METHOD OF ESTIMATING THE STATE OF A SYSTEM AND RELATIVE DEVICE FOR ESTIMATING POSITION AND SPEED OF THE ROTOR OF A BRUSHLESS MOTOR

• US 20090033259A1
• Filed: 08/02/2007
• Published: 02/05/2009
• Est. Priority Date: 08/03/2006
• Status: Active Grant

• Current Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Original Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Inventors: Di Meglio, Maurizio, Cesario, Nicola, Baratto, Stefano
• Granted Patent: US 7,659,685 B2
• Time in Patent Office: Days
• US Class Current: 318/400.40
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: H02P 21/13 Observer control, e.g. using Luenberger observers or Kalman filters
• CPC Class Codes: H02P 21/18 Estimation of position or speed
• CPC Class Codes: H02P 21/24 Vector control not involving the use of rotor position or rotor speed sensors

Abstract

A method for estimating the state of a system, as well as an extended Kalman filter (EKF), allows nonlinear mathematical models to be used for describing the system. Accurate precision is provided since the method is based on an EKF technique while using a filter that implements a first degree Stirling approximation formula. The method may be used for estimating position and speed of a brushless motor, and may be implemented in a relative device. Such a device may be introduced in a control loop of a brushless motor of a power steering system for a vehicle to provide a countering torque on the steering wheel based on speed of the vehicle and a steering angle of the vehicle.

Claims

1-11. -11. (canceled)

12. A method for estimating state variables of a system having a known input-state-output mathematical model, the method comprising:

installing an estimator circuit for estimating state variables of the system, the estimator circuit comprising a filter for implementing an estimation according to a Stirling approximation formula as a function of input and output data of the system and of filter parameters representing covariance matrices of process noise and of measurement noise of the system; and

training the estimator circuit by performing the following off-line

supplying the system with input patterns, and detecting corresponding output data and at least one state variable of the system,

estimating the at least one state variable of the system for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the state of the system during normal functioning using the off-line trained estimator circuit.

13. The method of claim 12, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

14. The method of claim 12, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.

15. The method of claim 12, wherein the filter comprises at least one of a first order filter and a second order filter.

16. A method for estimating position and speed of a rotor of a brushless motor controlled in a feedback loop comprising a circuit for generating feedback signals representing currents circulating in the brushless motor with respect to a rotoric reference system as a function of currents sensed in windings of the brushless motor, a controller receiving as input respective reference signals and generating respective control voltages with respect to the rotoric reference system, and a power stage receiving as input the respective control voltages, and driving the brushless motor in a PWM mode for nullifying a difference between the feedback signals and the respective reference signals, the method comprising executing at each PWM cycle the following:

sampling the respective control voltages;

calculating currents circulating in the brushless motor with respect to the rotoric reference as a function of an estimated position of the rotor at a current PWM cycle; and

estimating the position and speed of the rotor at a next PWM cycle as a function of the sampled respective control voltages and of the calculated currents, the estimating comprising

implementing an estimation according to a Stirling approximation formula as a function of input and output data of the brushless motor and of filter parameters representing covariance matrices of process noise and of measurement noise of the brushless motor, and

training the estimator circuit by performing the following off-line

supplying the brushless motor with input patterns, and detecting corresponding output data and at least one state variable of the brushless motor,

estimating the at least one state variable of the brushless motor for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the position and speed of the rotor during normal functioning using the off-line trained estimator circuit.

17. The method of claim 16, wherein the calculated currents are obtained from the sensed current flowing through the windings of the brushless motor based on a Clarke transform followed by a Park transform defined as a function of the estimated position of the rotor at the current PWM cycle.

18. The method of claim 16, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

19. The method of claim 16, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.

20. The method of claim 16, wherein the estimator circuit comprises a filter comprising at least one first order filter and a second order filter.

21. A device for generating signals representative of speed and position of a rotor of a brushless motor controlled in a feedback loop that includes a circuit for generating feedback signals representing currents circulating in the brushless motor with respect to a rotoric reference system as a function of currents sensed in windings of the brushless motor;

a controller receiving as input respective reference signals and generating respective control voltages with respect to the rotoric reference system;

a power stage receiving as input the respective control voltages, and driving the brushless motor in a PWM mode for nullifying a difference between the feedback signals and the respective reference signals;

the device comprising;

a filter for estimating the position and speed of the rotor at a next PWM cycle as a function of the sampled respective control voltages and of the calculated currents, the estimating comprising

implementing an estimation according to a Stirling approximation formula as a function of input and output data of the brushless motor and of filter parameters representing covariance matrices of process noise and of measurement noise of the brushless motor, and

training said filter by performing the following off-line

supplying the brushless motor with input patterns, and detecting corresponding output data and at least one state variable of the brushless motor,

estimating the at least one state variable of the brushless motor for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the position and speed of the rotor during normal functioning using said off-line trained filter.

22. The device of claim 21, wherein the calculated currents are obtained from the sensed current flowing through the windings of the brushless motor based on a Clarke transform followed by a Park transform defined as a function of the estimated position of the rotor at the current PWM cycle.

23. The device of claim 21, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

24. The device of claim 21, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.

25. The device of claim 21, wherein said filter comprises at least one first order filter and a second order filter.

26. A device for feedback controlling a brushless motor comprising:

a circuit for generating feedback signals representing currents circulating in the brushless motor with respect to a rotoric reference system as a function of currents sensed in windings of the brushless motor;

a controller for receiving as input respective reference signals and generating respective control voltages with respect to the rotoric reference system;

a power stage for receiving as input the respective control voltages, and driving the brushless motor in a PWM mode for nullifying a difference between the feedback signals and the respective reference signals; and

a filter for estimating the position and speed of the rotor at a next PWM cycle as a function of the sampled respective control voltages and of the calculated currents, the estimating comprising

implementing an estimation according to a Stirling approximation formula as a function of input and output data of the brushless motor and of filter parameters representing covariance matrices of process noise and of measurement noise of the brushless motor, and

training said filter by performing the following off-line

supplying the brushless motor with input patterns, and detecting corresponding output data and at least one state variable of the brushless motor,

estimating the at least one state variable of the brushless motor for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the position and speed of the rotor during normal functioning using said off-line trained filter.

27. The device of claim 26, wherein the calculated currents are obtained from the sensed current flowing through the windings of the brushless motor based on a Clarke transform followed by a Park transform defined as a function of the estimated position of the rotor at the current PWM cycle.

28. The device of claim 26, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

29. The device of claim 26, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.

30. The device of claim 26, wherein said filter comprises at least one first order filter and a second order filter.

31. A power steering system for a vehicle comprising:

a brushless motor for generating an auxiliary steering torque as a function of control signals received as input;

a sensor for generating a signal representing a total steering torque; and

a control circuit for said brushless motor comprising

a circuit for generating feedback signals representing currents circulating in said brushless motor with respect to a rotoric reference system as a function of currents sensed in windings of said brushless motor,

a controller for receiving as input respective reference signals and generating respective control voltages with respect to the rotoric reference system,

a power stage for receiving as input the respective control voltages, and driving said brushless motor in a PWM mode for nullifying a difference between the feedback signals and the respective reference signals, and

a filter for estimating the position and speed of the rotor at a next PWM cycle as a function of the sampled respective control voltages and of the calculated currents, the estimating comprising

implementing an estimation according to a Stirling approximation formula as a function of input and output data of said brushless motor and of filter parameters representing covariance matrices of process noise and of measurement noise of said brushless motor, and

training said filter by performing the following off-line

supplying said brushless motor with input patterns, and detecting corresponding output data and at least one state variable of said brushless motor,

estimating the at least one state variable of said brushless motor for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the position and speed of the rotor during normal functioning using said off-line trained filter.

32. The power steering system of claim 31, further comprising a circuit for generating the respective reference signals as a function of the total steering torque.

33. The power steering system of claim 31, further comprising a circuit for establishing the respective reference values as a function of vehicle speed and a steering angle of the vehicle.

34. The power steering system of claim 31, wherein the calculated currents are obtained from the sensed current flowing through the windings of said brushless motor based on a Clarke transform followed by a Park transform defined as a function of the estimated position of the rotor at the current PWM cycle.

35. The power steering system of claim 31, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

36. The power steering system of claim 31, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.

37. The power steering system of claim 31, wherein said filter comprises at least one first order filter and a second order filter.

38. A computer-readable medium having computer-executable instructions for causing a computer to perform steps comprising:

estimating state variables of a system using an estimator circuit, the estimator circuit comprising a filter for implementing an estimation according to a Stirling approximation formula as a function of input and output data of the system and of filter parameters representing covariance matrices of process noise and of measurement noise of the system; and

training the estimator circuit by performing the following off-line

supplying the system with input patterns, and detecting corresponding output data and at least one state variable of the system,

estimating the at least one state variable of the system for the input patterns,

adjusting the filter parameters with an evolutionary algorithm for reducing a cost function defined as a function of estimated output data, of sensed output data, of the at least one estimated state variable and of a corresponding at least one sensed state variable, and

estimating the state of the system during normal functioning using the off-line trained estimator circuit.

39. The computer-readable medium of claim 38, wherein the cost function is based on a root means square of a difference between the estimated output data and the sensed output data, and between the at least one estimated state variable and the corresponding at least one sensed state variable.

40. The computer-readable medium of claim 38, wherein the evolutionary algorithm comprises at least one of a particle swarm optimization algorithm (PSOA) and evolution strategy (ES) algorithms.