Method for calculating an induction flux vector angle, and corresponding device

Abstract

The invention relates to a method for calculating an induction flux vector angle, comprising a plurality of iterations, each iteration of the plurality of iterations except for the first one comprising: a / obtaining (S100) input data comprising armature voltage values, armature current values, armature resistance values and inductance values associated with the armature; b / estimating (S101) the induction flux vector of the iteration by adding a function of said input data to an induction flux vector obtained in the previous iteration, c / a correction comprising i. determining (S102) the centre of a circle passing through at least three points defined by the flux vector obtained through said estimation, and at least two flux vectors obtained in two previous iterations, and, ii. obtaining the flux vector of the current iteration, in which the flux vector obtained through said estimation is corrected (S106, S107) using said centre vector.

Description

Description Title of the invention: Method for calculating an angle of the inductor flux vector, and corresponding device Technical Field

[0001] This presentation concerns polyphase synchronous or asynchronous motors, and more particularly the observation of the flux vector of these motors. It concerns even more particularly the actuators which use these motors. Prior art

[0002] Polyphase synchronous or asynchronous motors, receiving rotating alternating currents (sometimes referred to as AC machines), can be integrated into electric actuators. An electric actuator comprises, for example, a motor of the aforementioned type with a mechanical reducer (for example, a gear), an electronic power inverter (or variator) ensuring the variable supply of electrical energy, and a control module configured to control switches of the inverter so that rotation is obtained.

[0003] These motors are used in many applications, in a power range from tens of watts to hundreds of kilowatts.

[0004] In particular, these electric actuators are used to embed them in systems such as motor vehicles, in which the electrification of many functions (replacing hydraulic systems) requires the use of these electric actuators with variable speed control (speed variations, load torque variations). In electric motor vehicles, the vehicle's traction or propulsion actuator is also an actuator that must be configured to handle speed variations and load torque variations.

[0005] Electric actuators are also used for robotics applications, for example in the context of household appliances (motors of washing machines controlled in position and speed, opening motors (gates, shutters)).

[0006] Another possible application of electric actuators is aeronautics and aerospace, typically for the motorization of aircraft drones.

[0007] Variable speed means situations in which the speed control can vary by at least ten percent of a maximum speed per electrical revolution, or even several tens of percent, with braking, stops, or even speed reversals. Variable speed also means situations in which the mechanical load applied to the motor varies suddenly, i.e. by several tens of percent of the nominal torque in one electrical revolution, or even situations in which the mechanical load can change sign by passing from a load torque to a driving load, and vice versa. Thus, the electric motors referred to here can be used as a motor but also as a generator (alternator or asynchronous generator depending on the type of machine).AC machines can also intermittently alternate between motor operation and generator operation (alternator or asynchronous generator depending on the type of machine). Often, due to the variations in electrical power generated, the DC supply voltage of the inverter can be disturbed and also become variable. These cases are in contrast to the succession of permanent regimes, where the motor speed, the load, and the inverter supply voltage are constant or slowly variable relative to the rotation of the motor.

[0008] For information, by "armature", which is also cited as "exciter armature", or "armature" in English, we designate the set of polyphase windings which will carry the alternating current of the AC machine with the associated magnetic circuit. Generally, the armature is the stator on synchronous machines and asynchronous machines. In a context where the armature is the stator on a synchronous or asynchronous machine, we have: the "armature" AC currents and voltages which are the currents and voltages "statoric", and the calculation reference linked to the armature which is the so-called "statoric" reference, or "fixed", "linked to the stator", or "stationary", in the plane perpendicular to the axis of rotation of the motor and centered on it, most generally whose abscissa axis (noted "α" or "a") coincides with the winding axis of phase 1 of the armature.

[0009] Generally, AC machines consist of a three-phase armature, but not exclusively: many machines are "two-phase" or "two-phase", with two windings wound at 90° electrical angles, and others are sometimes six-phase, or more widely multi-phase.

[0010] By "inductor", which is also cited as "exciter inductor", or in English "inductor", "magnetic field" or "exciting / excitation field", more specifically "field winding" for asynchronous or synchronous machines with wound rotor or "magnet field" for synchronous machines with permanent magnets, we designate the electromagnet generating the magnetic field in the armature, typically controlled to be at constant flux amplitude or slowly variable over time.

[0011] In a permanent magnet synchronous machine, the inductor consists of a set of magnets and a magnetic circuit.

[0012] In a wound rotor synchronous machine, the inductor consists of a set of "electromagnet" windings, this set being powered by an adjustable excitation current controlled as predominantly direct ("DC"), and a magnetic circuit.

[0013] In an asynchronous machine, the inductor consists of a set of short-circuited windings, such as "squirrel cage" conductors, for example, and a magnetic circuit.

[0014] In a widespread manner, the inductor is the rotor of AC machines. In this context, the calculation reference frame linked to the inductor is the so-called "rotor", "rotor-related" or "rotor flux" reference frame, whose abscissa axis generally noted "d" is aligned with the inductor flux, and the axis noted "q" at 90° from "d" is collinear with the back-electromotive force vector. It may be noted that there are other structures for which the AC machine is constructed in an inverted manner, as described for example in document EP1235332A1 or in document US2017077773A1, in which a polyphase armature is a rotating rotor, while the inductor is in the stator. It is thus understood why it is preferable in the present description to use the terms "armature" and "inductor".

[0015] In all the above-mentioned actuator applications involving AC machines controlled in variable speed, the performances commonly sought for the use of AC machines are their durability (absence of brushes or collector), their compactness resulting in a high volume or mass density of power, their high efficiency (at least ≥70%), their low inertia in certain topologies minimizing the exchange of electrical energy for variable and very dynamic operating regimes, their low electromagnetic radiation; and finally the possibility of modifying their apparent torque constant with respect to the amplitude of the phase currents, so that the geared motor can behave as a continuously electrically variable reduction,working at constant power over a very wide speed range (beyond the so-called "base" speed for a given supply voltage) without requiring a bulky or expensive mechanical reduction ratio changing device (gearbox). For this, in the case of the wound rotor synchronous machine the field flux is defined by controlling a field current independent of the polyphase armature. For other AC machines, the apparent torque constant is set by controlling one of the two degrees of freedom of rotating armature current, referred to as "direct" / "d" axis current in the literature (Park transform / dqo): on the asynchronous machine it defines the field flux, and on the permanent magnet synchronous machine it opposes the field flux to generate a lower standard air gap flux ("defluxing",(by misuse of language). This characteristic of AC machines has made it possible to eliminate the multi-speed gearbox on the majority of electric vehicles, and also to make the actuators more compact for a given specification.

[0016] In return for these performances, the constraint commonly shared by these applications is that the control part of the actuator always requires information on the angular position of the inductor relative to the armature, to ensure control of the AC motor in variable speed or with high efficiency. More precisely, for the control part the angle sought is that of the magnetic flux of the inductor, or also of its derivative the back electromotive force (BEMF), relative to phase 1 of the polyphase armature. On a conventional synchronous machine, the electrical angle of the rotor coincides with the angle of the inductor flux: the inductor flux is positive and maximum in phase 1 when this angle is zero by convention. The angle of the back electromotive force will always be 90° ahead of the inductor flux in the direction of speed, and therefore also provides the information sought.We will refer to one or the other of these magnetic angles as the “rotor angle” for the sake of simplicity.

[0017] To obtain this information, known implementations include either a physical measurement sensor of the angular position of the rotor, or a so-called "sensorless" estimator or observer, based on measurements of armature phase voltages and / or currents associated with a calculation process resolved by the control part in real time, to reconstruct the angle information (or both, to ensure redundancy for operational safety or personal safety in certain cases).

[0018] The known advantages of sensorless rotor angle estimators / observers are: compactness and weight, as the presence of an angle sensor takes up space in the motor environment and requires additional electrical connections, and possibly a cost reduction advantage of the actuator, as a hardware device is replaced by software.

[0019] In an automotive context, the use of a sensorless rotor angle observer for the control of a three-phase synchronous motor with magnets permanent constitutes a significant technical advantage in enabling the compactness of an electric actuator for controlling the locking or connection of shafts in the transmission of a vehicle, contributing to its low on-board weight (contributing to the autonomy of the electric vehicle) and thus its high mass and volume density of power, while limiting the cost.

[0020] Even more so in some cases, the actuator may have parts that are distant from each other on the vehicle: namely, the three-phase AC electric motor and the reducer (gearmotor) are located in the vicinity of their mechanical load (typically, the vehicle transmission), while the inverter and the control part are implemented on a remote ECU elsewhere in the vehicle. In this context, the sensorless observer makes it possible to limit the number of wirings, sometimes several meters long, between the ECU and the three-phase machine: reduced to only 3 power connections to supply the three-phase machine, instead of 6 to 8 if an angular position sensor was included.

[0021] Similarly, the use of a sensorless rotor angle observer provides a weight advantage for drones, where the number of geared motors varies from 4 to more than 6; thus, as many sensors, and numerous electrical connections to the motherboard are avoided, reducing the cost of the drone, reducing its weight and improving its autonomy. That being said, this autonomy and the high volume or mass power density of the geared motor thus designed, require very good angular accuracy of the measuring means. However, the estimators and observers without rotor angle sensors of the prior art lack precision compared to physical sensors. Also, the initialization time of these estimators and observers remains problematic.

[0022] Real sensors are commonly accurate to within an uncertainty of plus or minus one electrical degree, while state-of-the-art sensorless observers, depending on their technological sensitivity and the variation in machine behavior (load, speed, temperature, aging), have a guaranteed accuracy of the order of plus or minus fifteen electrical degrees. minimum, up to plus or minus 30°, in variable mode. A precision of plus or minus 15° electrical is, for example, quite sufficient to guarantee the correct operation of the AC machine control in dynamic mode. On the other hand, this lack of precision can result in a drop in efficiency, therefore an increased electrical consumption of the machine for the same torque (for 15° error, drop in efficiency of 3.5% minimum), and a loss of autonomy if the system embedding the actuator is battery powered. Such overconsumption is problematic, for example, on an electric car, or certain drones.In the case of the use by the control part of the "defluxing" technique previously explained, on a permanent magnet synchronous machine, the loss of efficiency due to the angular inaccuracy of a sensorless observer can be significantly greater than without defluxing: for example, for a current advance of 45° on the counter electromotive force (CEMF), an angular error of 15° can induce a drop in efficiency of more than 20%.

[0023] With regard to electric vehicles, the observations presented above apply to the main traction (or propulsion) actuator of the vehicle, but to all of the smaller actuators of the vehicle, the sum of whose consumption has an impact on the battery.

[0024] It should be noted that for the main electric traction (or propulsion) actuator of an electric vehicle, an additional issue is the influence of the inaccuracy of the angular position measurement of the motor on the torque ripples. Typically, a torque ripple of plus or minus 3.5% of the demand due to an angular observation uncertainty of plus or minus 15° electrical, can induce vibrations that can be felt by the passengers of the electric vehicle, which is not desirable.

[0025] For synchronous machines, real sensors also have a functional advantage over sensorless observers, namely that they can provide the motor position as soon as the control part is powered up. Their initialization time is zero, and they do not require any prior active power supply to the AC machine. Thus, starting the motor is possible with maximum efficiency and without unwanted torque ripple.

[0026] Conversely, so-called "passive" sensorless observers require a displacement of the rotor to determine the angular position. It is known that when switched on and without angular displacement (without speed or acceleration), the position is unobservable (Mohamad Koteich, "Modeling and observability of electrical machines for sensorless mechanical control" Automatics / Robotics, Paris Saclay University (COmUE), 2016. French. ffNNT: 2016SACLC043ff. fftel-01320377v2).

[0027] In order to initialize the observation of the angular position, some so-called "active" techniques based on the injection of power harmonics to the armature (Amir Messali, Malek Ghanes, Mohamad Koteich, Mohamed Assaad Hamida. A Robust Observer of Rotor Position and Speed ​​for IPMSM HFI Sensorless Drives. 2018 IEEE 9th International Symposium on Sensorless Control for Electrical Drives (SLED), Sep 2018, Helsinki, Finland. pp.90-95, ff10.1109 / SLED.2018.8486140ff. ffhal-02378496), or to the inductor in the case of a wound rotor synchronous machine, make it possible to determine the initial position of the rotor without it rotating (as described in document WO 2016079374), with however an uncertainty that can go up to the angular sector of plus or minus 30° electric.As far as the main electric traction of a vehicle is concerned, knowledge of the initial position of the rotor, either by a sensor most often, or by one of these active techniques, remains mandatory. For other application cases where these techniques are not used, the control part must apply an armature current, called "open loop" with regard to the angular position, in order to induce a rotation in one direction or another, and allow the "passive" sensorless observer to determine the position, to resume control in normal angular servo control in the desired direction of rotation. The time during which the position is unobservable and unknown, until it is determined with the desired nominal accuracy, is called the initialization time. During this time, the rotor angle is provided with a. uncertainty of up to 100% error over 360° electrical, and the issues mentioned above regarding precision arise in this context.

[0028] In order not to penalize the final application on which they are intended to replace the sensor, it is desirable to obtain passive sensorless estimators and observers usable in actuators with better angular accuracy, and the shortest possible initialization time. This last technical problem is particularly critical for position control actuators, where an actuation must be carried out in a specified maximum travel time.

[0029] We will now describe prior art sensorless estimators and observers that do not sufficiently address the technical problem of angular accuracy.

[0030] The state of the prior art makes it possible to distinguish between direct measurement methods of one of the phase back electromotive force (BEMF), estimators and state observers, the latter two categories calculating as output vector either the BEMF vector or the flux vector, the polar angle calculation of which provides the desired rotor angle. This typically constitutes 5 families of sensorless rotor angle determination methods.

[0031] The first family, of direct intermittent measurement of the voltage of a phase FCEM is known in particular from documents US 7301298 and CN 114531063. It is also implemented by the component marketed under the trade name A4964 by the company ALLEGRO MICROSYSTEMS, used for synchronous machines.

[0032] This direct measurement technology requires "under-powering" the polyphase electrical machine, by forcing the inverter to alternately leave one of its bridge arms in high impedance, therefore one of the phases in open circuit without current circulation by 60° electrical sector, so that the phase voltage represents the phase back-fault current, and can be directly measured and acquired by the analog-digital converter of the part control. The current then has a high harmonic content. While this power supply technology is suitable for many applications, it is not applicable to on-board applications requiring high efficiency, high power density, high power such as electric traction actuators, or prioritizing battery autonomy. It is also not applicable with the so-called "defluxing" control technique, for angles where the current is ahead by more than 20°, as the phase current cannot be at its maximum at the same time as the phase would be open circuit to allow the phase back-electromagnetic field strength to be read.

[0033] Furthermore, once these limitations on the power supply have been implemented, the direct measurement of the rotation back-fed emf of the inductor is by nature intermittent, since being projected onto one of the 3 armature phases put in open circuit, it only gives access to a one-dimensional projection and not to the vector in the machine frame. Consequently, the rotor angle can only be estimated under conditions by angular sectors of 60°, that is to say basically at plus or minus 30° electrical of one of the 6 angles corresponding to the so-called zero crossings or in English "zero crossings" of the back-fed emf voltage of a measurable phase with the floating neutral potential of the machine.Certain techniques (filtering, or signal slope analysis or signal at several points, as described in document US 2020343840) make it possible to reduce this estimation uncertainty to typically 10° electrical when the rotation speed approaches at least a third of the base speed for the supply voltage considered. The two observability conditions are sufficient speed, and moderate speed variation (i.e. less than 30° electrical angular variation per sector traveled by 60° electrical), in other words the assumption of pseudo-permanent regimes. If the speed is too low (less than approximately 7%) compared to the base speed, the zero crossing instant will be difficult to identify, and the unfavorable signal-to-noise ratio will not allow the position to be observed. The initialization times can therefore be significant compared to those of the Observers.Finally, if the regime is very dynamic, with strong variations in load torque or controlled variations in speed with strong acceleration or strong deceleration, the. phase which will be commanded to be in high impedance and allow the measurement will appear too far ahead or behind (beyond plus or minus 30°) compared to the zero crossing of the real FCEM signal of the machine, and the intermittency of the measurement will induce a disconnection of the control part compared to the angular reality of the motor.

[0034] It appears that this technology based on the measurement of FCEM is not satisfactory.

[0035] The solutions of the estimators and observers allow the machine to be powered by the inverter which transmits 100% of the available electrical power to the machine, making use of all its bridge arms in low impedance, either in order to generate sinusoidal armature currents (optimal for the generation of average torque), or by so-called "overmodulation" techniques, where for maximum power at the expense of efficiency, a maximum voltage amplitude is supplied to all the armature windings. The estimators and observers use as input data the knowledge (or sometimes the measurements) of the phase voltages produced by the inverter, and the measurements of the armature phase currents of the machine. If the machine is synchronous with wound rotor, the measurement of the field current will also be a beneficial input data.The power supply technology, as well as the sensorless estimation or observation, is also compatible with “defluxing” control. These solutions are therefore suitable for the general context of high power density actuators.

[0036] Generally speaking, estimators and observers are based on the modeling of the system, here the AC electric motor, in state representation, that is to say by using state variables which are defined as computer memories of internal quantities describing the system, most often representative of physical quantities image of energy storage evolving in the system, which cannot physically vary suddenly over a very short time interval.

[0037] The state vector, usually denoted "X", represents the set of state variables used for modeling. When the state of a system is not not measurable, we design an observer that allows us to reconstruct the state from a model of the dynamic system and measurements of other quantities. In the case of electrical machines, depending on the modeling chosen, these quantities of the vector "X" will most often be linked to internal storage of inductive energy (flux, currents), kinetic energy (speed, back-to-back force), or potential energy (position).

[0038] The state vector can also contain data that is low-pass filtered or results from an action or integral filtering, since this data is also representative of evolving information that cannot suddenly vary over a very short time interval. This is called an "extended" state vector (to signal processing information not internal to the physical system). Depending on the modeling and implementation chosen, sometimes this filtering data is integrated into the state vector "X" directly, or in other cases it will be located in separate memories (like the evolving matrices of a Kalman filter) while the vector "X" contains only the physical state variables of the system to be observed.

[0039] More precisely, digital observers are generally iterative algorithms for calculating non-measurable data, of a system modeled in state representation (the electric motor for example), which are based firstly on a step of estimating the state vector noted ^ ^̂ at the current time 'n', from the state vector denoted ^ ^ ~ ^ ^ observed at the previous instant, of an assumed calculation model and of a certain number of measurements (electrical in our case) of entry into a vector noted ^ ^ We will now describe an observer operation.

[0040] An estimate can be implemented from vector functions that represent the digitized computational model of the known system at the current time, for example denoted ^ ^ , ℎ ^ , ^ ^ , ^ ^ .

[0041] We can implement a calculation of the estimated state vector:

[0042] ^^̂ = ^^^^~^^^ ^ + ℎ^^^^ ^

[0043] We can then implement a calculation of estimated outputs from the model:

[0045] With ^ ^̂ the output vector. It should be noted that the estimated outputs of the model correspond to data that can be measured elsewhere, either directly or indirectly, while the variables in the state vector are in principle internal and not measurable. If a state variable in the vector X is measurable, then it will generally also appear as an element of the output vector^ ^̂ .

[0046] Then, the observers (unlike the estimators) present a second step called correction of the state vector, starting first from the calculation of a deviation or "error" vector noted ^ ^ between outputs estimated at the first stage and additional measurements (or sometimes a function of measurements) ^ ^^^^^ .

[0047] A correction step can be implemented by calculating the error vector between the measured outputs ^ ^^^^^ and the estimates ^ ^̂ :

[0048] ^^ = ^^^^^^ −^^^̂

[0049] This error vector ^ ^ then serves as the input variable of a choice of correction functions ^ ^ which will differentiate between the different observers: constant gain matrix (Luenberger observer), sliding mode observer with error function in absolute value or sigmoid, variable Kalman gain matrix, etc. The output of this correction function is added to the estimated state vector ^ ^̂ , to obtain the observed state vector ^ ^ ~ at the current calculation step. Finally, the observed output vector ^ ^ ~ , containing the data that we wish to observe, is a matrix combination of all or part of the observed state vector ^ ^ ~and possibly input measurements. We therefore have a calculation of the observed / corrected state vector:

[0050] ^~ ^ = ^^̂ + ^^^^^ ^

[0051] A calculation of observed outputs:

[0053] Thus, the estimated state variables ^ ^̂ are obtained by “open loop” calculation from a model assumed to be known of the system, from the previous state variables ^ ^ ~ ^ ^ and new input data ^ ^ without having been corrected yet, while the state variables ^ ^ ~ ^observed at the current time are corrected, which in a sense constitutes feedback.

[0054] The estimators, simpler and therefore ultimately less precise, actually correspond to observers without a correction step: only the first step is carried out. In other words, all available measurements ^ ^^serve as input data to estimate the state vector X^ with the assumed model, without any additional measurement (available) ^ ^^^ to make a correction, a “feedback”. Equivalently, an estimator is an observer whose ^ = [0], or e ~^ ncore whose ^^ = ^^̂.

[0055] Estimators are of little use in industrial applications, in favor of more precise observers capable of responding to the technical problem.

[0056] Observers without rotor angle sensors according to the prior art are distinguished mainly by the choice of physical (and filtered) quantities modeled in the state vector "X", incidentally by the calculation reference in the plane in which these quantities are expressed (three-phase 1,2,3 (or polyphase), (α,β) or (d,q), etc.), and finally by the correction function ^ ^used: linear matrix corrector (called Luenberger), stochastic corrector called 'Kalman filter', non-linear sliding mode observer, etc.

[0057] Many mathematical formulations are used to estimate and store intermediate physical quantities in the state vector X: vector of armature currents, or / and total flux vector at the armature, inductor flux particularly for asynchronous machines, etc.

[0058] Also, it is mainly the final step which consists of obtaining the angle of the inductor relative to the armature from the angular coordinate of a vector of the plane element of the vector ^ ^ ~ observed output data, which will divide the sensorless observer into two broad categories: those that produce as output data the vector ^^ of FCEM ('BEMF' in English), and those that produce as output data the vector Inductor Flux often noted in the literature ' ^ ^ ^^^^ ^ ^ ', noted ^^^ in this description. This angular coordinate is then either filtered (low-pass, phase-locked loop known as "PLL: Phase-Locked loop" in English, etc.), or used as is as output data from the process.

[0059] As an indication, the desired precision for actuators can be a maximum of plus or minus 8° electrical in dynamic mode, which can ensure a loss of efficiency of less than < 1% without “defluxing”.

[0060] The following previous papers present sensorless observers: Liu: Liu, Siwei & Zhijian, Qiu & Chen, Wei. (2019). Sensorless Control with Sliding Mode Observer for a Brushless DC Motor based on Concave Function. 872-876. 10.1109 / IMCEC46724.2019.8984058; Urbanski: Urbanski, K.; Janiszewski, D. Sensorless Control of the Permanent Magnet Synchronous Motor. Sensors 2019, 19, 3546. https: / / doi.org / 10.3390 / s19163546; Shen: Shen, Jian-qing, Lei Yuan, Ming-Liang Chen and Zhen Xie. “Flux Sliding-mode Observer Design for Sensorless Control of Dual Three-phase Interior Permanent Magnet Synchronous Motor.” Journal of Electrical Engineering & Technology 9 (2014): 1614-1622; Renesas; "Motor Control Application, from RENESAS company, published on October 31, 2018, version 1.02.

[0061] From these documents, we can see that passive sensorless observers (an observer is called "passive" because it is based on voltage and current measurements obtained without requiring modulation of the electrical power at the armature or inductor to provide the information), produce an observation of the electrical angle of the rotor with an angular error typically greater than plus or minus 25° in steady state, and up to plus or minus 40° in dynamic state (strong variation in speed or load).

[0062] As regards permanent magnet synchronous machines, the prior art is mainly based on the knowledge of two parameters (Rs, Ls) or three parameters (Rs, Ld, Lq) seen from the armature: Rs being the resistance of an armature phase (all phases have the same resistance Rs); Ls being the cyclic phase inductance, or average cyclic inductance over one revolution if the pole saliency is neglected (=(Ld+Lq) / 2).

[0063] Also, we can use the three-parameter model taking into account the extreme cyclic phase inductances during an electrical revolution, with: Ld in the direction of the inductor flux; and Lq at 90° electrical angle to the inductor flux.

[0064] For synchronous machines, a common point to all estimators and observers of the prior art, of FCEM or flux, is that the equation of the machine is based only on what happens at the armature (at the stator in general), to try to determine the behavior of the inductor (rotor in general). It can be noted that in the Urbanski document mentioned above, and also in the Delpoux document (Delpoux, Romain & Floquet, Thierry. (2014). “High order sliding mode control for sensorless trajectory tracking of a PMSM”. International Journal of Control. 10.1080 / 00207179.2014.903563.), the state vector is even extended to the speed in order to include the fundamental principle of mechanical dynamics, with the calculation of the electromagnetic torque and the knowledge of the rotor inertia noted J as an additional parameter, in order to constrain the acceleration and obtain an estimate of the load modeled as a disturbance term in pseudo-steady state. That being said, the observer is not thereby made more angularly precise.

[0065] There is therefore a real need for a sensorless observer which is free, at least in part, from the aforementioned drawbacks. Statement of the invention

[0066] The present disclosure relates to a method for calculating an angle of the inductor flux vector in an observation frame of a polyphase synchronous or asynchronous motor provided with an armature and said inductor, comprising a plurality of iterations, each iteration of the plurality of iterations except the first comprising: a / obtaining input data comprising armature voltage values, armature current values, and at least one armature resistance value (one or more); b / estimating in the observation frame, the inductor flux vector of the iteration by adding to an inductor flux vector obtained in the previous iteration a function of said input data, the inductor flux vector pointing towards an estimated inductor flux point and starting from the origin of the observation frame, c / a correction comprising: i.a determination of the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the current iteration to deduce a center vector, and, ii.obtaining the flux vector of the current iteration in which the flux vector obtained by said estimation is corrected using said center vector, so that the corrected inductor flux vector points towards said estimated inductor flux point and starts from a point which approaches (approaches or even for example coincides) the center of said circle, this point being considered as the origin of the observation reference frame for the following iteration, the angle of the calculated inductor flux vector being the angular coordinate of the flux vector obtained by the correction (here, the flux vector obtained by the correction is expressed in the modified reference frame, that is to say the reference frame whose origin approaches the center of the circle, the modification of the reference frame only modifying the origin).

[0067] For example, step a / may include an optional obtaining of one or more inductance values ​​associated with the armature (by inductance values ​​associated with the armature, we mean values ​​seen from the armature, measurable from the armature, or even leakage and magnetizing or cyclical of the armature phases). The process can nevertheless operate without taking into account inductance values ​​associated with the armature.

[0068] By being expressed in a frame of reference which approaches the center of said circle, we will have, after a plurality of successive iterations, a center of the frame of reference which will converge towards the center of rotation of the motor (and therefore of the real flux vector), and the angle of the real inductor flux vector is thus estimated in an increasingly precise manner.

[0069] This method can be implemented by a computer, for example a computer embedded in a system such as a vehicle. More generally, this method can be implemented by a device (or several devices) having a computer system structure (or each having a computer system structure, with one or more processors and one or more non-volatile memories.

[0070] It has been observed that the trajectory of the inductor flux of a permanent magnet synchronous motor, or of the flux of a wound rotor synchronous machine whose inductor current is controlled constant, or of the inductor (rotoric) flux of an asynchronous machine controlled as constant by the control part, is substantially circular on average in a rotating electrical machine (as would be a point linked to the rotor located at a constant radius from the center, seen from a reference frame linked to the stator) and centered on the center of rotation of the motor (which is also the center of the observation frame (αβ), (dq), etc.).

[0071] It is therefore proposed to take into account the geometry of an expected trajectory of the inductor flow, this expected trajectory being circular.

[0072] It can be noted that the method does not propose, in a fixed center frame, to conform the norm of the estimated flux vector to a constant value, (so that the flux observed at the end of the operation is always on a circle of the same radius), the radius being for example predefined, or even low-pass filtered; and then to apply an equal norm correction, on the components (x,y) (or (d,q)) of the flux. Such an approach would be incorrect for two reasons: first, the radius of the circle described by the flux, is a parameter of the machine which is the torque constant of the machine, which would then have to be added as input data to the model, with the disadvantage that it can also vary with external tolerances and environmental influences (temperature, aging), or internal ones depending on the control of the inductor flux applied for asynchronous and synchronous machines with wound rotor; on the other hand, it is not by modifying the length of the same vector (estimated flux) in a fixed center frame that its angular coordinate is corrected.

[0073] In the method defined above, we conform to this circular trajectory, by bringing the origin of the reference frame closer to the center of this trajectory. In particular, the correction of the flux vector obtained uses this center vector, which makes it possible to favor a substantially circular trajectory for the flux vector, and makes the calculation of the angle more precise.

[0074] In fact, many factors tend to distort an estimated trajectory from the armature: tolerances on the evolving parameters of the armature model, distorting harmonics and measurement noise on the currents, on the application of voltages with PWM generation and the uncertainty generated by dead times, etc. Therefore, the method proposes to use the circular nature of the flux trajectory at the inductor, to correct the 'distorted' estimate of the internal state variable that is the inductor flux, seen from the armature.

[0075] The use of a circle in the process (or at least an arc of a circle passing through said points) is implemented regardless of the observation frame (called "Park") used to describe the armature of the machine: stationary (α,β) (simpler formulation), linked to the inductor flux (d,q), or rotating in any way relative to the rotor and the stator.

[0076] The method also provides a correction by an independent geometric modeling and not taken into account when estimating the inductor flux vector (and not taken into account in the state of the prior art). The geometric modeling (the determination of the center of the circle) is implemented directly for the flux vector, which is the state variable and also the output data from which the angle of the inductor (of the rotor most often) will be directly extracted. By further constraining the mathematical problem, This additional geometric modeling, independent of the machine parameters, makes it possible to reduce their influence on the uncertainty of the final result, and improves the angular precision in dynamic mode by almost an order of magnitude.

[0077] According to a particular implementation mode, the flow vector obtained by said estimation is corrected by means of a correction function added to the flow vector obtained by said estimation, the correction function being configured to receive as input an error vector comprising said center vector, the error vector possibly being assigned a coefficient (positive, negative, real).

[0078] For information, the correction function can be configured to correct the flux vector so that it approaches a vector from this center or even so that it is a vector from this center.

[0079] The person skilled in the art will know how to choose an appropriate correction function, in particular a function configured to receive as input an error vector comprising said center vector.

[0080] According to a particular implementation mode, the correction function is chosen from: filtering of the error vector, multiplication by a constant or time-varying gain matrix of the error vector, multiplication by a variable gain matrix based on a variation calculation or based on a Kalman filter, a non-linear function, for example a gain function times sign function, a sigmoid function, a sign function times square root of absolute value, a sign function times integer or non-integer power function of the absolute value, a hyperbolic tangent function, a sine function receiving an input bounded by the interval [− a combination of one or more of said nonlinear functions.

[0081] It has been observed that these functions can implement the correction efficiently and simply.

[0082] According to a particular embodiment, the method further comprises a correction of said at least two flow vectors obtained at two iterations which precede the current iteration.

[0083] In this particular implementation mode, the flow vectors obtained at two iterations preceding the current iteration are also corrected. These flow vectors may have been memorized. This correction may be a correction analogous to that implemented in step c / .

[0084] As an indication, a possible correction can use a ratio between the norm of the currently commanded inductor flux vector and the norm of the inductor flux commanded in the previous iteration, so as to maintain a theoretically circular desired trajectory even if the intensity of the machine's inductor flux varies.

[0085] Here, we will approximate a circle trajectory for several points in the past. In fact, the center of the circle may coincide with the center of the observation frame with an appropriate frame. Due to the deformation of the estimate (tolerances of the armature model and electrical measurements), this center will not coincide exactly, and the error vector may therefore be the center vector of this circle itself, relative to the center of the observation frame. The correction has the effect, iteration after iteration, of "recentering" the calculation frame on the centers of the successive identified circular arcs. The angle of the flux observed in this corrected two-dimensional frame is therefore closer to reality than the angle of the vector estimated before the frame has been readjusted.The effectiveness of the correction allows the use of a fairly simple correction function, such as, for example, a constant gain (Luenberger), or a gain and a low-pass filter of the error vector (i.e. indirect extension of the state vector), to smooth the change of reference over time.

[0086] Also, in this mode of implementation, the angle of the observed flux is that of the vector of the last inductor flux estimated with respect to the center of the determined circle (towards which the algorithm will converge the center of the reference frame); and this angle will therefore be very close to the actual rotor angle, with reduced error and delay that do not require any specific filtering such as the use of a Phase Locked Loop. Since the angular error is lower with the circular geometric model (i.e. the circle) of the inductor than through the distorting prism of the armature, it is not necessary to use a high-performance high-gain slip mode corrector, which can generate chattering requiring filtering. We therefore have a numerically stable angle, with low ripple and also low delay.

[0087] According to a particular embodiment, the motor is a synchronous motor whose inductor is a wound rotor or an asynchronous machine, in which the correction of said at least two flux vectors obtained at two iterations preceding the current iteration is implemented, by determining a ratio between the norm of an inductor flux vector controlled at the current iteration and the norm of the inductor flux controlled or obtained during a previous iteration, or by determining a ratio between the inductor current controlled or obtained at the current iteration and the inductor current obtained during a previous iteration.

[0088] It should be noted that for wound rotor synchronous machines and asynchronous machines, the observer presents no contraindication to the controlled inductor flux being changed, as long as the variation is slow in time, i.e. takes place over several electrical revolutions of the machine, which is generally the case in practice, the inductor flux being generally reduced at very high speed. In this case, it is not necessary to compensate for the slightly non-circular, but spiral trajectory of the flux. If the inductor flux were to be varied more rapidly, compensation would be made on the memories of past fluxes, so that these past fluxes always constitute a circular (and not a pronounced spiral) trajectory with the flux at the current instant.

[0089] According to a particular mode of implementation, comprising a processing of the flux vector obtained for the current iteration, in which the radius of said circle is obtained, and in which this radius is limited and filtered in time, modifies the norm of the flux vector taking into account the said radius obtained, bounded and filtered.

[0090] In this implementation mode, the radius is bounded and filtered (typically with a low-pass filter). The bounding can be implemented at a value depending on the known manufacturing tolerances of the machine (plus or minus 10% for example), and which can be a function of the measured temperature of the machine for example.

[0091] Also, for wound rotor synchronous machines and asynchronous machines, the radius of the low-pass filtered flux can be limited by a typical value with a tolerance, for example a linear function of the inductor control current for the wound rotor synchronous machine, and a typical value corresponding to the controlled inductor flux for the asynchronous machine.

[0092] It should be noted that for wound rotor synchronous machines and asynchronous machines, the observer presents no contraindication to the controlled inductor flux being changed, as long as the variation is slow in time, i.e. takes place over several electrical revolutions of the machine, which is generally the case in practice, the inductor flux being generally reduced at very high speed. In this case it is not necessary to compensate for the slightly non-circular, but spiral trajectory of the flux. If the inductor flux were to be varied more rapidly, compensation would be made on the memories of past fluxes, so that these past fluxes always constitute a circular (and not a pronounced spiral) trajectory with the flux at the current instant.

[0093] Thus, a possible correction can use a ratio between the norm of the currently commanded inductor flux vector and the norm of the inductor flux commanded in the previous iteration, so as to maintain a theoretically circular desired trajectory even if the intensity of the machine's inductor flux varies.

[0094] With regard to permanent magnet synchronous machines (for example those called in English by the person skilled in the art "brushless DC"), with FCEM magnetization of rather trapezoidal phases, the location of the rotor flux (which is the integral of the locus of the back-EMF) in the plane of the armature, turns out to be of "almost circular" appearance, without inversion of curvature and with a limited harmonic content. For these machines, the recognition of trajectory in an arc of a circle is also fully effective without any particular limitation. In particular, it turns out that the radius of the low-pass filtered flux will converge towards the amplitude of the harmonic 1 of the inductor flux of the machine.

[0095] The choice of analyzing the arc trajectory of the inductor flux, with its center as the error vector, is particularly suitable for the dynamic regimes of the machine (which is suitable for actuators). Indeed, state-of-the-art observers most often treat the inductor flux as a "disturbance" that must be compensated (by the correction) and this is why, in order to ensure convergence, it is necessary that the flux vector or FCEM has varied little between two execution steps, so that the corrector can compensate a sufficient part of it in the given time step. This imposes very high calculation frequencies on non-linear correctors, for example of the "slip mode" type. In order to ensure the stability of the non-linear state model, the assumption of quasi-constant speed is explained in the flux observer of the Shen document.For the implementation mode presented here, advantageously, this assumption is not necessary. The speed can vary abruptly between two iterations, and the execution iterations can be very long (at very low frequency), so that the machine can even have completed a third of an electrical revolution from one execution to the next, and have suddenly slowed down at the next step, without compromising the recognition of the arc trajectory of the flow. This recognition is spatial, and totally independent of the time taken to complete the different points of the trajectory. The proposed observer is therefore particularly suitable for remaining precise even for very dynamic regimes.

[0096] According to a particular embodiment, the method further comprises a modification of the angular coordinate of the observed flux vector by the addition of a compensation term depending on a rotation speed of the motor.

[0097] This speed can be estimated and obtained as the output of a low-pass filter admitting as input data the difference between the current angle of the observed flow obtained by the process and the angle previously obtained, or as the output of a phase-locked loop admitting as input data a vector proportional to the flow obtained by the process.

[0098] According to a particular embodiment, the method comprises a comparison of a distance between the flow vector obtained by said estimation with at least one of said at least two flow vectors obtained at two iterations preceding the present iteration with a given threshold, and in which step ii. is triggered depending on the result of said comparison (typically, the step is triggered if the distance is greater than a threshold, and otherwise a correction can be used with a zero error vector, and / or nothing can be done until a next iteration).

[0099] In this particular embodiment, a distance is used that the person skilled in the art will be able to determine. For example, this distance may be chosen as being obtainable by a direct or indirect calculation (for example the norm of the vector product between segments of the chord of the circle between considered points of the inductor flux), or by a calculation of an arc angle (on the circle, between the considered points of the inductor flux, or a radius of curvature.

[0100] The threshold can be obtained by a calculation using a maximum motor flux (possibly a function of temperature).

[0101] This particular implementation mode is advantageous in that it only makes corrections if the trajectory is long enough for a circular arc to be recognized (to determine the circle). If the stored trajectory is not long enough, risking making the arc unrecognizable due to measurement noise or armature modeling errors, this implementation mode allows the flux to continue to be integrated during the estimation step, without making a correction, until the circular arc traveled by the flux vector is significant enough for a correction to be made. In this implementation, at low speed, the corrections do not are not necessarily performed at each iteration: several estimations (with provision of the estimated angle) are made before a correction is made. In this case it is preferable to also use a low-pass filter in the correction function, in order to avoid jolts in the observed state vector and the observer's exit angle during occasional corrections. This improvement makes it possible to obtain excellent precision at very low rotation speeds. Observability is no longer based on the speed itself, but on the angular displacement achieved. We no longer speak of a temporally corrected observer, but of a spatially corrected one: the correction will be postponed as long as the signal-to-noise ratio condition allowing the circular trajectory to be recognized is not spatially fulfilled.

[0102] According to a particular embodiment, a trajectory is defined by said at least three points, the method further comprising a verification of a condition relating to said trajectory, according to which an arc angle of the trajectory is greater than or equal to a given threshold, and / or according to which the radius of curvature is less than or equal to a given threshold, and in which step ii. is triggered depending on the result of said comparison (typically step ii. is triggered if the radius of curvature is smaller than the given threshold, and it is not triggered otherwise).

[0103] This particular implementation mode proposes to check the level of the curvature radius of the memorized trajectory, including the last estimated flux at the current time. If the trajectory is too straight (therefore the curvature is insufficient), or if the curvature changes sign compared to that of the previous steps, then the correction will not be carried out at the current step, and will be deferred. This makes it possible to reduce the influence of measurement noise and modeling errors, in particular when starting the AC machine during the initialization of the observer, by avoiding false observations. The device will wait for the following iteration steps to more effectively analyze the trajectory as a whole.

[0104] This mode of implementation is therefore particularly useful for initializing the process (other methods of initializing the process can be implemented). work, and for the first iterations, one can use stored values ​​for the necessary elements linked to previous iterations).

[0105] According to a particular mode of implementation, a trajectory is defined by said at least three points (for example the same trajectory as that presented above for another mode of implementation), the method further comprising a verification of a direction of curvature of the trajectory, and in which step ii. is triggered depending on the result of said verification (typically, a change in the direction of curvature can prevent the triggering of step c / ).

[0106] This particular implementation mode can be achieved by using a sign of the curvature determinable by calculation, or by using a notion of concave surface, for any pair of points of the surface limited by the trajectory, the straight line segment which connects these two points is included in the surface. We can also achieve this implementation mode by using tangents to the trajectory which divide the plane into 2 parts at any point; the last straight line segment constituted by the flux vector estimated at the current step being located inside the half-plane defined by the tangent at the previous point and including the center of the (previous) circle.

[0107] According to a particular implementation mode, the iterations are implemented at a frequency lower than the frequency of a PWM signal supplying the motor phases.

[0108] Prior art solutions traditionally use the frequency of PWM signals (e.g. 10kHz) for the observer, which implies the use of appropriate components.

[0109] It has been observed that iterations can be implemented at a lower speed, for example at 2kHz, while maintaining a good level of accuracy.

[0110] The invention also proposes a device for calculating an angle of the inductor flux vector in an observation frame of a polyphase synchronous or asynchronous motor provided with an armature and said inductor, comprising a controller configured to implement a plurality of iterations, each iteration of the plurality of iterations except the first comprising: a / obtaining input data comprising armature voltage values, armature current values, and at least one armature resistance value; b / estimating in the observation frame, the inductor flux vector by adding to an inductor flux vector of the iteration obtained at the previous iteration a numerical integration of a function of said input data, the inductor flux vector pointing towards an estimated inductor flux point and starting from the origin of the observation frame, c / a correction comprising: determining the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the present iteration to deduce therefrom a center vector,so that the corrected inductor flux vector points towards said estimated inductor flux point and starts from a point which approaches the center of said circle, this point being considered as the origin of the observation reference frame for the following iteration, and, obtaining the flux vector of the present iteration in which the flux vector obtained by said estimation is corrected using said center vector, the angle of the calculated inductor flux vector being the angular coordinate of the flux vector obtained by the correction.

[0111] This device can be configured to implement all the methods of implementing the method as defined above.

[0112] The invention also proposes a system comprising a device as defined above, and said motor.

[0113] The invention also proposes an actuator comprising the system as defined above.

[0114] According to a particular embodiment, the actuator is configured to be used within a vehicle transmission, or configured to be a traction or propulsion actuator of the vehicle.

[0115] The invention also provides a computer program comprising instructions for executing the steps of a method according to the invention, when said program is executed by at least one processor.

[0116] In the context of the invention, a computer program may be formed of one or more sub-parts stored in the same memory or in separate memories. The program may use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other desirable form.

[0117] The invention also provides an information medium readable by a processor or a computer and on which is recorded a computer program in accordance with the invention or a set of computer programs in accordance with the invention.

[0118] The information carrier may be any entity or device capable of storing the program. For example, the carrier may comprise a storage means, such as a non-volatile memory or ROM, for example a CD-ROM or a microelectronic circuit ROM. On the other hand, the information carrier may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by a telecommunications network or by a computer network or by other means. The program according to the invention may in particular be downloaded onto a computer network. Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

[0119] The foregoing and other features and advantages will become apparent from the following detailed description. This detailed description refers to the accompanying drawings. Brief Description of the Drawings

[0120] The attached drawings are schematic and are intended primarily to illustrate the principles of the presentation.

[0121] In these drawings, from one figure to another, identical elements (or parts of elements) are identified by the same reference signs.

[0122] [Fig. 1] Figure 1 is a schematic representation of a system according to an example.

[0123] [Fig. 2] Figure 2 shows the steps of a process according to an example.

[0124] [Fig. 3] Figure 3 is a graphical representation of the process according to an example.

[0125] [Fig. 4] Figure 4 graphically shows the angular accuracy of the invention and a prior art solution. Description of the embodiments

[0126] A method for calculating an angle of the inductor flux vector of a motor, a device for calculating an angle of the inductor flux vector of a motor, and a system comprising this device will now be described.

[0127] This method and device make it possible to obtain a more precise calculation of the angle than using the techniques of the prior art, including with an implementation at a frequency lower than the frequency of the PWM signals controlling the motor.

[0128] Figure 1 shows a system SYS comprising a polyphase synchronous or asynchronous motor M provided with an armature and said inductor, and a device 100 configured to implement the method which will be described below. This motor M is mechanically connected (by a shaft) to a reducer 200 and it is controlled by an inverter 300 in a manner known per se, the inverter receiving control signals from the device 100.

[0129] The device 100 has a computer system structure and comprises a controller 101 and a non-volatile memory 102.

[0130] In the non-volatile memory 102 computer program instructions are stored.

[0131] Here, the non-volatile memory 102 comprises instructions of a computer program PG configured to be executed by the processor 101 and to implement a method for calculating an angle of the inductor flux vector in an observation frame of a polyphase synchronous or asynchronous motor provided with an armature and said inductor, comprising a plurality of iterations, each iteration of the plurality of iterations comprising the steps implemented during the execution of the computer program instructions 103, 104, and 105.

[0132] The instructions 103, when executed by the processor 101, result in the implementation of obtaining input data comprising armature voltage values, armature current values, one or more armature resistance values, and possibly one or more inductance values ​​associated with the armature.

[0133] The instructions 104, when executed by the processor 101, lead to the implementation of an estimation in the observation frame of reference, of the inducing flux vector of the iteration by an addition to an inducing flux vector obtained in the previous iteration of a function of said input data, the inducing flux vector pointing towards an estimated inducing flux point and starting from the origin of the observation frame.

[0134] The instructions 105, when executed by the processor 101, lead to the implementation of a correction comprising: a determination of the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the current iteration to deduce a center vector therefrom, and, obtaining the flux vector of the current iteration in which the flux vector obtained by said estimation is corrected using said center vector so that the corrected inductor flux vector points towards said estimated inductor flux point and starts from a point which approaches the center of said circle, this point being considered as the origin of the observation reference frame for the iteration next, the angle of the calculated inductor flux vector being the angular coordinate of the flux vector obtained by the correction.

[0135] Figure 2 is a flowchart showing the steps of a process according to an example. In this process, the Concordia coordinate system is used in % and &.

[0136] The process is initialized in a step S0, this step being implemented only for the first iteration. In this step, we obtain:

[0138] où ,~ ( = ^^0102_4^4 is an approximate and known initial value of the inductor flux, for example estimated with the temperature measurement, controlled on an asynchronous machine or estimated by the initial measurement of the inductor current in a wound rotor synchronous machine. We also have ^ ( ≈ ^ ) the flow in % at the initial iteration and ^ ( ≈ ^ + the flow in & at the initial iteration (the axes of the reference frame are noted in subscript, just like the iteration).

[0139] We will now describe the process for an iteration of index n.

[0140] In a first step S100, input data are obtained comprising armature voltage values, armature current values, one or more armature resistance values, and possibly one or more inductance values ​​associated with the armature.

[0141] In the figure, we have represented the input data E1 which includes armature phase voltages which we note {(V1,V2,V3) i=[1.. K m ]} (in fact, we can memorize these voltages for different instants having indices ranging from 1 to m, corresponding to a measurement window going up to the iteration). We also obtain three-phase armature phase currents {(I1,I2,I3) i=[1.. K m ]},m. It can be noted that we can obtain the E1 data in a table in which we store this data for all the iterations or for some of the iterations which precede the current iteration.

[0142] In step S100, the input data E2 are also obtained, comprising the phase resistance of the armature windings Rs, the cyclic inductance of an armature phase Ls, as well as, for example, a time step ∆T between two electrical measurements (two consecutive indices for the data E1).

[0143] This data may be constant from known information, or it may vary slightly from one iteration to the next.

[0144] In the proposed implementation, no distinction is made between the cyclic inductance in the magnetization axis "Ld" and that at 90° to the magnetization axis "Lq": if the machine has salient poles, it is proposed as an approximation to consider a single inductance value Ls = (Ld + Lq) / 2. Optionally, the inductance can be ignored.

[0145] In step S101, the Concordia transform is used to convert the three-phase input data into data in the equivalent two-phase (α,β) stationary reference frame of the armature:

[0150] We can then implement an estimation S102 of the inductor flux vector of the iteration by an addition to an inductor flux vector obtained in the previous iteration of a function of said input data.

[0151] For example, a state vector X for estimation is the flux vector ^of inductor ^ = ^^^ = F )^+G in the stationary frame of the armature.

[0152] We calculate the estimated state vector for the current iteration:

[0154] Or :

[0159] Ici, le vecteur de sortie estimée est également le vecteur d’état estimé : ^^̂ = ^~ ^ (ie the functions ^^^H^^ = H^ and ^^ = 0).

[0160] It can be noted that advantageously, the output vector includes hidden states of the system, that is to say which are not directly measurable, this output being able to be used in an error vector for comparison with an indirect measurement from an absolute geometric model of the trajectory of the inductor, for a subsequent step, not used in the estimation step.

[0161] A correction is then implemented, the first step of which is a step S102 of determining the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the current iteration to deduce a center vector therefrom. By substantially, we mean that if we use three points, the circle can pass through these points, and if we use more than three points, the circle can approach these three points (we can use a minimization function for this purpose).

[0162] Here, the center of the circle is determined from 3 points (as can be observed in Figure 3 described below), namely the estimated flux ^^^^^ ^)^^ = F^+G and two other fluxes observed and stored in the past ^^^^~ = ^^^^~ ^^^^" = (possibly fixed to given values ​​for the first iterations).

[0163] en notant a vector ^ _ ^^^^ 1 ^^^^ ` ^^ (visible in figure 3) having the following coordinates:

[0167] This vector is a radius of the circle, and therefore:

[0168] ,^̂ =

[0169] Le centre du cercle f^ = passing through these 3 points in the observation, has for coordinates

[0170] Step S103 and step S104, optional, are test steps which allow to have the numerator and the denominator of the coordinates of ^ _ ^^^^ 1 ^^^^ ` ^^ non-zero, and therefore that _ ^^^^^ 1 ^^^^ ` ^^ and f^ are numerically defined.

[0171] In step S103, a comparison is carried out of a distance between the flow vector obtained by said estimation with at least one of said at least two flow vectors obtained at two iterations which precede the present iteration with a given threshold.

[0172] In step S104, a trajectory is defined by said at least three points, the method further comprising a verification (S104) of a condition relating to said trajectory, according to which an arc angle of the trajectory is greater than or equal to a given threshold, and / or according to which the radius of curvature is less than or equal to a given threshold.

[0173] Par exemple, pour l’étape S103, si le carré de la longueur _1g" = %]^^^² + &]^^^² is greater than a minimum threshold ≥ X`[^4^4², corresponding to the square of a minimum chord length parameterized as a fraction of the known minimum flux of the motor, and if the norm of the vector product i^_^^^^^1^^^g^^ × ^ _^^^^1^^^_^^^^^2^i =^|% ]"^]^ ^& ]^^^ −^% ]^^^ ^& ]"^]^ | (step S104) is greater than a minimum curvature threshold (or non-collinearity between these two vectors) X`k7 ^4^4 ,, then f^ and are calculated at the current iteration as expressed previously.

[0174] Sinon (étape S110), ils sont remplacés par les valeurs suivantes : f^ ← ^ 0^ et ,^̂ ← ,~^^^ , where ,~^^^ is the average radius of the circular trajectory of the flow, observed at the output of the low-pass filter at the iteration.

[0175] If both checks S103 and S103 are positive, we can implement the optional step S105 of checking a direction of curvature of the trajectory. In fact, we check here that the sign of the curvature is always the same. For example, a concave surface is defined.

[0176] If the curvature direction check fails, step S120 can be implemented in which the following assignments are made:

[0177] ^^^ ^ ^ ^ ~ ^ ^" ←^^^^ ^ ^ ^ ~ ^ ^^

[0178] ^^^ ^ ^ ^ ~ ^ ^^ ←^^ ^ ~

[0179] And step S110 is implemented.

[0180] The correction itself is implemented otherwise. In this example, the correction will be carried out on the flux vector in a non-linear manner, in two sub-steps:

[0181] First, the correction is carried out (step S106) by the center of the circle, where to do this the error vector is equal to the opposite of the center vector of the circle:

[0182] ^^ = 0 −^f^

[0183] The correction function proposed in this example is simply a gain K, less than or equal to 1:

[0185] We calculate the pre-corrected state vector:

[0187] The same correction is applied to the previously stored flows of the trajectory (this correction being optional):

[0190] Then we carry out the correction (S107) by the radius, by determining an observed ray using a digital low-pass filter of 1 ernumerical gain order 'a' less than 1, on the previously estimated radius, ^.̂

[0191] By taking up the formulation of the observer in estimated state model and correction, we can write for example the filter in the form of an error variable on the radius ^ ^^^ and a correction function ^ ^^^ being a digital gain constant 'a' of low-pass filter, less than 1:

[0194] State modeling extended to the ray realizing a digital low-pass filter of the 1 er order then being:

[0195] ,~ = ~ ^ 1 × ,^^^ + ^^^^^^^^^^

[0196] We then calculate the corrected state vector in a non-linear manner:

[0198] L’angle recherché de l’inducteur par rapport à l’induit, est égal à l’arc tangente n qnr2^^≈ ^^+ , ^≈ ^^) ^ applied to the coordinates of the flux vector observed at the current iteration.

[0199] If the center of the circle could be calculated at the current iteration, in other words if f^ is non-zero, then here the trajectory memories are updated as follows for future iterations (S108):

[0200] ^^^ ^ ^ ^ ~ ^ ^" ←^^^^ ^ ^ ^ ~ ^ ^^

[0201] ^^^ ^ ^ ^ ~ ^ ^^ ←^^^ ^ ≈ ^^^^^^

[0202] Figure 3 is a schematic representation of the determination of the circle and its center. In this figure, the estimated flux vector is represented and the memorized vectors ^^^ ^ ^ ^ ~ ^ ^" and ^^^ ^ ^ ^ ~ ^ ^^ The center of the reference point O is not the center of the circle passing through these three points which is expected at point H.

[0203] The invention aims to have a flux vector starting from H, to take into account the circular trajectory of the flux vector. It involves an update of the reference frame at the iteration, which refocuses on the center of the circle.

[0204] Figure 4 shows in its upper part: the position setpoint of an actuator in dotted line, the speed of the motor in broken line, and the measured position of the actuator shaft, for a given motor and as a function of time.

[0205] The lower part of Figure 4 shows two signals observed by the inventors. The signal in thick line shows the angular error of the flux vector obtained minus the actual electrical angle, by a method according to the invention while the signal in thin line shows the angular error of the flux vector minus the actual electrical angle obtained by the previous solution described in the Shen document. In both cases this angular error is defined as the difference between the angle provided by the method (according to the invention or according to the previous solution) and the actual electrical angle of the motor. The closer this value is to zero, the more accurate the method.

[0206] As can be seen from the figure, we have a more accurate and stable estimation of the flux vector angle with the method according to the example described here.

[0207] For information purposes, implementing the process on an engine can lead to behaviors that illustrate this implementation. Different ways of observing this implementation are possible.

[0208] In a first way, for example, for a synchronous machine of an actuator, the AC machine can be replaced by a wound rotor synchronous machine, equipped with a real angular sensor, of which the inductor current would be abruptly reduced in ramp, without providing the information in the control part so that it is always considered at the nominal inductor flux, in order not to counter-react by limiting the circle as explained previously.

[0209] This abrupt ramp variation can be applied well after initialization, in steady state at moderate speed (approximately 1 / 3 of the base speed for example), and with a load equal to a quarter of the nominal load of the actuator. The ramp should reduce the flux by two-thirds of the nominal value, in a time corresponding to approximately 90 to 180 electrical degrees.

[0210] The expected positive result (i.e. an implementation of the method, here using bounding) may be a stall of the control part due to a very large angular error, or some warning signal of a lack of observability. Indeed, the actual trajectory of the inductor flux will not be circular, but a very tight spiral, inducing particularly false circle center correction vectors. The angular error between the actual position sensor and the observer should be greater than 50°, as long as the flux is reduced.

[0211] A second way of observing an implementation of the process, for any synchronous or induction AC machine, can include adding to the machine a means of drastically modifying its stator inductance L: either by an adjustable external inductance, or by saturation of the magnetic circuit, with the inductor in overcurrent, or another means, etc. The self-inductance of the armature can thus vary quite rapidly in significant proportions (simple to triple, or simple to a third), (in a few hundred µs).

[0212] An implementation of the method would show an operation almost insensitive to this variation, in dynamic or permanent mode, with possibly an increase in the angular error. A machine with an observer according to the prior art is on the other hand particularly sensitive to the inductance parameter of the armature, which intervenes in the estimation step and in their correction step. The expected angular error should be excessive: greater than 50° electrical.

[0213] A third way to observe an implementation of the process, for any synchronous or induction AC machine, is to interpose an electronically controlled polyphase voltage source, in series between the machine M and its inverter 300 illustrated in figure 1. This polyphase voltage source (three-phase for example) can be produced by means of a transformer for example, an inverter connected to the secondary windings of the transformer, and an electronic control.

[0214] In a first step, the polyphase voltage source can be driven to achieve zero voltages, allowing the system under test (analogous to the SYS system of Figure 1) to operate normally. In a second step, the polyphase voltage source would be driven to add or subtract voltages in phase with the back-electromechanical forces of the machine from the voltages applied to the motor. In this way, the control device implementing the method would drive a machine that would virtually have an apparent inductor flux (being the sum of the real flux and the integral of the vector of the polyphase voltage source) that could be different from the actual inductor flux vector of the machine, and that could be suddenly varied by the electronic control of the polyphase voltage source interposed between (at a position analogous to that located between the motor M and the inverter 300).

[0215] By this means, the flux of the virtual machine thus constituted (like the series connection of the real machine with the polyphase voltage source) could be reduced for example by two thirds of the nominal value, in a time corresponding to approximately 90 to 180 electrical degrees, as previously explained.

[0216] An implementation of the method described here may lead to a stall of the control part 100 due to a very large angular error, or to the emission of an observability fault warning signal. Indeed, the trajectory of the estimated inductor flux will not be circular, but a very tight spiral, inducing particularly false circle center correction vectors.

Claims

Claims

1. Method for calculating an angle of the inductor flux vector in an observation frame of a polyphase synchronous or asynchronous motor provided with an armature and said inductor, comprising a plurality of iterations, each iteration of the plurality of iterations except the first comprising: a / obtaining (S100) input data comprising armature voltage values, armature current values, and at least one armature resistance value; b / estimating (S101) in the observation frame, the inductor flux vector of the iteration by adding to an inductor flux vector obtained in the previous iteration a function of said input data, the inductor flux vector pointing towards an estimated inductor flux point and starting from the origin of the observation frame, c / a correction comprising i.a determination (S102) of the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the current iteration to deduce therefrom a center vector, and, ii. obtaining the flux vector of the current iteration in which the flux vector obtained by said estimation is corrected (S106, S107) using said center vector so that the corrected inductor flux vector points towards said estimated inductor flux point and starts from a point which approaches the center of said circle, this point being considered as the origin of the observation reference frame for the following iteration, the angle of the calculated inductor flux vector being the angular coordinate of the flux vector obtained by the correction.

2. Method according to claim 1, in which the flow vector obtained by said estimation is corrected by means of a correction function added to the flow vector obtained by said estimation, the correction function being configured to receive an error vector as input. comprising said center vector, the error vector possibly being assigned a coefficient.

3. Method according to claim 2, in which the correction function is chosen from: a filtering of the error vector, a multiplication by a constant or time-varying gain matrix of the error vector, a multiplication by a variable gain matrix based on a variation calculation or based on a Kalman filter, a non-linear function, for example a gain function times sign function, a sigmoid function, a sign function times square root of absolute value, a sign function times integer or non-integer power function of the absolute value, a hyperbolic tangent function, a sine function r ecevant une entrée bornée à l’intervalle [− ! ! " ;"], une combinaison d’une ouseveral of said non-linear functions.

4. Method according to any one of claims 1 to 3, further comprising a correction of said at least two flux vectors obtained at two iterations preceding the current iteration.

5. Method according to claim 4, wherein the motor is a synchronous motor whose inductor is a wound rotor, or an asynchronous machine, wherein the correction of said at least two flux vectors obtained at two iterations preceding the current iteration is implemented, by determining a ratio between the norm of an inductor flux vector controlled at the current iteration and the norm of the inductor flux controlled or obtained during a previous iteration, or by determining a ratio between the inductor current controlled or obtained at the current iteration and the inductor current obtained during a previous iteration.

6. Method according to any one of the preceding claims, comprising processing the flux vector obtained for the current iteration, in which the radius of said circle is obtained, and in which the limit is set. and this radius is filtered in time, and the norm of the flux vector is modified taking into account said obtained, bounded and filtered radius.

7. Method according to any one of claims 1 to 6, further comprising a modification of the angular coordinate of the observed flux vector by the addition of a compensation term depending on a rotation speed of the motor.

8. Method according to any one of claims 1 to 7, comprising a comparison (S103) of a distance between the flux vector obtained by said estimation with at least one of said at least two flux vectors obtained at two iterations which precede the present iteration with a given threshold, and in which step ii is triggered according to the result of said comparison.

9. A method according to any one of claims 1 to 8, wherein a trajectory is defined by said at least three points, the method further comprising a verification (S104) of a condition relating to said trajectory, according to which an arc angle of the trajectory is greater than or equal to a given threshold, and / or according to which the radius of curvature is less than or equal to a given threshold, and in which step ii is triggered depending on the result of said comparison.

10. A method according to any one of claims 1 to 9, wherein a trajectory is defined by said at least three points, the method further comprising a verification (S105) of a direction of curvature of the trajectory, and in which step ii is triggered depending on the result of said verification.

11. Method according to any one of claims 1 to 10, in which the iterations are implemented at a frequency lower than the frequency of a PWM signal supplying the phases of the motor.

12. Device for calculating an angle of the inductor flux vector in an observation frame of a polyphase synchronous or asynchronous motor provided with an armature and said inductor, comprising a controller. configured to implement a plurality of iterations, each iteration of the plurality of iterations except the first comprising: a / obtaining input data comprising armature voltage values, armature current values, and at least one armature resistance value; b / estimating in the observation frame, the inductor flux vector by adding to an inductor flux vector of the iteration obtained in the previous iteration a numerical integration of a function of said input data, the inductor flux vector pointing towards an estimated inductor flux point and starting from the origin of the observation frame, c / a correction comprising i.a determination of the center of a circle passing substantially through at least three points defined by the flux vector obtained by said estimation, and at least two flux vectors obtained at two iterations preceding the present iteration to deduce therefrom a center vector, so that the corrected inductor flux vector points towards said estimated inductor flux point and starts from a point which approaches the center of said circle, this point being considered as the origin of the observation reference for the following iteration, and, ii. an obtaining of the flux vector of the present iteration in which the flux vector obtained by said estimation is corrected using said center vector, the angle of the calculated inductor flux vector being the angular coordinate of the flux vector obtained by the correction.

13. System comprising a device according to claim 12, and said motor.

14. Actuator comprising a system according to claim 13.

15. An actuator according to claim 14, configured for use within a vehicle transmission, or configured to be a traction or propulsion actuator of the vehicle.

16. A computer program comprising instructions for carrying out the steps of a method according to any one of claims 1 to 12, when said program is executed by a computer.

17. A computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of a method according to any one of claims 1 to 12.

Images