, nonlinear system that is used as a yardstick in control engineering laboratories to study, verify and confirm innovative control techniques. To implement a simple control algorithm, achieve upright stabilization and precise tracking control under external disturbances constitutes a serious challenge. Observer-based linear quadratic regulator (LQR) controller and linear matrix inequality (LMI) are proposed for the upright stabilization of the system. Simulation studies are performed using step input magnitude, and the results are analyzed. Time response specifications, integral square error (ISE), integral absolute error (IAE) and mean absolute error (MAE) were employed to investigate the performances of the proposed controllers. Based on the comparative analysis, the upright stabilization of the pendulum was achieved within the shortest possible time with both controllers however, the LMI controller exhibits better performances in both stabilization and robustness. Moreover, the LMI control scheme is effective and simple.
Study the performance of anfis controller for flexible link manipulatorIAEME Publication
This document discusses control of a flexible link manipulator using an adaptive network-based fuzzy inference system (ANFIS) controller. It first provides background on challenges controlling flexible link manipulators due to their infinite-order distributed parameter system dynamics. It then summarizes previous research using neural networks and fuzzy logic controllers. The document goes on to describe modeling a single flexible link manipulator using Lagrangian dynamics. It proposes using an ANFIS controller that combines a PID controller with a fuzzy neural network controller adapted based on PID output to control the flexible link manipulator's position and vibration.
We focus a modern methodology in this paper for adding the fuzzy logic control as well as sliding model control. This combination can enhance the MLS position control robustness and enhanced performance of it.In the start, for an application in an area to control the loops placement and position for the synchronous motor what has permanent magnetic linearity we tend to control the fuzzy sliding mode control. To resolve the chattering issues a designed controller is investigated and, in this way, steady state motion in sliding with higher accuracy is obtained. In this case, method of online tuning with the help of fuzzy logic is used in order to adjust the thickness of boundary layer and switching gains.For the suggested scheme technique, the outcomes of simulation suggest that with the classical SMC the accurate state and good dynamic performance is compared due to force chattering resistance, response by quick dynamic force and external disturbance elements and robustness against them.
A hybrid bacterial foraging and modified particle swarm optimization for mode...IJECEIAES
This paper study the model reduction procedures used for the reduction of large-scale dynamic models into a smaller one through some sort of differential and algebraic equations. A confirmed relevance between these two models exists, and it shows same characteristics under study. These reduction procedures are generally utilized for mitigating computational complexity, facilitating system analysis, and thence reducing time and costs. This paper comes out with a study showing the impact of the consolidation between the Bacterial-Foraging (BF) and Modified particle swarm optimization (MPSO) for the reduced order model (ROM). The proposed hybrid algorithm (BF-MPSO) is comprehensively compared with the BF and MPSO algorithms; a comparison is also made with selected existing techniques.
Similar to Neural Network Approach to Railway Stand Lateral SKEW Control (20)
Main news related to the CCS TSI 2023 (2023/1695)Jakub Marek
An English 🇬🇧 translation of a presentation to the speech I gave about the main changes brought by CCS TSI 2023 at the biggest Czech conference on Communications and signalling systems on Railways, which was held in Clarion Hotel Olomouc from 7th to 9th November 2023 (konferenceszt.cz). Attended by around 500 participants and 200 on-line followers.
The original Czech 🇨🇿 version of the presentation can be found here: https://www.slideshare.net/slideshow/hlavni-novinky-souvisejici-s-ccs-tsi-2023-2023-1695/269688092 .
The videorecording (in Czech) from the presentation is available here: https://youtu.be/WzjJWm4IyPk?si=SImb06tuXGb30BEH .
Driving Business Innovation: Latest Generative AI Advancements & Success StorySafe Software
Are you ready to revolutionize how you handle data? Join us for a webinar where we’ll bring you up to speed with the latest advancements in Generative AI technology and discover how leveraging FME with tools from giants like Google Gemini, Amazon, and Microsoft OpenAI can supercharge your workflow efficiency.
During the hour, we’ll take you through:
Guest Speaker Segment with Hannah Barrington: Dive into the world of dynamic real estate marketing with Hannah, the Marketing Manager at Workspace Group. Hear firsthand how their team generates engaging descriptions for thousands of office units by integrating diverse data sources—from PDF floorplans to web pages—using FME transformers, like OpenAIVisionConnector and AnthropicVisionConnector. This use case will show you how GenAI can streamline content creation for marketing across the board.
Ollama Use Case: Learn how Scenario Specialist Dmitri Bagh has utilized Ollama within FME to input data, create custom models, and enhance security protocols. This segment will include demos to illustrate the full capabilities of FME in AI-driven processes.
Custom AI Models: Discover how to leverage FME to build personalized AI models using your data. Whether it’s populating a model with local data for added security or integrating public AI tools, find out how FME facilitates a versatile and secure approach to AI.
We’ll wrap up with a live Q&A session where you can engage with our experts on your specific use cases, and learn more about optimizing your data workflows with AI.
This webinar is ideal for professionals seeking to harness the power of AI within their data management systems while ensuring high levels of customization and security. Whether you're a novice or an expert, gain actionable insights and strategies to elevate your data processes. Join us to see how FME and AI can revolutionize how you work with data!
Monitoring and Managing Anomaly Detection on OpenShift.pdfTosin Akinosho
Monitoring and Managing Anomaly Detection on OpenShift
Overview
Dive into the world of anomaly detection on edge devices with our comprehensive hands-on tutorial. This SlideShare presentation will guide you through the entire process, from data collection and model training to edge deployment and real-time monitoring. Perfect for those looking to implement robust anomaly detection systems on resource-constrained IoT/edge devices.
Key Topics Covered
1. Introduction to Anomaly Detection
- Understand the fundamentals of anomaly detection and its importance in identifying unusual behavior or failures in systems.
2. Understanding Edge (IoT)
- Learn about edge computing and IoT, and how they enable real-time data processing and decision-making at the source.
3. What is ArgoCD?
- Discover ArgoCD, a declarative, GitOps continuous delivery tool for Kubernetes, and its role in deploying applications on edge devices.
4. Deployment Using ArgoCD for Edge Devices
- Step-by-step guide on deploying anomaly detection models on edge devices using ArgoCD.
5. Introduction to Apache Kafka and S3
- Explore Apache Kafka for real-time data streaming and Amazon S3 for scalable storage solutions.
6. Viewing Kafka Messages in the Data Lake
- Learn how to view and analyze Kafka messages stored in a data lake for better insights.
7. What is Prometheus?
- Get to know Prometheus, an open-source monitoring and alerting toolkit, and its application in monitoring edge devices.
8. Monitoring Application Metrics with Prometheus
- Detailed instructions on setting up Prometheus to monitor the performance and health of your anomaly detection system.
9. What is Camel K?
- Introduction to Camel K, a lightweight integration framework built on Apache Camel, designed for Kubernetes.
10. Configuring Camel K Integrations for Data Pipelines
- Learn how to configure Camel K for seamless data pipeline integrations in your anomaly detection workflow.
11. What is a Jupyter Notebook?
- Overview of Jupyter Notebooks, an open-source web application for creating and sharing documents with live code, equations, visualizations, and narrative text.
12. Jupyter Notebooks with Code Examples
- Hands-on examples and code snippets in Jupyter Notebooks to help you implement and test anomaly detection models.
Best 20 SEO Techniques To Improve Website Visibility In SERPPixlogix Infotech
Boost your website's visibility with proven SEO techniques! Our latest blog dives into essential strategies to enhance your online presence, increase traffic, and rank higher on search engines. From keyword optimization to quality content creation, learn how to make your site stand out in the crowded digital landscape. Discover actionable tips and expert insights to elevate your SEO game.
Unlock the Future of Search with MongoDB Atlas_ Vector Search Unleashed.pdfMalak Abu Hammad
Discover how MongoDB Atlas and vector search technology can revolutionize your application's search capabilities. This comprehensive presentation covers:
* What is Vector Search?
* Importance and benefits of vector search
* Practical use cases across various industries
* Step-by-step implementation guide
* Live demos with code snippets
* Enhancing LLM capabilities with vector search
* Best practices and optimization strategies
Perfect for developers, AI enthusiasts, and tech leaders. Learn how to leverage MongoDB Atlas to deliver highly relevant, context-aware search results, transforming your data retrieval process. Stay ahead in tech innovation and maximize the potential of your applications.
#MongoDB #VectorSearch #AI #SemanticSearch #TechInnovation #DataScience #LLM #MachineLearning #SearchTechnology
Let's Integrate MuleSoft RPA, COMPOSER, APM with AWS IDP along with Slackshyamraj55
Discover the seamless integration of RPA (Robotic Process Automation), COMPOSER, and APM with AWS IDP enhanced with Slack notifications. Explore how these technologies converge to streamline workflows, optimize performance, and ensure secure access, all while leveraging the power of AWS IDP and real-time communication via Slack notifications.
Your One-Stop Shop for Python Success: Top 10 US Python Development Providersakankshawande
Simplify your search for a reliable Python development partner! This list presents the top 10 trusted US providers offering comprehensive Python development services, ensuring your project's success from conception to completion.
In the rapidly evolving landscape of technologies, XML continues to play a vital role in structuring, storing, and transporting data across diverse systems. The recent advancements in artificial intelligence (AI) present new methodologies for enhancing XML development workflows, introducing efficiency, automation, and intelligent capabilities. This presentation will outline the scope and perspective of utilizing AI in XML development. The potential benefits and the possible pitfalls will be highlighted, providing a balanced view of the subject.
We will explore the capabilities of AI in understanding XML markup languages and autonomously creating structured XML content. Additionally, we will examine the capacity of AI to enrich plain text with appropriate XML markup. Practical examples and methodological guidelines will be provided to elucidate how AI can be effectively prompted to interpret and generate accurate XML markup.
Further emphasis will be placed on the role of AI in developing XSLT, or schemas such as XSD and Schematron. We will address the techniques and strategies adopted to create prompts for generating code, explaining code, or refactoring the code, and the results achieved.
The discussion will extend to how AI can be used to transform XML content. In particular, the focus will be on the use of AI XPath extension functions in XSLT, Schematron, Schematron Quick Fixes, or for XML content refactoring.
The presentation aims to deliver a comprehensive overview of AI usage in XML development, providing attendees with the necessary knowledge to make informed decisions. Whether you’re at the early stages of adopting AI or considering integrating it in advanced XML development, this presentation will cover all levels of expertise.
By highlighting the potential advantages and challenges of integrating AI with XML development tools and languages, the presentation seeks to inspire thoughtful conversation around the future of XML development. We’ll not only delve into the technical aspects of AI-powered XML development but also discuss practical implications and possible future directions.
Trusted Execution Environment for Decentralized Process MiningLucaBarbaro3
Presentation of the paper "Trusted Execution Environment for Decentralized Process Mining" given during the CAiSE 2024 Conference in Cyprus on June 7, 2024.
Taking AI to the Next Level in Manufacturing.pdfssuserfac0301
Read Taking AI to the Next Level in Manufacturing to gain insights on AI adoption in the manufacturing industry, such as:
1. How quickly AI is being implemented in manufacturing.
2. Which barriers stand in the way of AI adoption.
3. How data quality and governance form the backbone of AI.
4. Organizational processes and structures that may inhibit effective AI adoption.
6. Ideas and approaches to help build your organization's AI strategy.
2. 328
Computer Science & Information Technology (CS & IT)
such method is not suitable for real time control. Thus, this paper aims to investigate the possible
use of a neural network approach for lateral control of such railway wheel sets. The suitability of
application to this problem is motivated by promising theoretical studies of higher-order neural
units (HONUs), especially the quadratic neural unit for engineering problems for [3]-[7]. These
studies are focused on the use of supervised learning based approaches for polynomial structured
neural units, also known as a class of HONUs, for adaptive identification and control of real
engineering systems. Further motivation arises from the successful implementation of a quadratic
neural unit controller (Neuro-controller), for control of a bathyscaphe system located in the
automatic control laboratories of CTU [12]. Where, here, such controller adhered more closely to
the desired behaviour of the system than the conventionally used PID controller. An extension on
this result can be recalled in the work [11]. Where, further study was made via introduction of a
new software for adaptive identification and control, along with further testing on both a
theoretical system and the previously mentioned bathyscaphe system. Given this, in this paper we
aim to investigate use of neural network (NN) approaches in the following manner. We will begin
by explaining in more depth the problem behind the recently employed state feedback and
cascade PID control of the linearization of the CTU roller rig. The proceeding section will then
describe the principles and control schemes behind the various experimented methods, for
adaptive identification and control. Following this, an experimental analysis of the various
approaches of NN, focusing firstly on adaptive identification of the CTU roller rig system and
then to test the various NN methods for control. The final component of this paper will then be, to
analyse and compare the produced experimental results, with a conclusion to be drawn at the end.
2. PROBLEM DESCRIPTION
This section aims to describe in more depth, the functionality behind the previously introduced
experimental railway stand (CTU roller rig). More importantly, for the scope of this paper, we
will elaborate on the necessity for control of such roller rig and the issues of conventional linear
PID control or state feedback control, for controlling the lateral skew.
1.1
Experimental Setup of the Roller Rig at CTU
The below figure (Figure 1) depicts the experimental setup of the CTU roller rig, along with the
3-D model for design and simulation purposes. This rig features 5 motor drives, two pairs of 0.5m
diameter rollers are independently driven, via the largest of the set of drives (FM1 & 2). To
simulate both straight and curved tracks, similar to those present on real railway networks. A
central servo motor (FM4) was introduced, to yaw the lower roller pair for replicating curved
track motion. This setup however, assumes simulation of a rail pair without effects of rail
buckling. For manipulation of the wheel set yaw, a separate servo motor, central to the wheel set
(FM3) is installed. For the scope of this paper, this is the most crucial component, as it is the
action of this servo motor that is used for control actuation for the lateral skew of the wheel sets.
Where, the discussed control setups in this paper will analyse control via manipulation of this
servo motor torque. Finally, a fifth drive FM5 is located between the front roller set, as a control
differential.
3. Computer Science & Information Technology (CS & IT)
329
Figure 1: CTU Roller Rig (real left, PTC Pro/Engineer 3-D model on the right)
1.2
Problem of the State Feedback and Cascade PID Control
In the work [1], the method of lateral skew control via state feedback with cascade PID setup is
introduced. In this presented setup, a linearization of the roller rig is used, via translation to a state
space representation, achieved via a 12x12 matrix of dynamics. However, according to theory,
this linearization of the roller rig system appeared to be uncontrollable and unobservable via
analysis by common linear algebra and control approaches.
Figure 2: Control loop on linearization of the roller rig with State feedback + Cascade PID control
The above figure (Figure 2), depicts the control scheme for this setup. After numerous testing, it
was found that a simulation step of approximately < 1E-6 was necessary, in order to achieve
stability of the continuous control loop, which cannot be used for practical application. Hence, it
is for this reason that an alternative approach is necessary for the lateral skew control of such
railway wheel set. Thus, this paper aims on investigating the possibility of adaptive identification
and control of the stand, based on measured data. For the purposes of this paper the data will be
generated via a 3-D simulation model via software SIMPACK, linked via SIMAT toolbox with
MATLAB Simulink, to provide real time simulation of the roller rig as a plant system for the
investigated control approaches. Then, this data will be used as the training data for neural
network models, to simplify otherwise complex simulation models and to allow a constant
sampling for real time usability, furthermore, investigation into the potentials for real-time neuralnetwork based control.
4. 330
Computer Science & Information Technology (CS & IT)
3. APPLIED METHODS
In this section, the various NN approaches, aimed for adaptive identification and control of the
previously introduced lateral skew problem, will be described. These approaches are based on the
well know Gradient Descent (GD) method used as a tool for defining the learning rule of the
applied neural units. These applied neural units (adaptive models) are trained via two methods,
which will be focused on in this paper. These methods are namely, the method of incremental
training or Real Time Recurrent Learning (RTRL) [8] applicable for dynamic adaptive models or
a batch form of training, which is a variation of the back propagation through time or (BPTT)
training [10], as an extension of RTRL training in combination with the famous LevenbergMarquardt equation.
1.1
Preliminaries
This subsection aims to review the important theories from works [11]-[13], behind the GD
method and structures of the neural units used within this paper. Firstly, we will begin by
introducing the famous GD algorithm for the linear and quadratic neural units. To understand
this, we must begin with the polynomial models of the linear and quadratic (LNU and QNU)
neural units respectively, as follows
n
y = ∑ xi wi = w0 .x0 +w1.x1 + ... + wn .xn = w.x
(1)
i =0
n
n
y = ∑∑ xi x j wi , j = w0,0 x0 x0 +w0,1 x0 x1 + ... + wn, n xn 2 = rowx.colW
(2)
i =0 j =0
Where, y represents the output of the LNU (1) and QNU (2) respectively. In regards to the LNU,
w stands for an updatable vector of neural weights and x, represents the input values of the
engineering system in the case of a purely static model and in the sense of a dynamic model, a
combination of inputs of the real system and neural model outputs. Looking at equation (2), rowx
is a long-vector representation of the utilised input vector. Where, colW represents a weight
matrix of the quadratic neural unit in general. From this, we may then understand the GD
algorithm as applied to such neural units.
∂y (k )
∂wi
∂y (k + 1)
colW (k +1) = colW (k ) + µ .e(k ).
∂colW
wi +1 = wi + µ.e(k ).
(3)
(4)
Here, equation (3) & (4) depicts the GD algorithm for both the LNU and QNU respectively. Here
the output of the GD algorithm is to incrementally over sample-by-sample, update the neural
weights such to adaptively teach the LNU or QNU model, to learn the engineering system. Here,
µ represents the learning rate of the GD algorithm. This is analogical to humans where, setting µ
relatively high, corresponds to faster learning of the human and consequently, means the less
detail the human can remember and digest from their learning. Furthermore, setting this
parameter to a smaller value, corresponds to a slowing rate of learning i.e. the human may
remember the information learned, quite well, but less information overall about the subject. The
next parameter is e(k) (k representing the number of the sample), this represents the current error
∂y (k )
between the real and calculated output of the model. The final term
, corresponds to the
∂wi
partial derivatives of the neural unit output, with respect to the individual neural weights.
5. Computer Science & Information Technology (CS & IT)
331
Regarding the QNU in equation (4), we see that the GD algorithm is analogical, with exception of
updating the matrix of neural weights as opposed to a vector in the sense of the LNU.
Till now, the structures of GD for LNU and QNU were reviewed in the format of RTRL or
sample-by-sample method of learning. Where, the neural weights are updated over each new
sample of the engineering system data. However for certain engineering processes, it is more
advantageous to use the BPTT batch method of training these neural weights, over runs of the
algorithm rather than over each sample. This is because the RTRL method focuses on the
contemporary governing law of the system as opposed the BPTT which focuses more so on the
main governing law of the input and outputs of the engineering system. The BPTT method is
achieved via an extension of the RTRL method with combination of the famous LevenbergMarquardt equation. It is also important to note that this method is more preferable in cases where
the data may be affected by noise. The following equation denotes the weight update rule for the
BPTT method;
∆w = (J T .J +
1
µ
.I ) −1 .J T .e
(5)
Here, the neural weights are updated over each run of the algorithm (or batch trained) in the
following way w=w+∆w. Here equation (5) describes the change, necessary for the update of the
batch trained weights. Where, J represents the Jacobian matrix of derivatives for the neural unit.
This may be the complete partial derivatives of the neural model with respect to each neural
weights, or in practical applications it seems useful to simply introduce this Jacobian matrix as
the input vector or matrix itself, being x and colx for LNU and QNU respectively. Furthermore, it
is important to note that colx = rowx T .
Often in such adaptive neural units, it is apparent that a modification of the normalised learning
rate may be used to solve issues associated with instability of learning. In practise, it is possible to
employ the simplified normalised learning rate as presented in the work [11], as follows;
η=
µ
x(k)x(k)T + 1
(6)
Where, equation (6) represents the normalised learning rate in the sense of LNU. This is
analogically represented in the QNU, by replacement of x(k) (i.e. each k th sample of the input
vector x) with colx(k). It should be noted that the above representation holds for RTRL training
for dynamic adaptive models, where the algorithms used in this paper for BPTT method of
training, take the learning rate µ itself.
1.2
Adaptive Identification and control
The previous section focused on usage of neural units in the sense of adaptive identification of a
real engineering system. In this subsection, we extend on these neural units as a method of
control, with brief review from works [11]-[13], and extension as applied to the lateral skew
problem, which is indeed our focus within this paper.
6. 332
Computer Science & Information Technology (CS & IT)
Figure 3: Adaptive Identification with supervised learning of neural networks (where, w.x=LNU,
rowX.colW=QNU)
The above figure (Figure 3), depicts the identification scheme for the previously reviewed neural
network architectures. Where, for the scope of this paper a simulation of the real roller rig will be
used for data generation of the investigated neural network approaches for control. u, represents
the input data of the roller rig, in our case this is the yaw torque of the servo motor system for
manipulating the lateral skew of the wheel set. The output yreal, is the simulated output from the
real time 3-D SIMPACK model and y, being the output of the neural unit, and the difference
being the error, e.
Figure 4: Adaptive control loop for experimental study of a neural network controller (where, w.x=LNU,
rowX.colW=QNU)
The above (Figure 4), shows an extension of the discussed neural architectures for application to
lateral skew control of the roller rig system. Here, once the neural unit as a model is identified, it
may be utilised as the foundation for a neural network based control setup. For the scope of this
paper, we will propose an offline tuned control scheme, as the goal here, is to investigate the
potentials of applying a neural network based control, for this application. However, the
extension to online control is indeed the ultimate aim of our research to this problem, beyond this
paper. Figure 4, depicts use of a second neural unit as a controller. Similarly to the previously
mentioned architectures, as a controller here too the neural unit may take shape to that of the
LNU or QNU adaptive models. However, in this case the adaptive neural weights are tuned
differently to that of those used for the neural unit as a model and hence, depicted as v.
Analogically, should a QNU structure be applied, these neural weights would be represented in a
matrix form. Further to this, the above figure (Figure 4) introduced a new vector ξ. This vector
comprises of a combination of outputs from the neural unit as a model and the difference between
the desired behaviour (in our case desired yaw torque or lateral skew of the roller rig) and the
7. Computer Science & Information Technology (CS & IT)
333
output of the neural model. v. ξ or collectively, the variable q, thus serves as a manipulator for the
newly feed samples of the neural unit as an identified model. Here the GD algorithms are
employed in the following manner to achieve sample-by-sample adaptation of the neural weights
for the controller, as follows;
vi +1 = vi + µ.ereg (k ).
∂y (k )
∂vi
(7)
Where vi , are adaptable neural weights of the neural unit as a controller and ereg(k) is the error
between the desired value of the real system (in our case the roller rig, where the desired value
will be denoted as d) and the real system output value at sample k.
∂y (k )
, is the partial
∂vi
derivative of the output of the neural unit as a model, with respect to the individual adaptive
neural weights of the neural unit as a controller. An extension of this weight update scheme for
BPTT training would result in the following form v=v+∆v, where the change of neural weights
for each batch would be analogical to equation (5).
4. EXPERIMENTAL ANALYSIS
This section aims to analyse the previously reviewed methods of adaptive identification and
control, via the introduced neural network architectures. Particularly, we aim to investigate the
applicability of the discussed neural architectures for this problem of the lateral skew control of
railway wheel sets. In this section, we will simulate our results using simulation data of the
previously described 3-D SIMPACK model (at mid-range speed) linked with MATLAB Simulink
in real time simulation. The data was produced with 0.001 sampling over a 10 second interval.
First, we will analyse the results for identification of the roller rig as a system, via the different
methods of the neural network models, followed by an extension of control, where the various
combinations of neural network architectures will be tested.
DLNU with RTRL
Almost exact identification of
Roller Rig data (lines are superimposed)
Deviation of the model with the Roller Rig
data, over each sample
Figure 5: Testing of the adaptive identification, where the plant (roller rig) is represented by blue colour
and green being the neural model. Trained by DLNU with RTRL Training; µ=1,
epochs =10, For ny=3 (previous samples of model output) and nu=5 (previous samples of real system input)
8. 334
Computer Science & Information Technology (CS & IT)
The above figure (Figure 5) and following figure (Figure 6), illustrates the adaptive identification
process of the roller rig system data. We can note, that both methods of the dynamic LNU or
(DLNU) and dynamic QNU or (DQNU), for RTRL learning methods, achieved almost exact
identification. Where, the DQNU performed slightly faster in terms of convergence to minima of
sum of square errors, as opposed to the DLNU.
DQNU with RTRL. Excellent identification with slightly
faster learning than the DLNU of same parameters
Error of model is still within an
excellent range
Sum of square errors decreases more quickly
with the DQNU for same parameters.
Figure 6 : Testing of the adaptive identification, where the plant (roller rig) is represented by blue colour
and green being the neural model. Trained by DQNU with RTRL Training; µ=1, epochs =10, For ny=3
(previous samples of model output) and nu=5 (previous samples of real system input)
9. Computer Science & Information Technology (CS & IT)
Desired (black dashed)
335
Roller Rig system data (blue)
Neural controller (magenta) applied to
identified Roller Rig data
Error plot illustrates deviation of the Neural
Controller and desired behaviour.
Sum of Square Errors reaches
close to 10-4 after 50 epochs
Figure 7: Testing of the adaptively tuned control loop, where the roller rig is represented by trained DLNU with constant and
previously trained parameters (Figure 5) and the adaptive feedback controller is trained via LNU, after 10 epochs Adaptive
Identification (of DLNU with RTRL) + Neural Unit Controller for lateral skew control of roller rig system - Using LNU with
Incremental training µ=0.1, Data re-sampling = each 5th sample, epochs =200, for nqy=0 (previous samples of neural model output)
and nqe=2 (previous samples of difference of the neural model output and desired behaviour of the system.
Closer to desired behaviour of the
system with QNU + BPTT
The difference of the Neural Controller
and desired behaviour is within +0.2 to
-0.2 mm.
Sum of Square Errors reaches close to 10after only few epochs, with QNU and
BPTT training.
4
Figure 8: Testing of the adaptively tuned control loop, where the roller rig is represented by trained DQNU with constant and
previously trained parameters (Figure 6) and the adaptive feedback controller is trained via QNU,
after 10 epochs Adaptive Identification (of DQNU with RTRL) + Neural Unit Controller for lateral skew control of roller rig system Using QNU with BPTT Training µ=0.00818,
epochs=5, for nqy=0 and nqe=3
10. e
336
Computer Science & Information Technology (CS & IT)
Here, Figure 7 & Figure 8 show the application of the various neural units for control of the
lateral skew. The black dotted line, illustrates the chosen desired behaviour of the roller rig lateral
skew. It was found that the BPTT training method used in combination with the QNU had the
best performance. Providing closer control to the desired behaviour, as compared with the
incremental training methods of both LNU and QNU architectures.
Desired lateral skew (black dashed line)
set to ideal position of zero
Model provides almost exact response to the
desired after the first few samples.
Figure 9: Testing of the adaptively tuned control loop, where the roller rig is represented by trained DQNU
with constant and previously trained parameters (Figure 6) and the adaptive feedback controller is trained
via QNU,
after 10 epochs Adaptive Identification (of DQNU with RTRL) + Neural Unit Controller for lateral skew
control of roller rig system - Using QNU with BPTT Training µ=0.001, epochs=5, for nqy=3, nqe=3
Figure 9 depicts a theoretical test of the roller rig system under an idealised case of zero lateral
skew. This is included in our results to investigate the capability of the used neural controller.
From this figure, we see that after several small samples at the beginning of the controller’s
application the controller provides almost exactly zero lateral skew. This result however, must be
reasoned with what is really possible in terms of actuation of the real system and real conditions
of the rig, such to achieve this ideal of a result.
5. DISCUSSION OF EXPERIMENTAL RESULTS
In the previous section, the experimental results of the various neural network architectures for
adaptive identification and control were presented. From the first set of figures, regarding the
adaptive identification for the roller rig system, it was found that the most suitable learning
method for both the DLNU and DQNU was the RTRL training method. Here, relatively few
epochs were adequate for both architectures to identify the system data, with almost exact
11. Computer Science & Information Technology (CS & IT)
337
identification. The DQNU model was slightly faster in learning the behaviour of the system, thus,
was used as the identified neural unit as a model, in the control section. The extension of these
various architectures for control showed more distinguishable results. The application of both
LNU and QNU with incremental learning for the neural controller was only possible when a
higher number of epochs were run, or higher learning of the model with a small learning rate.
Few runs of the algorithm were unable to match the desired behaviour even moderately.
However, after the first few samples of application as in Figure 7, reasonable behaviour of the
controller was achieved. The combination of identification via the DQNU followed by extension
of a QNU trained by BPTT on the neural controller, showed to follow more closely to the desired
behaviour of the roller rig system than the other tested combinations. Where, the setup of the
QNU with BPTT neural controller featured input vector ξ, being comprised only of several errors
or differences of the desired behaviour, with the output of the neural unit as a model. Further to
this, the training was substantially quicker than the incremental training method, showing that
only several epochs were sufficient for the model to achieve its optimal behaviour with respect to
the desired of the system.
In the above results, two forms of the desired behaviour were used. The first was under proposal
that the lateral derivation of the wheel set would be within a small tolerance range, which is most
realistic in real application of such control. The presented desired data is relative to the small
range of lateral skew in the simulation data. Where, within this exampled range, we could be sure
the wheel flanges would not contact the rails of the real roller rig. The desired behaviour is
however relative to the setup of the wheel sets and rails, thus the desired deviation and lateral
skew outputs may indeed vary for real railway applications. The second was an idealised situation
where no lateral skew at all would occur. This is principally unrealistic for the neural controller to
achieve zero lateral skew, due to a combination of natural factors on real railway wheel sets,
however for demonstrational purposes, we may note the capability of the neural controller,
providing almost exact behaviour of the idealised lateral skew, as applied on the identified roller
rig system.
6. CONCLUSION
Referring back to the original control loop depicted in Figure 1 (as presented in work [1]), of the
Simulink model of the linearization of the roller rig system and state feedback with cascade PID
controller. We can recall that a minimum sampling of <1E-6 seconds for numerical stability was
necessary, which is not possible for practical application. After investigation of the presented
neural network approaches, we may conclude that it has promising potential for real application
on controlling the lateral skew of IRW railway wheel sets. Now, we note that the achieved
sampling can be within order of 0.005 (5E-3) seconds, for adequate functionality of the neural
network based, adaptive identification and control system as applied to the roller rig. The tested
models in this paper were DLNU and DQNU for adaptive identification models and LNU or
QNU architectures as a controller. The experimental results of these architectures showed that
both DLNU and DQNU, precisely approximated the complex SIMPACK model of the roller rig
system. This result however could vary when using more or, real training data measured from the
rig itself. In terms of control of the lateral skew, the QNU architecture showed better results to
keep the lateral skew within close limits to the desired behaviour of the system, which is also in
general consensus with our previous findings on using QNU for control. We also can note the
significantly faster tuning of the control loop via QNU as compared to the LNU, more
particularly, the BPTT training method as applied to the controller compared to the incremental
method. Where, more desirable control was shown, in much fewer epochs or runs of the
controller algorithm. Thus, this case study shows that with proper tuning of the investigated
neural units, application to the real system is theoretically possible, to achieve adequate control
12. 338
Computer Science & Information Technology (CS & IT)
for this investigated problem. Thus, real application of such adaptive control loop of neural
network based architecture, is indeed the very next step of our research.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the following grant for support during this work:
SGS12/177/OHK2/3T/12SGS12/177/OHK2/3T/12, Non-conventional and Cognitive Signal
Processing Methods of Dynamic Systems.
And also the Technology Agency of Czech Republic. Project No: TE01020038 “Competence
Center of Railway Vehicles”.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
J. Kalivoda, P. Bauer (2013) “Curving Behaviour of 2-Axle Bogie with Independently Rotating
Wheels – Simulations and Scaled Roller Rig Tests”- 23rd International Symposium on Dynamics of
Vehicles on Roads and Tracks, Qingdao, China, August 19th-23rd.
S. Bruni, R. M. Godall, T.X. Mei, H. Tsunashima (2007) “Control and monitoring for railway vehicle
dynamics”, Vehicle System Dynamics, vol. 45 issue 7-8, pp. 743-779
Gupta, M., M., Bukovsky, I., Homma, N., Solo M. G. A., Hou Z.-G. (2012): “Fundamentals of
Higher Order Neural Networks for Modeling and Simulation“, in Artificial Higher Order Neural
Networks for Modeling and Simulation, ed. M. Zhang, IGI Global,.
Bukovsky, I., Bila, J., Gupta, M., M, Hou, Z-G., Homma, N. (2010): “Foundation and Classification
of Nonconventional Neural Units and Paradigm of Nonsynaptic Neural Interaction” in Discoveries
and Breakthroughs in Cognitive Informatics and Natural Intelligence within the series of the
Advances in Cognitive Informatics and Natural Intelligence (ACINI), ed. Y. Wang, IGI Publishing,
Hershey PA, USA, ISBN: 978-1-60566-902-1, pp.508-523.
Bukovsky, I., Homma, N., Smetana, L., Rodriguez, R., Mironovova M., Vrana S. (2010): “Quadratic
Neural Unit is a Good Compromise between Linear Models and Neural Networks for Industrial
Applications”, ICCI 2010 The 9th IEEE International Conference on Cognitive Informatics, Tsinghua
University, Beijing, China, July 7-9.
Bukovsky I., S. Redlapalli and M. M. Gupta (2003): Quadratic and Cubic Neural Units for
Identification and Fast State Feedback Control of Unknown Non-Linear Dynamic Systems, Fourth
International Symposium on Uncertainty Modeling and Analysis ISUMA 2003, IEEE Computer
Society, , Maryland USA, ISBN 0-7695-1997-0, p.p.330-334
Rodriguez , R., Bukovsky, I., Homma, N.: “Potentials of Quadratic Neural Unit for Applications”, in
International Journal of Software Science and Computational Intelligence (IJSSCI) ,vol 3, issue 3, IGI
Global, Publishing, Hershey PA, USA ISSN.
Bukovsky, I., Lepold, M., Bila J. (2010): “Quadratic Neural Unit and its Network in Validation of
Process Data of Steam Turbine Loop and Energetic Boiler”,WCCI 2010, IEEE Int. Joint. Conf. on
Neural Networks IJCNN, Barcelona,Spain.
R. J. Williams and D. Zipser (1989) “A learning algorithm for continually running fully recurrent
neural networks,” Neural Computation, vol. 1, pp. 270–280.
P. J.Werbos (Oct. 1990) “Backpropagation through time: What it is and how to do it” Proc. IEEE, vol.
78, no. 10, pp. 1550–1560.
Peter M. Benes (2013): Software Application for Adaptive Identification and Controller Tuning
Student's Conference STC, Faculty of Mechanical Engineering, CTU in Prague.
Ladislav Smetana (2008): “Nonlinear Neuro-Controller for Automatic Control Laboratory System”,
Master’s Thesis, Czech Tech. Univ. in Prague,
Laboratory System Batyscaphe, Automatic Control Laboratory, Dpt. of Instr. and Cont. Eng., FME,
Czech Technical University in Prague, online http://vlab.fsid.cvut.cz/en/ulohy/batyskaf.php, 2010.
13. Computer Science & Information Technology (CS & IT)
339
Short Biography of the Authors
Peter Mark Beneš received his Bachelor’s degree with honours at Czech Technical University (CTU) in
Prague in 2012. Currently he is a Master’s student, with expected PhD studies to follow in 2014. His
research focuses on non-conventional neural networks for adaptive identification and control of industrial
systems including hoist mechanisms and skew control of rail-based mechanisms as such cranes and
railway vehicles. Peter’s work has been awarded in local and international student competitions also with
an industrial BOSCH award in 2013.
Matouš Cejnek received his Bachelor’s degree at Czech Technical University (CTU) in Prague in 2012.
Currently he is a Master’s student at Czech Technical University in Prague. His research focuses on nonconventional neural networks for adaptive systems and novelty detection in time series and biomedical
applications. Matous’s work has been awarded in local and international student competitions 2013.
Jan Kalivoda graduated from Czech Technical University (CTU) in Prague where he received his
Master’s degree with honours in 1996 and Ph.D. in the field of Machines and Equipment for
Transportation in 2006. Currently he is a teacher and active researcher in the Department of Automotive,
Combustion Engine and Railway Engineering at CTU. His research interests include MBS models of
railway vehicles, Mechatronics for railway vehicles, Active control of railway vehicle suspensions and
wheel sets.
Ivo Bukovsky graduated from Czech Technical University in Prague (CTU) where he received his Ph.D.
in the field of Control and System Engineering in 2007 and became associate professor since 2013. His
research interests include higher-order neural networks, adaptive evaluation of time series and systems,
multi-scale-analysis approaches, control and biomedical applications. He was a visiting researcher at the
University of Saskatchewan (2003), at the University of Manitoba in Canada (2010), and he was a
visiting professor at Tohoku University (2011)