Marcin Kaminski

FPGA Implementation of the Multilayer Neural Network for the Speed Estimation of the Two-Mass Drive System

This paper presents a practical realization of a neural network (NN)-based estimator of the load machine speed for a drive system with elastic coupling, using a reconfigurable field-programmable gate array (FPGA). The system presented is unique because the multilayer NN is implemented in the FPGA placed inside the NI CompactRIO controller. The neural network used as a state estimator was trained with the Levenberg-Marquardt algorithm. Special algorithm for implementation of the multilayer neural networks in such hardware platform is presented, focused on the minimization of the used programmable blocks of the FPGA matrix. The algorithm code for the neural estimator implemented in C-RIO was realized using the LabVIEW software. The neural estimators are tested: offline (based on the measured testing database) and online (in the closed-loop control structure). These estimators are tested also for changeable inertia moment of the load machine of the drive system with elastic joint. Presented results of the experimental tests confirm that the multilayer NN, implemented in the FPGA with the use of the higher level programming language, ensures a high-quality state variable estimation of the two-mass drive system.

Implementation of a Sliding-Mode Controller With an Integral Function and Fuzzy Gain Value for the Electrical Drive With an Elastic Joint

This paper presents a modified sliding-mode structure implemented for the speed control of a two-mass drive. A characteristic feature of the presented control method is the higher rank of the switching function caused by the application of an integral element (sliding mode with an integral function control). The proposed control system is a combination of a sliding-mode controller and a linear controller. Furthermore, to eliminate the chattering phenomenon related to the sliding-mode control, a switching function with a variable slope based on the fuzzy system is implemented. This solution ensures the robustness and dynamics of a two-mass drive better than with a linear speed controller. The main stages of the design methodology of the presented speed control structure are described in the initial sections of this paper. In the subsequent sections, simulation and experimental tests for the proposed control structure are presented and discussed.

FPGA Implementation of ADALINE-Based Speed Controller in a Two-Mass System

The paper presents the application of an adaptive neural controller used for speed control of electrical drives with elastic joint. The described project is realized in CompactRIO controller (cRIO-real-time embedded controller with reconfigurable input and output modules) equipped with an FPGA chip. The proposed speed controller is based on Adaptive Linear Neuron (ADALINE) model with on-line updated weights coefficients. The main advantages of the tested controller are simplicity and a reduced number of parameters for selection in the design process. Several stages of the real implementation are described. The two-mass drive system is modeled using the main processor of the cRIO, to emulate the real system, while the structure of the ADALINE model and its adaptation law are implemented in the FPGA module. Thus, hardware in the loop simulation is obtained. The obtained results present correct speed control with high dynamics and show the influence of the adaptation coefficient of the ADALINE-based controller on drive transients. Except for this the robustness of the proposed controller against changes of mechanical time constant of the load machine is presented.

A Modified Fuzzy Luenberger Observer for a Two-Mass Drive System

In this paper, issues related to the design of a fuzzy Luenberger observer for a drive system with a flexible joint is presented. The proposed estimator ensures shifting observer closed-loop poles in accordance with the present condition of the plant. The dynamic states are recognized on the basis of the error between the plant and estimator outputs (classical approach), as well as additional signal (difference between electromagnetic, and shaft torque and its derivative). Depending on these signals, the fuzzy systems determine the location of the observer poles. The proposed control structure is investigated through a number of experimental tests.

Application of the OBD method for optimization of neural state variable estimators of the two-mass drive system

This paper presents neural estimators of the mechanical state variables of the electrical drive system with elastic joints. The non-measurable state variables, as the torsional torque and the load machine speed are estimated using multilayer feed-forward neural networks. The main stages of the design methodology of these neural estimators are presented. The optimal brain damage method is implemented for the structure optimization of each neural network. Then signals estimated by neural estimators are tested in the electrical drive control structure with additional feedbacks from the estimated shaft torque and the difference between the motor and the load speeds. The simulation results show good accuracy of both presented neural estimators for the wide range of changes of the reference speed and the load torque. The simulation results are then verified by laboratory experiments.

Adaptive Neurocontrollers for Drive Systems: Basic Concepts, Theory and Applications

In this chapter basic principles of neurocontrol are revised and discussed from the point of view of applications in converter-fed drive systems. The main neural network structures used as neural controllers are presented and classified into two groups: off-line and on-line trained controllers. From the point of view of drive system uncertainties, caused by simplifying assumptions under mathematical model formulation, errors in drive parameters identification and changes of the models and their parameters under different operation conditions, on-line adaptive neural controllers are proposed. Various neural structures and their on-line training methods are discussed. The chosen neurocontrollers were verified in simulation and experimental tests for converter-fed drives with rigid and resilient mechanical connections between the driving motor and loading machine.

FPGA realization of the neural speed estimator for the drive system with elastic coupling

In this paper implementation of neural network based estimator of the load machine speed for two-mass drive system on a reconfigurable field-programmable gate array (FPGA) is presented. Due to such implementation, the developed neural estimator performs the parallel signal processing and thus the realization time is much shorter than using classical DSP. The FPGA based neural network structure has been carried out in the controller NI CompactRIO. The neural estimator design process, including input vector preprocessing, training and testing procedure is presented. The main part of research is concentrated on implementation of neural networks estimators with required internal structure in FPGA matrices. Neural estimators are tested for changeable moment of inertia of the load machine in the drive system with elastic joint. High accuracy of reconstructed signals is obtained without the necessity of the electrical drive system parameters identification.

Effectiveness of Saliency-Based Methods in Optimization of Neural State Estimators of the Drive System With Elastic Couplings

This paper deals with the optimization problem of internal structure of neural-network (NN)-based state estimators of the drive system with elastic joints. Saliency-based optimization methods, developed in the NN theory, are tested in this paper to obtain the high-quality estimation of torsional torque and load-machine speed. The optimal brain damage and optimal brain surgeon methods are compared in this application. The state variables estimated by the optimized NN are used in feedback paths in the two-mass drive structure with a state controller. The simulation results show good accuracy of both presented neural estimators for a wide range of changes of the reference speed and the load torque. The simulation results are verified by laboratory experiments.

Optimization of neural state estimators of the two-mass system using OBD method

This paper presents neural estimators of the state variable for drive system with elastic joints. The main stages of the design methodology of neural estimators of the torsional torque and the load machine speed were presented. For the optimization of the structure of each neural networks the Optimal Brain Damage method was implemented. The signals estimated by neural networks were tested in the control structure with additional feedbacks from the shaft torque and the difference between the motor and the load speeds. The simulation results show good accuracy of both presented neural estimators for the wide range of changes of the reference speed and the load torque. The simulation results are verified by laboratory experiments.

Application of MLP and RBF neural networks in the control structure of the drive system with elastic joint

Purpose – The purpose of this paper is to obtain an estimation of not measured mechanical state variables of the drive system with elastic coupling between the driven motor and a load machine, using neural networks (NN) of different type for the sensorless drive system.Design/methodology/approach – The load‐side speed and the torsional torque are estimated using multi‐layer perceptron (MLP) and radial basis function (RBF) networks. The special forms of input vectors for neural state estimators were proposed and tested in open‐ and closed‐loop control structure. The estimation quality as well as sensitivity of neural estimators to the changes of the inertia moment of the load machine were evaluated and compared.Findings – It is shown that an application of RBF‐based neural estimators can give better accuracy of the load speed and torsional torque estimation, especially for the proper choice of the input vector of NN, also in the case of a big change of the load machine time constant.Research limitations/im...