METHOD AND DEVICE FOR ESTIMATING FORCE

Abstract

Some embodiments are directed to a method for estimating a periodic or substantially periodic force present in a mechanical or electromechanical system, the method comprising: estimating, by a processing device, one or more harmonic frequencies of an acceleration signal representing an acceleration in the system, the substantially periodic force contributing to said acceleration; and estimating, by the processing device, the force on the basis of a dynamic model of the system, the dynamic model being defined by said one or more estimated harmonic frequencies.

Description

BACKROUND

Some embodiments relate to the field of methods and devices for estimating force and in particular methods and devices for estimating a force in a mechanical or electromechanical system.

In certain fields it is desirable to be able to estimate one or more variable forces influencing a system. For example, in the case of a motor-assisted bicycle, it could be desirable to be able to estimate the force exerted by the cyclist. Similarly, in the case of a wind turbine, it could be desirable to be able to estimate the force generated by wind.

In the case of motor-assisted bicycles, using a torque or power detector to detect the force exerted by each foot of the cyclist has been proposed. Such torque or power detectors are based on the detection of a mechanical deformation, for example the twisting of the shaft in the pedal shaft. Placing force detectors in the pedals of a bicycle has also been proposed. However, one of the drawbacks of these solutions is that their precision is generally affected by temperature variations and by the ageing of the materials used, which means frequent calibration is may be necessary. In addition, such detectors add weight to the structure and are relatively costly to install.

There is therefore a need in this field for a method and a device for estimating forces that solve some or all or most of the drawbacks mentioned above.

SUMMARY

One purpose of the embodiments of the present description is to satisfy one or more of the needs of the related art, at least partially.

According to some embodiments, a method for estimating a periodic or substantially periodic force present in a mechanical or electromechanical system is provided, the method including: estimating one or more harmonic frequencies of an acceleration signal in the system using a processing device, with the substantially periodic force contributing to the acceleration; and, using the processing device, estimating the force based on a dynamic model of the system, the dynamic model being defined by the one or more estimated harmonic frequencies.

According to some embodiments, the dynamic model is updated based on an error signal representing the difference between a sensed speed signal and an estimated speed signal.

According to some embodiments, the method also includes the generation of an acceleration signal based on a differential calculation over time for two or more values of a speed signal representing an angular or linear speed in the system.

According to some embodiments, the estimation of the one or more harmonic frequencies of the acceleration signal involves the calculation of an error signal equal to:

e ₁(k)=r(k)−φ^T{circumflex over (θ)}(k)

• where r(k) is a current acceleration value and φ^T{circumflex over (θ)}(k) is an estimation of the acceleration based on a vector {circumflex over (θ)}(k) representing the harmonic frequencies and a vector

T representing one or more previous acceleration values.

According to some embodiments, the mechanical or electromechanical system is a motor-assisted bicycle and the periodic or substantially periodic force is the pedalling force generated by the cyclist, and the dynamic model includes one or more estimations {circumflex over (v)}(k), {circumflex over (F)}_H(k), {circumflex over (F)}_H(k) and z(k), in which each of these estimations is defined as follows:

$\hat{v}\left(k\right)=\hat{v}\left(k-1\right)+\frac{\mathrm{Ts}}{M}·F_M\left(k\right)+\frac{\mathrm{Ts}}{M}·\widehat{\stackrel{\_}{F}}\left(k-1\right)+\frac{\mathrm{Ts}}{M}·{\hat{\tilde{F}}}_H\left(k-1\right)$ $\widehat{\stackrel{\_}{F}}\left(k\right)=\widehat{\stackrel{\_}{F}}\left(k-1\right)$ ${\hat{\tilde{F}}}_H\left(k\right)=\hat{z}\left(k-1\right)$ $\hat{z}\left(k\right)=-{\hat{\theta }}_2\left(k\right)·\hat{\tilde{F}}\left(k-1\right)-{\hat{\theta }}_1\left(k\right)·\hat{z}\left(k-1\right)$

where Ts is the sampling period, M is the mass of the cyclist and the bicycle, F_M is the force generated by the motor, {circumflex over (θ)}₁(k) and {circumflex over (θ)}₂(k) represent one of the harmonic frequencies, {circumflex over ({tilde over (F)})}_H(k−1) is the estimation of the force exerted by the cyclist, and {circumflex over ({tilde over (H)})}(k−1) is an estimation of other forces in the system.

According to some embodiments, the mechanical or electromechanical system is a wind turbine and the periodic or substantially periodic force is a periodic component of the force of the wind on the blades of the turbine, and the dynamic model includes one or more estimations {circumflex over (ω)}_T(k), {circumflex over (F)}(k) and {circumflex over (z)}(k) in which each of these estimations is defined as follows:

${\hat{\omega }}_T\left(k\right)={\hat{\omega }}_T\left(k-1\right)-\left(\frac{\mathrm{Ts}}{I}\right){\tau }_{\mathrm{generator}}\left(k\right)+\left(\frac{\mathrm{Ts}}{I}\right)\left({\hat{\tilde{\tau }}}_{\mathrm{wind}}+{\widehat{\stackrel{\_}{\tau }}}_{\mathrm{wind}}\right)$ $\widehat{\stackrel{\_}{F}}\left(k\right)=\widehat{\stackrel{\_}{F}}\left(k-1\right){\hat{\tilde{F}}}_H\left(k\right)=\hat{z}\left(k-1\right)\hat{z}\left(k\right)=-{\hat{\theta }}_2\left(k\right)·\hat{\tilde{F}}\left(k-1\right)-{\hat{\theta }}_1\left(k\right)·\hat{z}\left(k-1\right)$

where Ts is the sampling period, {circumflex over (θ)}₁(k) and {circumflex over (θ)}₂(k) represent one of the harmonic frequencies, {circumflex over ({tilde over (F)})}_H(k−1) is the estimation of the force exerted by the wind, and {circumflex over (F)}(k−1) is an estimation of other forces in the system, where ω_T is the speed of the turbine and {circumflex over (ω)}_T(k) is an estimation of the speed of the turbine, I is the inertia of the turbine, I_generator is the torque applied by the generator, for example equal to Ke·I_generator, where Ke is the speed constant of the motor and I_generator is the current generated by the generator, and {circumflex over ({tilde over (τ)})}_wind+{circumflex over (τ)}_wind represents the total torque generated by the wind, {circumflex over ({tilde over (τ)})}_wind representing a cyclical component of this torque.

According to some embodiments, the one or more harmonic frequencies include at least one from among the fundamental frequency, the first harmonic frequency, the second harmonic frequency, and the third harmonic frequency.

According to some other embodiments, a processing device configured to estimate a periodic or substantially periodic force present in a mechanical or electromechanical system is provided, the processing device being configured to: estimate one or more harmonic frequencies of an acceleration signal representing an acceleration in the system, with the substantially periodic force contributing to the acceleration; and estimating the force based on a dynamic model of the system, the dynamic model being defined by the one or more estimated harmonic frequencies.

BRIEF DESCRIPTION OF THE FIGURES

The subject matter and advantages mentioned above, in addition to others, will become clear from the following detailed description of embodiments given for non-limiting, illustration purposes in reference to the enclosed Figures in which:

FIG. 1 shows a motor-assisted bicycle according to an embodiment example of the present description;

FIG. 2 shows a schematic diagram of a force estimation system according to an embodiment example;

FIG. 3 is a flowchart illustrating steps in a force estimation method in a mechanical or electromechanical system according to an embodiment example;

FIG. 4 is a graph showing frequency components of an acceleration signal according to an embodiment example;

FIG. 5 is a schematic illustration of a force estimation method in a mechanical or electromechanical system according to an embodiment example;

FIG. 6 is a more detailed schematic illustration of a portion of the method in FIG. 6 according to an embodiment example;

FIG. 7 is a more detailed flowchart showing the steps in the method of FIG. 3 according to an embodiment example;

FIG. 8A is a graph illustrating an acceleration signal and a resultant force according to an embodiment example;

FIG. 8B is a graph illustrating an acceleration signal and a dynamic constraint according to an embodiment example;

FIG. 9 is a schematic illustration of a computer device according to an embodiment example; and

FIG. 10 illustrates a wind turbine according to an embodiment example.

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS

The embodiments described here refer to a method and a device intended for estimating a periodic or substantially periodic force in a mechanical or electromechanical system. The term “periodic force” is used here to refer to a force which is cyclical in nature, at least within certain limits. For example, the force exerted by a cyclist on the pedals of a bicycle is periodic because the force will generally come from the downward thrust applied by the cyclist's foot when each pedal reaches a certain angular position range in its cycle of rotation. There are other mechanical or electromechanical systems in which a periodic or substantially periodic force is present and may be estimated by the techniques described herein. For example, the present inventor has discovered that the force exerted by wind on the blades of a wind turbine generally has a periodic or substantially periodic component which can be estimated, as will be described in greater detail below. One of ordinary skill in the art will be aware of other application in which a periodic or substantially periodic force could be estimated using the techniques described herein, such as for example the force exerted on an oar by a rower or by waves on a boat.

FIG. 1 shows a motor-assisted bicycle 100 according to an embodiment example. Bicycle 100 includes, for example, an electric motor 102. In the example of FIG. 1, motor 102 is, for example, a roller motor pressing against a tire and installed in the vicinity of pedal set shaft 103, so that the motor applies force directly to the rear wheel tire. In some other embodiments, motor 102 could apply force to pedal set shaft 103, or it could be installed elsewhere, such as on the hub of the front or rear wheel. Motor 102 is, for example, an electric motor powered by a battery 104, which is installed, for example, on a rack placed above the rear wheel of the bicycle. In some other embodiments, battery 104 could be installed elsewhere on the bicycle. The motor, for example, is controlled by a computer 106, which is installed, for example, on the handlebars of the bicycle, although, in this case as well, it could be installed elsewhere. Computer 106, for example, allows the cyclist to select the level of assistance provided by the motor.

Computer 106 is designed, for example, to estimate the useful force exerted by the cyclist. For instance, the useful force is the force that results in an acceleration of the bicycle, and, for example, does not include certain components of the force that are lost to due to friction, etc. The useful force will generally correspond to the torque applied to pedal set shaft 103 by the pedal cranks. In certain embodiments, computer 106 includes a display and is designed to display an indication of the estimated force.

In addition to, or in its place, computer 106 is designed, for example, to control motor 102 based on the estimated force. For example, the level of assistance provided by motor 102 is controlled in such a way that the overall acceleration of the bicycle stays relatively constant, for instance within a range of 10% of the average value. In other words, the motor is controlled, for example, in such a way as to provide greater assistance between the cyclist's pedal strokes. As a variation, computer 106 can be designed to control motor 102 in a different way based on the estimated force, for example in order to provide a level of assistance that is proportional or inversely proportional to the estimated force.

FIG. 2 shows a schematic illustration of a force estimation system 200 according to an embodiment example. For instance, system 200 is partly implemented by computer 106 and partly implemented by detectors and sensors, which will be described in greater detail below.

System 200 includes, for example, a motor force detector 202 which, for instance, provides signal F_M indicating the force applied by the motor at a given instant. System 200 also includes, for example, a pedalling sensor 204, which provides a signal ω_P indicating the angular velocity of the pedal cranks. This signal can be used, for example, to determine speed ω_B of the bicycle. In addition to or in place of the pedalling sensor, a wheel speed sensor 206 is, for instance, provided and generated a signal ω_B indicating the speed of the bicycle. In some other embodiments, signal ω_B indicating the speed of the bicycle could be provided directly by the motor, in which case the wheel speed sensor could be omitted. Signal ω_B can include or can consist of, for example, of digital samples and includes at least five samples for each complete rotation of a pedal crank, and possibly at least seven samples for each complete rotation of a pedal crank.

The system of 200 furthermore includes an unknown force estimator 208, implemented for example by computer 106 and designed to estimate an unknown force F in the system, the force corresponding to the useful force exerted by the cyclist. Estimator 208 receives, for example, signal F_M and an indication of the ground of the bicycle and of the cyclist. In certain embodiments, the mass is provided by a mass estimator 210 which, for example, receives the motor force F_M from motor force detector 202 and also the speed of the bicycle ω_B, and estimates the mass based on these parameters while the cyclist is not pedalling. As a variant, the mass can be obtained by an information input provided by the user.

In certain embodiments, the unknown force estimator also receives a pedal speed signal ω_P from pedalling sensor 204 and/or a bicycle speed signal ω_B from a wheel speed sensor. Estimator 208 also receives, for example, a cyclist constraint force F_C coming from a previous force calculation, which is used as a basis for calculating a subsequent force value during a subsequent iteration, as described in greater detail below.

Unknown force estimator 208 provides, for example, an estimation F of the useful force generated by the cyclist in the system, a force which contributes to the acceleration ω_B of the bicycle. A kinematic constraint generator 212 receives, for example, estimated forces F and the pedalling speed ω_P, and generates a cyclist constraint force F_C, which is provided through a feedback path to unknown force estimator 208. The cyclist constraint force signal F_C is also supplied, for example, to a cyclist power estimator 214, which also receives, for example, the speed of the bicycle ω_B and generates the estimated cyclist power Power_cyclist(k). In particular, it would be clear to one of ordinary skill in the art that, while a cyclist force is estimated in the described embodiments, in certain embodiments this force could be expressed in the form of a power value equal to the force multiplied by the distance over time. For example, in certain embodiments the power in watts can be obtained by the following equation:

Power_Cyclist(k)=|{tilde over (F)}_H19 v(k)|

where v(k) is the speed of the bicycle and {tilde over (F)}_H is the estimated useful force of the cyclist.

As a variation, rather than being expressed in the form of a linear force, the force exerted by the cyclist could be expressed in the form of a torque. For example, the torque in Nm can be obtained by the following equation:

$\mathrm{Pedalling}{\mathrm{Torque}}_{\mathrm{Cyclist}}\left(k\right)=\frac{{\mathrm{Power}}_{\mathrm{Cyclist}}}{{\omega }_p\left(k\right)}.$

In yet another example, the acceleration in m/s² generated by the cyclist could be obtained by the following equation:

${\mathrm{Acceleration}}_{\mathrm{Cyclist}}\left(k\right)=\frac{{\tilde{F}}_H}{M}$

where M is the mass of the cyclist and the bicycle.

Of course, in certain embodiments the motor could be turned off so that the force of the motor F_M is equal to zero.

FIG. 3 is a flowchart illustrating the steps in a method for determining the force of a cyclist on a motor-assisted bicycle, such as bicycle 100 in FIG. 1. The method is implemented, for example, by computer 106.

In step 301, a reading of speed v(k) of the bicycle is obtained, for instance. In certain embodiments, this speed can be generated based on the signal ω_B coming from a wheel speed sensor or another input. As a variation, speed v(k) can be calculated based on a reading of the motor speed ω_MOTOR, in radians per second, and based on the radius of the motor Rm, with, for example, v(k)=ω_MOTOR·Rm.

In step 302, an acceleration value a(k) is calculated based on the speed value, for instance by a differential calculation of the speed signal over time.

In step 303, the frequency of one or more harmonics of the acceleration signal a(t) is determined, for example. The term “harmonic” is used to refer to a fundamental frequency and/or the first, second, third, etc., harmonic frequencies. In an embodiment described in greater detail below, the harmonic frequencies are determined based on an iterative algorithm.

FIG. 4 is a graph illustrating an example of the amplitudes of frequency components of the acceleration signal. The frequency axis shows, for example, the normalized frequency in the form of fractions of the sampling frequency 1/Ts, where Ts is the sampling period. In some embodiments, the sampling frequency is equal to 20 Hz and therefore the sampling period Ts is equal to 50 ms. More generally speaking, the sampling frequency is for instance between 10 and 50 Hz.

In the example of FIG. 4, the harmonic frequencies correspond to peaks on the frequency distribution curve labelled 402, 404, and 406, which for instance are at about 3, 5, and 9 Hz, respectively. There is also, for example, a direct current component, DC, of the acceleration signal represented by a peak 408 at 0 Hz. Step 303 in FIG. 3 involves, for example, an estimation of the frequency of one or more of harmonics 402, 404, and 406.

Referring once again to FIG. 3, in step 304 a model is generated for the forces present in the bicycle system, based on the harmonic frequencies. This model includes a force component {tilde over (F)}_H(k).

In step 305, the force component {tilde over (F)}_H(k) is extracted from the model in order to obtain the estimation of the useful force exerted by the cyclist.

In certain embodiments, another step 306 involves determining an error value associated with the estimated force {tilde over (F)}_H(k).

FIG. 5 is a block diagram showing steps 302 to 305 of FIG. 3 in greater detail.

Block 502 in FIG. 5 represents the implementation of steps 302 and 303 in FIG. 3. Block 502 receives the speed value v(k) and, based on a model of a transfer function H({circumflex over (θ)},Z), calculates the harmonics θ(k) updated based on the speed value v(k).

Block 504 in FIG. 5 represents the implementation of steps 304, 305, and 306 in FIG. 3. Block 504 receives the speed value v(k), the harmonics θ(k) generated by block 502, and the force of the motor F_M. The force of the motor F_M can be determined, for example, based on the speed constant of the motor Ke and a measurement of the current I supplied to the motor. In particular, the force of the motor F_M is generated, for example, based on the motor torque T_M, equal to, for example, Ke*I, where Ke is the speed constant of the motor and I is the current. The force of the motor F_M is therefore equal, for example, to T_M/Rm, where Rm is the radius of the motor, in the case of a roller motor on a tyre, such as motor 102 in FIG. 1. Block 504 generates, for example, an estimation {tilde over (F)}_H(k) of the useful force exerted by the cyclist. In certain embodiments, block 504 also generates an error value indicating the margin of error in the force estimation.

FIG. 6 illustrates block 504 of FIG. 5 in greater detail according to an embodiment example.

A dynamic model 606 receives, for example, the force of the motor F_M(k), the harmonic frequencies θ(k), and an error value e₂(k) obtained by subtracting an estimation v(k) of the speed from the speed value v(k). Block 606 provides, for example, an updated dynamic model Xe(k). Block 608 represents the extraction of the force estimation {tilde over (F)}_H(k) from the dynamic model Xe(k). Another block 610 represents the determination of the error associated with the force estimation.

FIG. 7 is a more detailed block diagram representing the method of FIG. 3 according to an embodiment example. After step 301 in which the speed value v(k) is read, step 302 involves, for example, calculation of the acceleration value a(k) in the form of a value {dot over (ω)}_P(k) based on the pedalling speed ω_P(k). For example, the pedal acceleration value {dot over (ω)}_P(k) is calculated by the following equation:

${\stackrel{.}{\omega }}_P\left(k\right)=\frac{\left({\omega }_p\left(k\right)-{\omega }_p\left(k-1\right)\right)}{\mathrm{Ts}}$

where ω_P(k−1) is a previous value of the pedal speed and Ts is the sampling period, for example, equal to the time period between samples ω_p(k−1) and ω_p(k) of the pedal speed.

Of course, instead of being based on the pedal speed ω_p(k), in some other embodiments the acceleration value a(k) could be calculated based on another speed signal.

Steps 703 to 705 in FIG. 7 implement step 303 of FIG. 3, which involve finding one or more harmonics of the acceleration signal.

Step 703 involves, for example, calculating a vector φ in the form of [r(k−1); r(k−2)], where r(k−1)={dot over (ω)}_P(k−1) and r(k−2}={dot over (ω)}_P(k−2).

Step 704 involves, for example, calculating an error value e₁(k) and a parameter L(k).

The error value e₁(k) is, for example, based on the following formula:

e ₁(k)=r(k)−φ^Tθ(k)

• where r(k)={dot over (ω)}_P(k) and {circumflex over (θ)}(k) is, for example, a vector representing an estimation of the harmonic frequencies initialized to zero, and φ^T{circumflex over (θ)}(k) is an estimation of the acceleration.

The parameter L(k) is, for example, based on the following formula:

L(k)=P(k)φ(φ^TP(k)φ+FF)

• where FF is a forgetting factor, for example equal to 0.95, and P(k) is a matrix which is initialized, for example, to a certain value, such as a value of:

$P\left(k\right)=\left(\begin{array}{cc}100& 0\\ 0& 100\end{array}\right)$

and is recalculated for each new iteration by the following formula:

$P\left(k+1\right)=\left(\frac{1}{\mathrm{FF}}\right)·\left(P\left(k\right)-L\left(k\right){\varphi }^TP\left(k\right)\right).$

In step 705, a harmonic vector {circumflex over (θ)}(k+1) is generated, for example, based on the following formula:

{circumflex over (θ)}(k+1)={circumflex over (θ)}(k)+L(k)·e₁(k)

For example, in the case of a single harmonic, the harmonic vector {circumflex over (θ)}(k+1) has, for instance, the following form:

$\left(\begin{array}{c}{\hat{\theta }}_1\left(k+1\right)\\ {\hat{\theta }}_2\left(k+1\right)\end{array}\right)\hspace{1em}$

where {circumflex over (θ)}₁ represents the frequency of the harmonic, for example in the form {circumflex over (θ)}₁=−2 cos(ω·T_S) where ω=2Πf, with f being the frequency of the harmonic, and {circumflex over (θ)}₂ represents the quality factor of the harmonic, which is for instance close to 1.

Step 304 involves, for example, sub-steps 706 and 707. In these steps, a dynamic model Xe representing the bicycle system is modified, for example, based on the last speed value v(k). For example, the dynamic model is based on the following equation for the drive force of the bicycle:

$\mathrm{Total}\mathrm{Force}=M·\stackrel{.}{v\left(t\right)}=M\left(\frac{v\left(k\right)-v\left(k-1\right)}{\mathrm{Ts}}\right)=F_M+{\tilde{F}}_H+\stackrel{\_}{F}$

where {tilde over (F)}_H represents the force exerted by the cyclist and F represents the other forces contributing to the overall force of the bicycle, such as wind, terrain, etc. Based on this equation, the next dynamic model Xe can be defined, for example, as being a vector having the following components:

$\mathrm{Xe}=\left(\begin{array}{c}\hat{v}\left(k\right)\\ \widehat{\stackrel{\_}{F}}\left(k\right)\\ {\hat{\tilde{F}}}_H\left(k\right)\\ \hat{z}\left(k\right)\end{array}\right)$

where F_H(k−1), {circumflex over (v)}(k), {circumflex over (F)}(k), {circumflex over (Z)}(k) and {circumflex over (Z)}(k) are defined as follows:

$\begin{array}{cc}{\tilde{F}}_H\left(k-1\right)=F_C\left(k\right)& \sim 1\\ \hat{v}\left(k\right)=\hat{v}\left(k-1\right)+\frac{\mathrm{Ts}}{M}·F_M\left(k\right)+\frac{\mathrm{Ts}}{M}·\widehat{\stackrel{\_}{F}}\left(k-1\right)+\frac{\mathrm{Ts}}{M}·{\hat{\tilde{F}}}_H\left(k-1\right)& \sim 2\\ \widehat{\stackrel{\_}{F}}\left(k\right)=\widehat{\stackrel{\_}{F}}\left(k-1\right)& \sim 3\\ {\hat{\tilde{F}}}_H\left(k\right)=\hat{z}\left(k-1\right)& \sim 4\\ \hat{z}\left(k\right)=-{\hat{\theta }}_2\left(k\right)·\hat{\tilde{F}}\left(k-1\right)-{\hat{\theta }}_1\left(k\right)·\hat{z}\left(k-1\right)& \sim 5\end{array}$

In step 706, a matrix Ad(k) and vectors Ld(k) and P₂(k+1) are calculated, for example.

Matrix Ad(k) is calculated, for example, based on the following formula:

$\mathrm{Ad}\left(k\right)=\left[\begin{array}{cccc}1& \mathrm{Ts}/M& \mathrm{Ts}/M& 0\\ 0& 1& 0& 0\\ 0& 0& 0& 1\\ 0& 0& -{\hat{\theta }}_2\left(k\right)& -{\hat{\theta }}_1\left(k\right)\end{array}\right]$

where Ts is the sampling period and M is the mass of the bicycle and the cyclist.

Vector Ld(k) is calculated, for example, based on the following formula:

Ld(k)=((Cd·P₂·Cd^T+Vd)·Cd·P₂·Ad^T)^T

• where Cd is the vector [1 0 0 0], for example, and Vd is a constant representing the covariance expected from the speed measurements.

Vector P₂(k+1) is calculated, for example, based on the following formula:

P ₂(k+1)=Ad·P₂(k)·Ad^T+Wd−Ad·P₂(k)Cd^TLd^T

• where Wd is a constant matrix representing the covariance expected from the process disturbances, that is, the covariance of the exogenous forces.

In step 707, the vector Xe(k+1) is calculated, for example, based on the following formula:

Xe(k+1)=(Ad−Ld·Cd)Xe(k)+Bd·F_M(k)+Ld·e₂(k)

• where e₂(k) is an error value equal to v(k)−{circumflex over (v)}(k) and {circumflex over (v)}(k) is an estimation of the speed v(k).

In step 305, the force {tilde over (F)}_H exerted by the cyclist is extracted, corresponding for example to the third element Xe(3) of vector Xe. In addition, the other forces F can be extracted, the forces corresponding, for example, to the second element Xe(2) of vector Xe.

Step 306 involving determining, for example, an error value ERROR_{tilde over (F)} associated with the estimated force, as shown in block 610 in FIG. 6. The error value is determined, for example, based on the following equation for an error vector of vector Xe:

ERROR=√{square root over (diag(P₂(k))·e₂²(k))}

The error value ERROR {tilde over (F)} is then extracted, for example, as the third element ERROR(3) of the ERROR vector. In certain embodiments, the amplitude of the error value ERROR {tilde over (F)} is then compared to an admissible level to decide whether the force estimation is sufficiently accurate to be useful. For example, the following operation is implemented:

If |ERROR_{{tilde over (F)}}|ε Then Force Estimation is “admissible,” Otherwise “not admissible”

• where ε is the admissible level, for example equal to a value of between 2 and 10 percent, and in certain embodiments between 2 and 5 percent.

Optionally, in step 708 the motor of the bicycle is controlled based on the estimated force if, for example, the estimated force is determined in step 306 to be admissible.

FIG. 8A is a graph including a curve 802 representing an example of estimated acceleration produced by the cyclist before the constraints are considered. Dashed lines 804 and 806 represent the maximum and minimum accelerations, respectively, due to gravity, assuming a maximum road slope of 20 percent. Curves 808 and 810 represent other constraints, calculated for example in real time by module 212 in FIG. 2. In particular, curve 808 represents the maximum acceleration generated by the weight of the cyclist on the pedals, with this signal dropping for example to zero during periods when the cyclist stops pedalling. Curve 810 represents, for example, the maximum acceleration in the presence of aerodynamic losses.

FIG. 8B is another graph illustrating constraints 804, 806, 808, and 810 of FIG. 8A, and also illustrating a curve 812 representing the constraint force F_C after the constraints have been taken into account. In particular, curves 804, 808, and 810 represent maximum constraint values Cmax(t), whereas curve 806 represents a minimum constraint value Cmin(t).

The constraint force F_C is, for example, equal to the estimated unconstrained force of the cyclist F if none of the constraints is exceeded, for example, since Cmin(t)<=F<=Cmax (t). As a variation, if F<Cmin(t), FQ is for example equal to Cmin(t), and if F>Cmax(t), then FQ is for example equal to Cmax(t).

FIG. 9 is a schematic illustration of a computer 900 configured to implement the method of FIG. 3 and/or of FIG. 7 in order to calculate a cyclist force, and which corresponds, for example, to computer 106 in FIG. 1.

Computer 900 includes, for example, a processing device (P) 902 including one or more processors under the control of instructions stored in instruction memory 904. An input interface (I/O INTERFACE) 906 makes it possible, for example, to enter readings into processing device 902 coming from various measurement devices, such as those coming from, for example, the electric motor and/or a separate speed sensor. Memory 908, for example, stores the various parameters, vectors, and matrices described earlier for implementation of the method.

Instead of being used to calculate an estimation of the force exerted by a cyclist, in some other embodiments the methods of FIGS. 3 and 7 and the device of FIG. 9 could be applied to estimate the force exerted by the wind on the blades of a wind turbine, as will now be described in greater detail in reference to FIG. 10.

FIG. 10 illustrates an example of a wind turbine 1000 including two or more blades 1002 connected to a shaft 1004, which is in turn connected to an electrical power generator 1006.

Of equations 1 through 5 indicated earlier and representing the dynamic model of forces on the bicycle, equations 1, 2, and 5 remain unchanged, for example, while equations 2 and 3 are replaced, for example, with the following equations 1′, 2′, and 3′:

$\begin{array}{cc}{\hat{\omega }}_T\left(k\right)={\hat{\omega }}_T\left(k-1\right)-\left(\frac{\mathrm{Ts}}{I}\right){\tau }_{\mathrm{generator}}\left(k\right)+\left(\frac{\mathrm{Ts}}{I}\right)\left({\hat{\tilde{\tau }}}_{\mathrm{wind}}+{\widehat{\stackrel{\_}{\tau }}}_{\mathrm{wind}}\right)& \sim 1^{\prime }\\ {\widehat{\stackrel{\_}{\tau }}}_{\mathrm{wind}}\left(k\right)={\widehat{\stackrel{\_}{\stackrel{\_}{\tau }}}}_{\mathrm{wind}}& \sim 2^{\prime }\\ {\hat{\tilde{\tau }}}_{\mathrm{wind}}\left(k\right)=\hat{z}\left(k-1\right)& \sim 3^{\prime }\end{array}$

where ω_T is the speed of the turbine and {circumflex over (ω)}_T(k) is an estimation of the speed of the turbine, Ts is the sampling period, I is the inertia of the turbine, τ_generator is the torque applied by the generator, equal for example to Ke·I_generator, where Ke is the constant speed of the motor and I_generator is the current generated by the generator, and {circumflex over ({tilde over (τ)})}_wind+{circumflex over (τ)}_wind represents the total torque generated by the wind, with {circumflex over (τ)}_wind representing a cyclical component of this torque. The speed of the turbine ω_T is for example detected by a speed sensor located on the main shaft of the turbine and/or using a generator speed measurement based, for example, on rotation encoders.

The speed of the generator depends both on the torque produced by the wind and the torque produced by the generator, which is proportional to the electrical current I_generator drawn from the generator. In certain embodiments, the estimation of the torque {tilde over (τ)}_wind produced by the wind can be used to control the speed of the generator by controlling the current level I_generator. One advantage in controlling the torque of the generator in this way is that cyclical fluctuations in the speed of blades 1002 can be avoided, thus avoiding or reducing the risk of damage to the turbine.

One advantage of the embodiments described here is that a periodic or substantially periodic force in a mechanical or electromechanical system can be estimated practically in real time and with relatively high precision in a relatively easy way.

A description of a sample embodiment having now been given, various alterations, modifications, and improvements or enhancements will become readily clear to one of ordinary skill in the art. For instance, although a particular example of a dynamic model is given in the aforementioned equations 1 through 5, it will be clear to one of ordinary skill in the art that modifications could be made to these equations to take into account, for instance, additional forces present in the system, and that the equations could be adapted to applications other than the examples of a bicycle and wind turbine described herein.

Claims

1. A method for estimating a periodic or substantially periodic force present in a mechanical or electromechanical system, the method comprising: estimating one or more harmonic frequencies of an acceleration signal in the system using a processing device, with the substantially periodic force contributing to the acceleration; andusing the processing device, estimating the force based on a dynamic model of the system, the dynamic model being defined by the one or more estimated harmonic frequencies.

2. The method according to claim 1, wherein the dynamic model is updated based on an error signal (e2(k)) representing the difference between a sensed speed signal (v(k)) and an estimated speed signal ({circumflex over (v)})(k)).

3. The method according to claim 1, further including the generation of an acceleration signal based on a differential calculation over time for two or more values of a speed signal representing an angular or linear speed in the system.

4. The method according to claim 1, wherein the estimation of the one or more harmonic frequencies of the acceleration signal involves the calculation of an error signal equal to: e1(k)=r(k)−φT{circumflex over (θ)}(k)

5. The method according to claim 1, wherein the mechanical or electromechanical system is a motor-assisted bicycle and the periodic or substantially periodic force is the pedalling force generated by the cyclist, and in which the dynamic model includes one or more of estimations {circumflex over (v)}(k), {circumflex over (F)}(k), {circumflex over ({tilde over (F)})}H(k) and {circumflex over (Z)}(k), in which each of these estimations is defined as follows:

6. The method according to claim 1, wherein the mechanical or electromechanical system is a wind turbine and the periodic or substantially periodic force is a periodic component of the force of the wind on the blades of the wind turbine, and in which the dynamic model includes one or more of estimations {circumflex over (ω)}T(k), {circumflex over (F)}H(k), {circumflex over ({circumflex over (F)})}H(k) and {circumflex over (Z)}(k), in which each of these estimations is defined as follows:

7. The method according to claim 1, wherein one or more harmonic frequencies include at least one from among the fundamental frequency, the first harmonic frequency, the second harmonic frequency, and the third harmonic frequency.

8. A processing device configured to estimate a periodic or substantially periodic force present in a mechanical or electromechanical system, the processing device being configured to: estimate one or more harmonic frequencies of an acceleration signal representing an acceleration in the system, with the substantially periodic force contributing to the acceleration; andestimate the force based on a dynamic model of the system, the dynamic model being defined by the one or more estimated harmonic frequencies.