Permanent magnet synchronous linear motor mover mass decoupling identification method

Abstract

The application discloses a kind of permanent magnet synchronous linear motor moving mass decoupling identification method, to solve the problem that traditional method is limited to specific excitation signal in identification process, and there is coupling with other parameters, nonlinear friction force influence etc..First, linear extended state observer (LESO) is constructed to obtain extended state variable, then it is modified;Second, state expansion extended Kalman filter (SE-EKF) algorithm is designed, with moving mass and linear velocity as state variable, to calculate prior estimation result;Then, through covariance matrix, identification gain calculation, the result of prior estimation is modified using gain, and finally the identified moving mass value can be obtained.

Description

Technical Field

[0001] This invention relates to the field of permanent magnet synchronous linear motor technology, and specifically to a method for decoupling and identifying the mass of a permanent magnet synchronous linear motor mover. Background Technology

[0002] Permanent magnet synchronous linear motors (PMSLMs) are widely used in servo control due to their high thrust density and fast dynamic response. In practical applications, tuning the control loop coefficients of a servo system often requires information about the mover mass; inaccurate parameter values ​​can lead to decreased control performance. Furthermore, the mover mass of a PMSLM also affects the motor's speed response. Against this backdrop, various mover mass identification algorithms have been proposed.

[0003] Traditional mover mass identification algorithms are primarily based on the Coulomb-viscous friction model and can be categorized into offline and online methods. Offline algorithms require injecting specific excitation signals into the motor's control loop, utilizing the signal's periodic symmetry and orthogonality to identify the mover mass. However, offline algorithms are highly sensitive to external disturbances and limited by the excitation signal, requiring multiple runs and post-processing of the data, resulting in generally low accuracy and efficiency. Online identification algorithms, on the other hand, identify mechanical parameters based on data measured during motor operation, without requiring specific excitation signals. However, online identification typically identifies multiple parameters at once, leading to coupling between these parameters. Furthermore, inaccuracies in the friction model within the mover motion equations can also result in significant errors. In addition, both offline and online identification methods often neglect parameters coupled with the mover mass in the motion equations for the sake of algorithmic convenience, making it difficult to guarantee accuracy. Summary of the Invention

[0004] To address the shortcomings and deficiencies of existing technologies and to quickly and accurately identify the mover mass, this invention proposes a decoupled identification method for the mover mass of a permanent magnet synchronous linear motor. This method solves the problems of traditional methods being limited by specific excitation signals, coupling with other parameters, and the influence of nonlinear friction during the identification process. First, a Linear Extended State Observer (LESO) is constructed to obtain the extended state variables, which are then corrected. Second, an Extended Kalman Filter (SE-EKF) algorithm for state extension is designed, using the mover mass and linear velocity as state variables to calculate the prior estimation results. Then, after calculating the covariance matrix and identification gain, the prior estimation results are corrected using the gain, ultimately yielding the identified mover mass value.

[0005] To achieve the above objectives, the present invention specifically adopts the following technical solution:

[0006] A method for decoupling and identifying the mover mass of a permanent magnet synchronous linear motor is characterized by: constructing a linear extended state observer (LESO) to obtain extended state variables, and then correcting them; using the extended Kalman filter (SE-EKF) algorithm for state extension, with the mover mass and linear velocity as state variables, calculating the prior estimation results; then, after calculating the covariance matrix and identification gain, using the gain to correct the prior estimation results, finally obtaining the identified mover mass value.

[0007] Furthermore, the specific steps include:

[0008] Step S1: Taking the position p of the mover as the state variable of the system, consider the frictional force F. fric Influenced by the motion of the permanent magnet synchronous linear motor, the motion equations of the mover are established:

[0009]

[0010] In the formula, M is the mass of the mover, and K f i is the electromagnetic thrust coefficient. q For the q-axis current, in a surface-mounted permanent magnet synchronous linear motor, the electromagnetic thrust F is satisfied. e =K f i q The ".." symbol above the variable indicates a second derivative;

[0011] Step S2: Define the state f that needs to be expanded according to equation (1). se :

[0012]

[0013] In the formula, Initial values ​​set for the mover mass identification algorithm;

[0014] Step S3: Expand the state term f se As the third extended state in addition to the mover displacement p and linear velocity v, construct the linear extended state observer of the system:

[0015]

[0016] In the formula, These are the observed values ​​of mover displacement, mover linear velocity, and expansion state, respectively; β1, β2, and β3 are... The corresponding observation gain, the "." symbol above the variable indicates the first derivative;

[0017] Step S4: Adjust the expanded observations The correction yields a new state term:

[0018]

[0019] In the formula, Let the output result of the mover mass identification algorithm be set; combining equation (2), equation (4) is further approximated as follows:

[0020]

[0021] Step S5: Based on formula (2), using the mass M and linear velocity v of the mover as state variables, construct the discrete prior estimation equation:

[0022]

[0023] In the formula, T s The algorithm is discrete; the variable at time k is labeled with "k"; the prior state variable matrix is ​​defined.

[0024] Step S6: Calculate the covariance matrix:

[0025] P k / k-1 (k)=D(k-1)P k (k-1)D T (k-1)+Q (7)

[0026] In the formula, P k / k-1 To estimate the covariance matrix a priori; P k Q is the posterior correction covariance matrix; Q is the model noise covariance matrix; and D is the partial derivative matrix of the state equation.

[0027] Step S7: Identify gain calculation:

[0028] G(k)=P k / k-1 (k)H T HP k / k-1 (k)H T +R] -1 (8)

[0029] In the formula, G is the identification gain matrix; H =

[01] is the partial derivative matrix of the state variables; R is the measurement noise covariance matrix;

[0030] Step S8: Using the identification gain G to correct the equation (6) P of equation (7) k / k-1 :

[0031]

[0032] Step S9: Read the posterior-corrected state variable matrix The decoupled mover mass identification results are obtained:

[0033]

[0034] Furthermore, the current loop employs a PI control algorithm, and the position loop is designed based on the characteristics of the mover mass identification algorithm, where the feedback signals are the observed state variables of LESO at each order; the input signal of the mover mass identification algorithm is the q-axis current i. q q-axis given current mover displacement p, mover given displacement p * And the linear velocity v of the mover; where the q-axis current is obtained by converting the three-phase current collected by the current sensor, and the mover displacement is obtained by measuring the high-precision grating ruler.

[0035] Furthermore, the state observation algorithm consists of two parts: LESO and state correction; firstly, the mover displacement p and displacement observation are performed. The difference is used for gain and integration to obtain the extended state observation. Based on this, linear velocity observations were further obtained. and displacement observation In this system, each order of observation states is used to form a feedback closed loop for position control; and The middle part represents the mass M of the moving part and the frictional force F. fric Electromagnetic thrust F e Initial values ​​for quality identification The coupling information is used, therefore electromagnetic thrust, initial identification value, and identification results from the SE-EKF algorithm are introduced. right Make corrections to obtain the corrected state items. After state correction, the mass of the moving part is only related to the frictional force F. fric There is coupling; in order to The motioner mass contained therein is completely decoupled and identified using the decoupled SE-EKF algorithm;

[0036] The SE-EKF algorithm first uses the mover mass M and linear velocity v as state variables, and then utilizes the electromagnetic thrust F at time k. e , And the posterior correction matrix of the previous time step The prior estimates of M and v are calculated, and together they constitute the prior state variable matrix. On the other hand, using F e , Calculate the partial derivative matrix D(k-1) of the state equation, and combine it with the posterior corrected covariance matrix P from the previous time step. k (k-1) The prior estimated covariance matrix P is calculated from the model noise covariance matrix Q. k / k-1(k); then, combining the partial derivative matrix H of the state variables and the covariance matrix R of the measurement noise, calculate the gain matrix G(k) for identification update; then, multiply the information obtained by subtracting the prior estimate of the linear velocity from the actual value of the linear velocity with G(k) to calculate the correction amount for identification; based on the correction amount, adjust the prior state variable matrix... After making corrections, we obtain the posterior correction matrix at the current time step. And the covariance matrix P k / k-1 After correction, the posterior corrected covariance matrix P at the current time is obtained. k (k), then obtained and P k (k is used in the identification iteration of the next time step, eventually causing the algorithm to converge to a steady state; where the posterior correction matrix generated in each algorithm iteration) Each value contains identification values ​​for the mover mass and linear velocity. The identification value for the mover mass is output as the identification result. To ensure the state correction in the state observation algorithm The accuracy of the identification results This feedback is then fed back into the state observation.

[0037] Furthermore, firstly, the DC bus voltage U is sampled. dc Waiting for U dc After the drive requirements are met, the relay is closed to start the motor; next, initialization steps including drive configuration, capacitor charging, and DC offset compensation for AD sampling are performed; then, the mover of the linear motor is positioned, and the precise position of the mover is obtained through the QEP module; finally, the operation control of the linear motor is performed.

[0038] During the algorithm's execution, firstly, the driver determines the position of the mover (p) and the given d-axis and q-axis voltages. Calculate the α and β axis voltages required for control. The switching signals of the inverter power transistors are obtained based on the duty cycle modulation algorithm, and then controlled by the isolation drive circuit to control the operation of the power switching transistors, thereby controlling the operation of the linear motor; secondly, the actual position p of the mover and the given position p are used. * q-axis current given The mover displacement was obtained through the observer stage. Linear velocity observation and extended state observation Secondly, regarding the observations Make corrections to obtain new state terms. Finally, Substituting into the SE-EKF algorithm, the mover mass of the permanent magnet synchronous linear motor is identified through four steps: prior estimation, covariance matrix calculation, updating the identification gain, and posterior correction of the state variables.

[0039] Compared with the prior art, the present invention and its preferred embodiments have the following beneficial effects:

[0040] (1) In the process of identifying the mass of the mover, the displacement control of the mover does not require a specific excitation signal (such as a symmetrical triangular wave, a specific sine signal, etc.);

[0041] (2) The identification of the mass of the mover is not affected by other coupling parameters (such as friction, electromagnetic thrust, etc.), avoiding errors caused by inaccurate mechanical motion equations and effectively ensuring identification accuracy.

[0042] (3) The influence of nonlinear friction is fully considered, and the coupling information of friction and motion mass can be accurately obtained even when the friction model is unknown. High-precision identification results can be obtained by further decoupling identification. Attached Figure Description

[0043] Figure 1 This is a control block diagram of the permanent magnet synchronous linear motor mover mass identification system according to an embodiment of the present invention;

[0044] Figure 2 This is a structural diagram of the state observation algorithm according to an embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of the SE-EKF algorithm according to an embodiment of the present invention;

[0046] Figure 4 This is a hardware structure diagram of the driver system according to an embodiment of the present invention. Detailed Implementation

[0047] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below for detailed explanation:

[0048] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0049] In this invention, a linear extended state observer (LESO) is first constructed to obtain the extended state variables, which are then corrected. Next, an extended Kalman filter (SE-EKF) algorithm for state extension is designed, using the mover mass and linear velocity as state variables to calculate the prior estimation results. Then, after calculating the covariance matrix and identification gain, the prior estimation results are corrected using the gain, ultimately yielding the identified mover mass value.

[0050] Specifically as follows:

[0051] The control block diagram of the permanent magnet synchronous linear motor mover mass identification system proposed in this invention is as follows: Figure 1 As shown, the current loop employs a PI control algorithm, while the position loop is designed based on the characteristics of the mover mass identification algorithm. The feedback signals are the observed state variables of LESO at each order. The input signal of the mover mass identification algorithm is the q-axis current i. q q-axis given current mover displacement p, mover given displacement p * And the linear velocity v of the mover. The q-axis current is obtained by converting the three-phase current acquired by the current sensor, and the mover displacement is measured by a high-precision grating ruler.

[0052] Figure 2 The diagram shows the structure of the state observation algorithm, which consists of two parts: LESO and state correction. First, the mover displacement p and displacement observation... The difference is used for gain and integration to obtain the extended state observation. Based on this, linear velocity observations were further obtained. and displacement observation In this system, each order of observation states is used to form a feedback closed loop for position control. The middle part represents the mass M of the moving part and the frictional force F. fric Electromagnetic thrust F e Initial values ​​for quality identification The coupling information is used, therefore electromagnetic thrust, initial identification value, and identification results from the SE-EKF algorithm are introduced. right Make corrections to obtain the corrected state items. After state correction, the mass of the moving part is only related to the frictional force F. fric There is coupling. In order to... The motioner mass contained therein is completely decoupled and identified, and a decoupled SE-EKF algorithm is further designed.

[0053] The SE-EKF algorithm structure diagram is as follows: Figure 3 As shown, "reshape" indicates matrix reshaping, and "inv" indicates matrix inversion. Initially, the mover mass M and linear velocity v are used as state variables. At time k, electromagnetic thrust F is applied... e , And the posterior correction matrix of the previous time step The prior estimates of M and v are calculated, and together they constitute the prior state variable matrix. On the other hand, using F e , Calculate the partial derivative matrix D(k-1) of the state equation, and combine it with the posterior corrected covariance matrix P from the previous time step. k (k-1) The prior estimated covariance matrix P is calculated from the model noise covariance matrix Q. k / k-1(k); then, combining the partial derivative matrix H of the state variables and the covariance matrix R of the measurement noise, calculate the gain matrix G(k) for identification update; then, multiply the information obtained by subtracting the prior estimate of the linear velocity from the actual value of the linear velocity with G(k) to calculate the correction amount for identification; based on the correction amount, adjust the prior state variable matrix... After making corrections, we obtain the posterior correction matrix at the current time step. And the covariance matrix P k / k-1 After correction, the posterior corrected covariance matrix P at the current time is obtained. k (k), then obtained and P k (k is used in the identification iteration of the next time step, eventually causing the algorithm to converge to a steady state; where the posterior correction matrix generated in each algorithm iteration) Each value contains identification values ​​for the mover mass and linear velocity. The identification value for the mover mass is output as the identification result. To ensure the state correction in the state observation algorithm The accuracy of the identification results This feedback is then fed back into the state observation.

[0054] like Figure 4 As shown, the overall control system of this invention includes: a DC bus voltage sampling circuit, a three-phase inverter, an isolation drive circuit, a three-phase winding current sampling circuit, a microprocessor, and a three-phase permanent magnet synchronous linear motor.

[0055] The front-end DC bus voltage can be provided by selecting a DC voltage source, or by using an AC power supply-isolation transformer-AC-rectifier bridge. In the three-phase inverter, the power transistors are IGBTs or MOSFETs with diodes connected in parallel, and the microprocessor is a high-performance DSP chip or a microcontroller. The three-phase winding current sampling circuit is composed of a current Hall sensor and an operational amplifier circuit, and the DC bus voltage sampling circuit is composed of a voltage Hall sensor and an operational amplifier. The selection of Hall sensors ensures electrical isolation between the main circuit and the control circuit, improving control reliability. The current and voltage sampling circuits can capture the weak electrical signals output by the circuit and then send them to the controller's A / D conversion module to become digital signals. The mover position detection circuit can be composed of a high-precision grating ruler and a level converter; the output pulse signal is sent to the controller's QEP module to obtain the position signal. According to the control method of this invention, firstly, the DC bus voltage U is sampled. dc Waiting for U dcAfter meeting the drive requirements, the relay is closed; secondly, initialization steps such as drive configuration, capacitor charging, and DC offset compensation for AD sampling are performed; thirdly, the linear motor's mover is positioned, and its precise position is obtained through the QEP module; finally, the linear motor's operation is controlled. During algorithm execution, firstly, the driver determines the mover's position based on its actual position p and the given d-axis and q-axis voltages. Calculate the α and β axis voltages required for control. The switching signals of the inverter power transistors are obtained based on the duty cycle modulation algorithm, and then controlled by the isolation drive circuit to control the operation of the power switching transistors, thereby controlling the operation of the linear motor; secondly, the actual position p of the mover and the given position p are used. * q-axis current given The mover displacement was obtained through the observer stage. Linear velocity observation and extended state observation Secondly, regarding the observations Make corrections to obtain new state terms. Finally, Substituting this into the SE-EKF algorithm, the mover mass of the permanent magnet synchronous linear motor is identified through four steps: prior estimation, covariance matrix calculation, updating the identification gain, and posterior correction of the state variables.

[0056] Based on the above design, the implementation steps of this invention can be summarized as follows:

[0057] (1) Taking the position p of the mover as the state variable of the system, consider the frictional force F. fric Influenced by the motion of the permanent magnet synchronous linear motor, the motion equations of the mover are established:

[0058]

[0059] In the formula, M is the mass of the mover, and K f i is the electromagnetic thrust coefficient. q For the q-axis current, in a surface-mounted permanent magnet synchronous linear motor, the electromagnetic thrust F is satisfied. e =K f i q The ".." symbol above the variable indicates the second derivative.

[0060] (2) Define the state f that needs to be expanded according to equation (1). se :

[0061]

[0062] In the formula, The initial values ​​set for the mover mass identification algorithm.

[0063] (3) Expand the state term f se As the third extended state in addition to the mover displacement p and linear velocity v, construct the linear extended state observer of the system:

[0064]

[0065] In the formula, These are the observed values ​​of mover displacement, mover linear velocity, and expansion state, respectively; β1, β2, and β3 are... The corresponding observation gain, the "." symbol above the variable indicates the first derivative.

[0066] (4) For the expanded observation term The correction yields a new state term:

[0067]

[0068] In the formula, Let the output result of the mover mass identification algorithm be set. Combining equation (2), equation (4) is further approximated as follows:

[0069]

[0070] (5) Based on formula (2), using the mass M and linear velocity v of the mover as state variables, construct the discrete prior estimation equation:

[0071]

[0072] In the formula, T s The algorithm is discrete; the variable at time k is labeled with "k"; the prior state variable matrix is ​​defined.

[0073] (6) Calculate the covariance matrix:

[0074] P k / k-1 (k)=D(k-1)P k (k-1)D T (k-1)+Q (7)

[0075] In the formula, P k / k-1 To estimate the covariance matrix a priori; P k denoted as posterior corrected covariance matrix; Q is the model noise covariance matrix; and D is the partial derivative matrix of the state equation.

[0076] (7) Identification gain calculation:

[0077] G(k)=P k / k-1 (k)H T HP k / k-1 (k)H T +R]-1 (8)

[0078] In the formula, G is the identification gain matrix; H =

[01] is the partial derivative matrix of the state variables; and R is the measurement noise covariance matrix.

[0079] (8) Using the identification gain G to correct formula (6) P of equation (7) k / k-1 :

[0080]

[0081] (9) Read the a posteriori-corrected state variable matrix The decoupled mover mass identification results are obtained:

[0082]

[0083] To enable those skilled in the art to further understand the present invention, the basic principles involved in the present invention are provided below:

[0084] 1. Mathematical Model of Permanent Magnet Synchronous Linear Motor Servo System

[0085] Assuming the air gap magnetic field of the motor is approximately sinusoidally distributed and neglecting magnetic saturation, the voltage model in a two-phase rotating coordinate system oriented by the mover magnetic field is as follows:

[0086]

[0087] In the formula, u d u q These are the d-axis and q-axis voltages, respectively; i d i q These are the d-axis and q-axis currents, respectively; R is the resistance of the moving phase winding; L d L q These are the d-axis and q-axis inductances, respectively; ψ f τ is the flux linkage of the permanent magnet; v is the linear velocity of the mover; τ is the pole pitch of the motor.

[0088] The electromagnetic thrust controlling the movement of the rotor of a permanent magnet synchronous linear motor can be expressed as:

[0089]

[0090] In the formula, F e For electromagnetic thrust of the mover; n p This represents the number of pole pairs of the motor.

[0091] During the motion of the mover, the relationship between the electromagnetic thrust, the mass of the mover, and the total frictional force is as follows:

[0092]

[0093] In the formula: M is the mass of the mover; p is the displacement of the mover; F fric This represents the total frictional force experienced by the mover during its motion. The "." and ".." symbols above the variables represent the first and second derivatives, respectively.

[0094] 2. Basic Principles of Motor Mass Decoupling Identification

[0095] 2.1 Observation of the expansion state of coupled mass and friction

[0096] Equation 3 can be simplified to:

[0097]

[0098] As can be seen from Equation 4, the actual mover mass exists simultaneously in F. e / M and F fric In the linear motor control process, the mover mass, electromagnetic thrust, and friction are all coupled. This is because the mover displacement p and mover thrust F are coupled during actual linear motor control. e These can be considered as known quantities, and the mover mass M and electromagnetic thrust F can be realized. e The decoupling, but M and frictional force F fric The unknown quantity is the friction force. Clearly, when using the state equation described by Equation 4 to identify the mover mass, the friction force must be known beforehand, resulting in a coupling between the observations of the mover mass and the friction force. To eliminate this coupling, a two-stage mover mass identification strategy is proposed: the first stage observes the extended state term f, which couples the mover mass and the friction force. se (See Formula 6 for the definition), and thereby establish a new state term for coupled frictional forces. (See Formula 12 for details); the second level utilizes state terms. The SE-EKF identification algorithm for the mass of the mover decoupled from friction is explained in detail below.

[0099] The identification algorithm requires setting an initial value for the mover mass, denoted as . Formula 4 can then be rewritten as follows:

[0100]

[0101] In Formula 5, the extended state term is defined as:

[0102]

[0103] As can be seen from Formula 6, the actual value of the mover mass M is implicit in the extended state term f. se In the middle, the actual value of the mover mass M and the initial value of the mover mass identification are... Electromagnetic thrust F e and total friction force F fric There is coupling. The extended state term f...se As a third extended state in addition to the mover displacement p and linear velocity v, a linear extended state observer of the second-order system can be constructed:

[0104]

[0105] In the formula, β1, β2, and β3 are the observer gains, which are then multiplied by the observer bandwidth ω. o By relating them, we obtain the gain matrix.

[0106] Further utilizing the observation information from Formula 7 to compensate for the control, the position loop output is designed as follows:

[0107]

[0108] In the formula, K p K v These are the position and speed control coefficients, respectively.

[0109] Substituting the control signal from Equation 8 into Equation 5, we get:

[0110]

[0111] In the formula, e1, e2, and e3 represent the mover position p, the mover linear velocity v, and the extended state term f, respectively. se The observation error, and satisfying:

[0112]

[0113] As shown in Equation 9, the displacement control of the system will be affected by the presence of observation errors. Once the system reaches steady state, assuming... Bounded and satisfied The range of observation error is as follows:

[0114]

[0115] From Formula 11, we know that as long as ω o If the error is large enough, the observation errors of each order can converge to a smaller value, thereby achieving accurate observation of Formula 6.

[0116] Under the condition that the observations are accurate, the expanded observation term The correction yields a new state term:

[0117]

[0118] If we define the extended state term f in formula 6 se Observations Substituting into Formula 12, Formula 12 is further approximated as follows:

[0119]

[0120] 2.2 Decoupling identification of the mover mass for decoupling friction force

[0121] It is evident that the coupled frictional force exists only in Equation 13. Therefore, based on the state equation in Equation 5, an extended Kalman filter algorithm with state expansion is designed to completely decouple and identify the mover mass. The specific steps are as follows:

[0122] (1) According to Formula 5, using the mass M and linear velocity v of the mover as state variables, the discretized prior estimation equation is obtained:

[0123]

[0124] (2) Calculate the covariance matrix used to update the identification gain:

[0125] P k / k-1 (k)=D(k-1)P k (k-1)D T (k-1)+Q (Formula 15)

[0126] In the formula, D is the partial derivative matrix of the state equation, expressed as:

[0127]

[0128] (3) Identification gain calculation:

[0129] G(k)=P k / k-1 (k)H T HP k / k-1 (k)H T +R] -1 (Formula 17)

[0130] (4) Use the identification gain G to correct the prior state variable matrix of Formula 14. The covariance matrix P of Formula 15 k / k-1 The corrected state variable matrix Output mover mass identification results:

[0131]

[0132]

[0133] SE-EKF introduces a modified state in the prior estimation. Replace the original estimate Not only did it solve the problem of friction force F when the friction force model was unknown. fricThe inability to accurately obtain the information also reduces the computational load of prior estimation of state variables, effectively improving the speed and accuracy of the identification algorithm.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0135] This patent is not limited to the above-described preferred embodiments. Anyone can derive other forms of permanent magnet synchronous linear motor mover mass decoupling identification methods under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A permanent magnet synchronous linear motor moving mass decoupling identification method, characterized in that: The extended state variables are obtained by constructing a linear extended state observer (LESO) and then corrected. The prior estimation results are calculated by using the extended Kalman filter (SE-EKF) algorithm for state extension, with the mover mass and linear velocity as state variables. Then, after calculating the covariance matrix and identification gain, the results of the prior estimation are corrected using the gain, and finally the identified mover mass value is obtained. Specifically, the following steps are included: Step S1: Taking the mover position p as the state variable of the system, considering the influence of friction , the motion equation of the permanent magnet synchronous linear motor mover is established: （1） In the formula, M is the mass of the mover. This is the electromagnetic thrust coefficient. For the q-axis current, the electromagnetic thrust in a surface-mounted permanent magnet synchronous linear motor is satisfied. The ".." symbol above the variable indicates a second derivative; Step S2: Define the state that needs to be expanded according to equation (1). : （2） In the formula, Initial values ​​set for the mover mass identification algorithm; Step S3: Expand the state term As the third extended state in addition to the mover displacement p and the mover linear velocity v, a linear extended state observer of the system is constructed: （3） In the formula, , , These are the observed values ​​of the mover displacement, the mover linear velocity, and the expansion state, respectively. , , for , , The corresponding observation gain, the "." symbol above the variable indicates the first derivative; Give a current to the q-axis; Step S4: Observations of the extended state The correction yields a new state term: （4） In the formula, The result is the output of the mover mass identification algorithm; combined with equation (2) and equation (4), it is further approximated as follows: （5） Step S5: Based on formula (2), using the mass M and linear velocity v of the mover as state variables, construct the discrete prior estimation equation: （6） In the formula, The algorithm is discrete; the variable at time k is labeled with "k"; the prior state variable matrix is ​​defined. = ; Step S6: Calculate the covariance matrix: （7） In the formula, To estimate the covariance matrix a priori; For the posterior correction of the covariance matrix; Q Let D be the model noise covariance matrix. The partial derivative matrix of the state equation; Step S7: Identify gain calculation: （8） In the formula, G To identify the gain matrix; H= R is the partial derivative matrix of the state variables; R is the measurement noise covariance matrix. Step S8: Using the identification gain matrix G Correction The sum of (7) : （9） Step S9: Read the posterior-corrected state variable matrix The decoupled mover mass identification results are obtained: （10）。 2. The method for decoupling and identifying the mass of the permanent magnet synchronous linear motor mover according to claim 1, characterized in that: The current loop employs a PI control algorithm, while the position loop is designed based on the characteristics of the mover mass identification algorithm. The feedback signals in this loop are the observed state variables of LESO at various orders. The input signal of the mover mass identification algorithm is the q-axis current. q-axis given current , mover displacement p, mover given displacement And the linear velocity v of the mover; where the q-axis current is obtained by converting the three-phase current collected by the current sensor, and the mover displacement is obtained by measuring the high-precision grating ruler.

3. The method for decoupling and identifying the mass of the permanent magnet synchronous linear motor mover according to claim 2, characterized in that: The state observation algorithm consists of two parts: LESO and state correction. First, the observed values ​​of the mover displacement p and the mover displacement are... The difference is used for gain and integration to obtain the observed value of the expanded state. Based on this, the observed values ​​of the moving part's linear velocity were further obtained. Observed values ​​of mover displacement In this process, each order of observation states is used to form a feedback closed loop for position control; and The middle part represents the mass M of the moving part and the frictional force. Electromagnetic thrust Initial values ​​for quality identification The coupling information is used, therefore electromagnetic thrust, initial identification value, and identification results from the SE-EKF algorithm are introduced. right Make corrections to obtain the corrected state items. After state correction, the mass of the moving part is only related to the frictional force. There is coupling; in order to The motioner mass contained therein is completely decoupled and identified using the decoupled SE-EKF algorithm; The SE-EKF algorithm first uses the mover mass M and the mover linear velocity v as state variables, and then utilizes electromagnetic thrust at time k. , And the posterior correction matrix of the previous time step The prior estimates of M and v are calculated, and together they constitute the prior state variable matrix. On the other hand, using , , Calculate the partial derivative matrix D of the state equation. Combined with the posterior corrected covariance matrix of the previous time step The model noise covariance matrix Q is used to calculate the prior estimated covariance matrix. Then, by combining the partial derivative matrix H of the state variables and the covariance matrix R of the measurement noise, the identification gain matrix G for identification update can be calculated. Then, the information obtained by subtracting the prior estimate of linear velocity from the actual value of linear velocity is compared with G. Multiply to calculate the correction amount for identification; based on the correction amount, apply the correction to the prior state variable matrix. After making corrections, we obtain the posterior correction matrix at the current time step. And the covariance matrix After correction, the posterior corrected covariance matrix at the current time is obtained. , obtained and The algorithm is used in the identification iterations of the next time step to eventually converge to a steady state; the posterior correction matrix generated in each iteration of the algorithm is used in these iterations. Each value contains identification values ​​for the mover mass and linear velocity. The identification value for the mover mass is output as the identification result. To ensure the state correction in the state observation algorithm The accuracy of the identification results This feedback is then fed back into the state observation.

4. The method for decoupling and identifying the mass of the permanent magnet synchronous linear motor mover according to claim 3, characterized in that: First, sample the DC bus voltage. ,wait After the drive requirements are met, the relay is closed to start the motor; next, initialization steps including drive configuration, capacitor charging, and DC offset compensation for AD sampling are performed; then, the mover of the linear motor is positioned, and the precise position of the mover is obtained through the QEP module; finally, the operation control of the linear motor is performed. During the algorithm's execution, firstly, the driver determines the position of the mover (p) and the given d-axis and q-axis voltages. , Calculate the α and β axis voltages required for control. , The switching signal of the inverter power transistor is obtained according to the duty cycle modulation algorithm, and the power switching transistor is controlled to operate through the isolation drive circuit, thereby controlling the operation of the linear motor. Secondly, the actual position p of the mover and the given position p are used. * q-axis current given The observed values ​​of the mover displacement are obtained through the observer stage. Observed values ​​of the linear velocity of the moving part and observations of the expansion state Secondly, regarding the observations Make corrections to obtain new state terms. Finally, Substituting into the SE-EKF algorithm, the mover mass of the permanent magnet synchronous linear motor is identified through four steps: prior estimation, covariance matrix calculation, updating the identification gain, and posterior correction of the state variables.

Images