EPE Association: EPE 1999 - Topic 05e: Estimation Techniques

EPE 1999 - Topic 05e: Estimation Techniques
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1999 - Conference > EPE 1999 - Topic 05: ELECTRICAL MACHINES AND ADJUSTABLE SPEED DRIVES > EPE 1999 - Topic 05e: Estimation Techniques

   [return to parent folder]

   A Stator Resistance Estimator for Sensorless Vector Controlled Drives Using ANN
By M. Sumner; J. Campbell; M. Curtis [View]
[Download]

Abstract: This paper describes the use of an Artificial Neural Network (ANN) to provide an independent estimate of the stator resistance variation in an induction motor. The design and training of the ANN is outlined, and verification through simulation is described. The performance of the ANN is then compared to the performance of a stator resistance estimator which uses a simple first order curve fitting routine. Experimental results show that the ANN based system provides a superior resistance estimate.

   An Autotuning Method for Vector Controlled Induction Motor Drives...
By S-i. Hashimoto; K. Tanaka; I. Miki [View]
[Download]

Abstract: This paper presents an auto-tuning method for vector controlled induction motor drives considering the stator core loss for the start up. Effect of the stator core loss of the induction motor, which appears in the transient motor terminal voltage caused by the undesirable rotor flux variation, must be considered to obtain accurate electrical motor parameters that are needed to achieve the instantaneous torque control. The equivalent motor model is based on the assumption that the stator core loss occurs due to the eddy current of the motor stator, and the tuning algorithm uses the rotor flux variation for a stepwise torque command. The proposed method can obtain the gains used in the vector control system, and the validity of this method is confirmed by experiment.

   An Experimental Study of the Synchonizing Condition to the Linear...
By M. Kajiok; K’i. Oka; M. Maruyama; M. Watada; S. Torii; D. Ebihara; T. Shinya; M. Karita [View]
[Download]

Abstract: Linear synchronous motor (LSM) is used for transportation system such as parts transfer, material transfer. Stator of the LSM is usually arrangement continuously. As long stroke transportation system may not needed to constant speed operation, authors proposed the discontinuous arrangement applied to the Long Stator LSM. One of technical problem of this system is synchronization of speed at each discontinuous stator. In this paper, the authors present the method of the synchronization and experimentally ascertained of the synchronization.

   Comparison Between Two Analytical Determinations of Reduced-order Observer for Induction Machine Rotor Flux
By E. Delmotte; B. Vulturescu; A. Bouscayrol; B. Semail [View]
[Download]

Abstract: In the case of the vector control of an induction machine, two analytical methods are compared to find the gain of determinist reduced-order observer for the rotor flux: one considers the observer dynamics in the continuous-time domain; the other takes into account parameter variations in the discrete-time domain. Experimental ersts and simulations are performed in order to compare the dynamics and robustness of flux estimation.

   Design of a Kalman Filter for Direct Mean Torque Control
By E. Flach [View]
[Download]

Abstract: Direct Mean Torque Control (DMTC) combines a good dynamic performance of Direct Torque Control (DTC) with the advantages of time equidistant control algorithms for digital implementation. DMTC requires a predictive machine model. In order to correct the predicted states, the model can be designed as a state observer. One way to contrive the feedback is a Kalman filter. the computation time can significantly be reduced by using a steady-state Kalman filter, i.e. a Wiener-Bucy filter. This paper deals with an off-line approach of such a steady-state filter. Experimental results confirm the validity of the scheme.

   Estimation of Mechanical Parameters of Induction Motor Using Neural Networks Principles
By D. Balara; J. Timko [View]
[Download]

Abstract: A method for estimation of mechanical parameters of induction motor is presented in this paper. Adaptive model of mechanical subsystem of motor and simple adaptation rule are presented as well as results of simulation using model of squirrel-cage induction motor. Adaptation rule was derived based on the gradient descent algorithm widely used in training of Artificial Neural Networks. Currents and fluxes of stator and rotor used in estimation were supposed to be either measured or observed.

   Identification of a small Power Induction Machine.
By C. Millet; J. Pierquin; C. Bergmann [View]
[Download]

Abstract: We realised a non-linear decoupling structure in a field-oriented control with a small power induction machine. First, parameters identification was established with modern experimental tools to establish a reference vector of our induction machine electrical parameters in order to synthesise an experimental no linear decoupling law of the dq current loops [4]. In a second time, these parameters were used to be compare with others values obtained with other identification test limits. These experimental tests, after stator resistance measure, propose a global identification parameter with only one test more speed than our first classical parameter identification.

   Neural Network Based Estimation of the Flux Space Vector Angle for an Indirect Field Oriented Control Independant form Rotor Resistance Variations
By A. Cataliotti; V. Cecconi; M. Cirrincione; A. Flaccomio [View]
[Download]

Abstract: The aim of the paper is to obtain an indirect field oriented control scheme independant from rotor resistance changes by means of an Artifial Neural-Network estimation of the rotor flux space vector angle. This estimation is based on two non-recursive supervised neural networks. The training phase has been carried out off-line using an informative enough training set. Drive system simulations have shown improved dynamical responses as a consequence of the robustness of such a neural estimation with regard to rotor resistance variations, resulting form the generalisation properties of supervised neural networks.

   Parameters Identification for Induction Machines at Standstill
By M. Zélia; P. Assis; P. F. Seixas [View]
[Download]

Abstract: This paper presents new procedures for electrical parameters estimation of induction motors based on its continuous-time model and by application of the recursive least-squares algorithm. The use of the continuous time model requires the calculation of the derivatives of current and voltage measured signals, which are determined using polynomial interpolation (spline-type) techniques. Three different linear regression forms are presented. The first one enables the simultaneous estimation of all the parameters of the model. The others consider some a priori knowledge. For each procedure, all the electrical parameters are obtained considering the equivalent machine concept that will be defined. The estimation procedures are applied in the self-comissioning phase, prior to the machine start-up. In order to avoid any machine intervention, only no-torque signals are used in the standstill tests. The conditions to produce these no-torque voltages with the PWM inverter, avoiding ripple torque generation, are analyzed. The paper includes experimental results, which confirm the good performance of the proposed procedures.

   Real-time Parameter Estimation of Vector-Control Induction Motor Fed by CR-PWM Using Extended Kalman Filter
By D. Ovidiu Kisck; S. Bucurenciu; M. Bucurenciu; G. Sirbu; M. Kisck [View]
[Download]

Abstract: For the control of high performance induction motor drives, one of the most used method is indirect rotor field oriented control which requires an accurate knowledge of the rotor resistance. A linear model of the induction motor becomes non-linear when the parameters change and it effects the dynamic performances of the drives. To estimate the rotor resistance, the paper presents a method which uses the Extended Kalman Filter. It is proposed that instead of using measured three-phase stator currents as measured variables to Extended Kalman Filter, the stator reference currents, to which white Gaussian noise is added, may be used. So that the 3/2 phase transformation and three low-pass filters can be eliminated. the proposed method is implemented with a fixed-point DSP-TMS320F240 and the digital simulation and experimental results carried out show its performances.

   Sensorless Brushless DC Motor with Improved Speed Estimation Accuracy Using Stator Resistance Estimation
By B. Terzic; M. Jadric [View]
[Download]

Abstract: This paper describes a method for simultaneous state variables and parameters estimation of a brushless DC motor. The extended Kalman filter is applied for motor state variables and parameters estimation using measured values of the stator line voltages and currents. The experimental results of the speed, rotor position and stator resistance estimation are presented. The results are obtained using two variants of the estimation algorithm. The first one deals with only state motor variables, and in the other one the motor variables and stator resistance are estimated simultaneously. By introducing the estimation of the stator resistance the speed estimation accuracy is increased, particularly at low speed.

   Simple Neural Cascade Architecture for Estimating of Stator and R...
By Lech Grzesiak [View]
[Download]

Abstract: We present a simple cascade neural architecture for stator and rotor flux vectors estimation. The system consists of two neural networks. The first one estimates stator flux from the stator voltage and current data, then the second one applies the stator current and already estimated stator flux to rotor flux estimation. the system is learning from examples in a non-traditional way. We use dynamic neural architecture where there is no need to set up number of hidden units in advance. The incremental learning process is very fast and takes place without local error minima.

   Stable Identification of Rotor Time Constant for IM Using FTA
By J. Kim; H. Kubota; K. Matsuse [View]
[Download]

Abstract: Various approaches to estimate a rotor time constant of induction motor have been using a PI controller as an adaptation controller, in which even small input error causes the divergence or oscillation of its controller when detected voltage and/or current have DC offeset and/or magnitude error. Therefore, new compensation method, which is mixed a reactive power method (an identification method of a rotor time constant) with a fixed trace algorithm (an adaptation controller), which has a characteristic of variable gain depending on the amount of its input, is presented and confirmed through simulations and experiments.

   The Selected Systems of the Magnetic Flux Vector and the Speed of the AC Motor Estimation
By A. Solbut; J. Werdoni; T. Citko [View]
[Download]

Abstract: The results of the computer simulations and the laboratory model investigation of the selected systems of the motor flux vector components and the angular speed reproduction by means of the traditional software and with the neural network application will be presented in the article. The Application Development System DSP96000ADSA Motorola Firm was used to the realisation of this problem in the real time.

EPE 1999 - Topic 05e: Estimation Techniques
You are here: EPE Documents > 01 - EPE & EPE ECCE Conference Proceedings > EPE 1999 - Conference > EPE 1999 - Topic 05: ELECTRICAL MACHINES AND ADJUSTABLE SPEED DRIVES > EPE 1999 - Topic 05e: Estimation Techniques
	[return to parent folder]

	A Stator Resistance Estimator for Sensorless Vector Controlled Drives Using ANN By M. Sumner; J. Campbell; M. Curtis	[View] [Download]
Abstract: This paper describes the use of an Artificial Neural Network (ANN) to provide an independent estimate of the stator resistance variation in an induction motor. The design and training of the ANN is outlined, and verification through simulation is described. The performance of the ANN is then compared to the performance of a stator resistance estimator which uses a simple first order curve fitting routine. Experimental results show that the ANN based system provides a superior resistance estimate.

	An Autotuning Method for Vector Controlled Induction Motor Drives... By S-i. Hashimoto; K. Tanaka; I. Miki	[View] [Download]
Abstract: This paper presents an auto-tuning method for vector controlled induction motor drives considering the stator core loss for the start up. Effect of the stator core loss of the induction motor, which appears in the transient motor terminal voltage caused by the undesirable rotor flux variation, must be considered to obtain accurate electrical motor parameters that are needed to achieve the instantaneous torque control. The equivalent motor model is based on the assumption that the stator core loss occurs due to the eddy current of the motor stator, and the tuning algorithm uses the rotor flux variation for a stepwise torque command. The proposed method can obtain the gains used in the vector control system, and the validity of this method is confirmed by experiment.

	An Experimental Study of the Synchonizing Condition to the Linear... By M. Kajiok; K’i. Oka; M. Maruyama; M. Watada; S. Torii; D. Ebihara; T. Shinya; M. Karita	[View] [Download]
Abstract: Linear synchronous motor (LSM) is used for transportation system such as parts transfer, material transfer. Stator of the LSM is usually arrangement continuously. As long stroke transportation system may not needed to constant speed operation, authors proposed the discontinuous arrangement applied to the Long Stator LSM. One of technical problem of this system is synchronization of speed at each discontinuous stator. In this paper, the authors present the method of the synchronization and experimentally ascertained of the synchronization.

	Comparison Between Two Analytical Determinations of Reduced-order Observer for Induction Machine Rotor Flux By E. Delmotte; B. Vulturescu; A. Bouscayrol; B. Semail	[View] [Download]
Abstract: In the case of the vector control of an induction machine, two analytical methods are compared to find the gain of determinist reduced-order observer for the rotor flux: one considers the observer dynamics in the continuous-time domain; the other takes into account parameter variations in the discrete-time domain. Experimental ersts and simulations are performed in order to compare the dynamics and robustness of flux estimation.

	Design of a Kalman Filter for Direct Mean Torque Control By E. Flach	[View] [Download]
Abstract: Direct Mean Torque Control (DMTC) combines a good dynamic performance of Direct Torque Control (DTC) with the advantages of time equidistant control algorithms for digital implementation. DMTC requires a predictive machine model. In order to correct the predicted states, the model can be designed as a state observer. One way to contrive the feedback is a Kalman filter. the computation time can significantly be reduced by using a steady-state Kalman filter, i.e. a Wiener-Bucy filter. This paper deals with an off-line approach of such a steady-state filter. Experimental results confirm the validity of the scheme.

	Estimation of Mechanical Parameters of Induction Motor Using Neural Networks Principles By D. Balara; J. Timko	[View] [Download]
Abstract: A method for estimation of mechanical parameters of induction motor is presented in this paper. Adaptive model of mechanical subsystem of motor and simple adaptation rule are presented as well as results of simulation using model of squirrel-cage induction motor. Adaptation rule was derived based on the gradient descent algorithm widely used in training of Artificial Neural Networks. Currents and fluxes of stator and rotor used in estimation were supposed to be either measured or observed.

	Identification of a small Power Induction Machine. By C. Millet; J. Pierquin; C. Bergmann	[View] [Download]
Abstract: We realised a non-linear decoupling structure in a field-oriented control with a small power induction machine. First, parameters identification was established with modern experimental tools to establish a reference vector of our induction machine electrical parameters in order to synthesise an experimental no linear decoupling law of the dq current loops [4]. In a second time, these parameters were used to be compare with others values obtained with other identification test limits. These experimental tests, after stator resistance measure, propose a global identification parameter with only one test more speed than our first classical parameter identification.

	Neural Network Based Estimation of the Flux Space Vector Angle for an Indirect Field Oriented Control Independant form Rotor Resistance Variations By A. Cataliotti; V. Cecconi; M. Cirrincione; A. Flaccomio	[View] [Download]
Abstract: The aim of the paper is to obtain an indirect field oriented control scheme independant from rotor resistance changes by means of an Artifial Neural-Network estimation of the rotor flux space vector angle. This estimation is based on two non-recursive supervised neural networks. The training phase has been carried out off-line using an informative enough training set. Drive system simulations have shown improved dynamical responses as a consequence of the robustness of such a neural estimation with regard to rotor resistance variations, resulting form the generalisation properties of supervised neural networks.

	Parameters Identification for Induction Machines at Standstill By M. Zélia; P. Assis; P. F. Seixas	[View] [Download]
Abstract: This paper presents new procedures for electrical parameters estimation of induction motors based on its continuous-time model and by application of the recursive least-squares algorithm. The use of the continuous time model requires the calculation of the derivatives of current and voltage measured signals, which are determined using polynomial interpolation (spline-type) techniques. Three different linear regression forms are presented. The first one enables the simultaneous estimation of all the parameters of the model. The others consider some a priori knowledge. For each procedure, all the electrical parameters are obtained considering the equivalent machine concept that will be defined. The estimation procedures are applied in the self-comissioning phase, prior to the machine start-up. In order to avoid any machine intervention, only no-torque signals are used in the standstill tests. The conditions to produce these no-torque voltages with the PWM inverter, avoiding ripple torque generation, are analyzed. The paper includes experimental results, which confirm the good performance of the proposed procedures.

	Real-time Parameter Estimation of Vector-Control Induction Motor Fed by CR-PWM Using Extended Kalman Filter By D. Ovidiu Kisck; S. Bucurenciu; M. Bucurenciu; G. Sirbu; M. Kisck	[View] [Download]
Abstract: For the control of high performance induction motor drives, one of the most used method is indirect rotor field oriented control which requires an accurate knowledge of the rotor resistance. A linear model of the induction motor becomes non-linear when the parameters change and it effects the dynamic performances of the drives. To estimate the rotor resistance, the paper presents a method which uses the Extended Kalman Filter. It is proposed that instead of using measured three-phase stator currents as measured variables to Extended Kalman Filter, the stator reference currents, to which white Gaussian noise is added, may be used. So that the 3/2 phase transformation and three low-pass filters can be eliminated. the proposed method is implemented with a fixed-point DSP-TMS320F240 and the digital simulation and experimental results carried out show its performances.

	Sensorless Brushless DC Motor with Improved Speed Estimation Accuracy Using Stator Resistance Estimation By B. Terzic; M. Jadric	[View] [Download]
Abstract: This paper describes a method for simultaneous state variables and parameters estimation of a brushless DC motor. The extended Kalman filter is applied for motor state variables and parameters estimation using measured values of the stator line voltages and currents. The experimental results of the speed, rotor position and stator resistance estimation are presented. The results are obtained using two variants of the estimation algorithm. The first one deals with only state motor variables, and in the other one the motor variables and stator resistance are estimated simultaneously. By introducing the estimation of the stator resistance the speed estimation accuracy is increased, particularly at low speed.

	Simple Neural Cascade Architecture for Estimating of Stator and R... By Lech Grzesiak	[View] [Download]
Abstract: We present a simple cascade neural architecture for stator and rotor flux vectors estimation. The system consists of two neural networks. The first one estimates stator flux from the stator voltage and current data, then the second one applies the stator current and already estimated stator flux to rotor flux estimation. the system is learning from examples in a non-traditional way. We use dynamic neural architecture where there is no need to set up number of hidden units in advance. The incremental learning process is very fast and takes place without local error minima.

	Stable Identification of Rotor Time Constant for IM Using FTA By J. Kim; H. Kubota; K. Matsuse	[View] [Download]
Abstract: Various approaches to estimate a rotor time constant of induction motor have been using a PI controller as an adaptation controller, in which even small input error causes the divergence or oscillation of its controller when detected voltage and/or current have DC offeset and/or magnitude error. Therefore, new compensation method, which is mixed a reactive power method (an identification method of a rotor time constant) with a fixed trace algorithm (an adaptation controller), which has a characteristic of variable gain depending on the amount of its input, is presented and confirmed through simulations and experiments.

	The Selected Systems of the Magnetic Flux Vector and the Speed of the AC Motor Estimation By A. Solbut; J. Werdoni; T. Citko	[View] [Download]
Abstract: The results of the computer simulations and the laboratory model investigation of the selected systems of the motor flux vector components and the angular speed reproduction by means of the traditional software and with the neural network application will be presented in the article. The Application Development System DSP96000ADSA Motorola Firm was used to the realisation of this problem in the real time.