CN101226068A - System and method for testing dynamic friction parameter - Google Patents

Abstract

The invention discloses a test system and method of dynamic friction parameter, the test system is characterized in that's: an absolute encoder 14 is arranged on a motor 13, a moment sensor 4 is arranged axially between a test part, a weight fixing device 6 and a locking mechanism 2, the test part is arranged at upper and lower layers of the weight fixing device, when testing, the operation of the motor is controlled by a controller, moreover the parameters including the sampling time, friction moment, rotating speed, rotating angle and motor winding current are recorded synchronously, the evaluated value of current friction parameters are set at Sigma 0, Sigma 1, Fc, Fs, Omegas, and Sigma, and then differential equation of systematic movement is solved so as to obtain the evaluated value Theta(t) of moment motor rotating angle of each sampling time, thereby respectively calculating the errors of the evaluated values of the moment motor rotating angle and the friction moment of each sampling time, the error parameters are contrasted with the set testing parameters so as to meet the required evaluated value of the output friction parameters, otherwise correction is re-made. The test system and method of dynamic friction parameter has the advantages of convenient testing and high precision. The test system is applicable for dynamic friction parameter test on various materials.

Description

Dynamic friction parameter test system and method for testing

Technical field

The invention belongs to technical field of measurement and test, particularly relate to the measurement of dynamic friction parameter, be used to instruct the design of high-performance servo-drive system.

Background technology

In servo-drive system, have on two surface of contact of relative motion or relative motion trend and can produce friction force.Friction has reduced the stability and the stable state accuracy of system, has caused the low speed jerking motion phenomenon, and the raising of system performance has been constituted serious obstacle.A large amount of studies show that friction is a kind of physical phenomenon of complexity, friction factor not only and the speed of related movement nonlinear dependence between surface of contact, the static characteristics that is referred to as to rub, but also the dynamic effects such as stiction that have the friction memory, increase.Armstrong B, Dupont P, Canudas de Wit C. is at " Automatica ", the paper of delivering on 1994 30 the 7th phases of volume " A Survey of Models, Analysis Tools and Compensation Methods forthe Control of Machines with Friction " has carried out good summary to this.When Servo System Design, do not consider or only consider simple friction phenomenon, as stiction or Coulomb friction force, will cause tracking accuracy to descend, controlled variable is adjusted problems such as difficulty, slow running stationarity difference.In order to describe the friction that exists in the servo-drive system better and the harm that it causes to be compensated, be necessary to study the test and the identification of accurate friction model and parameter thereof, promptly at first modeling is accurately carried out in friction, next accurately measures the friction force between surface in contact under the particular case, according to specific friction model, adopt parameters identification method to obtain the numerical value of each parameter in the model then.Up to the present, people have proposed nearly kind of friction model surplus in the of 30, model wherein with strongest influence power is that Canudas De Wit C etc. is published in the LuGre model that proposes in the paper " A New Model forControl of Systems with Friction " of 40 the 3rd phases of volume of " IEEE Transaction on Automatic Control " nineteen ninety-five, and it can describe the situation of change of surface of contact from static to each state friction force of motion all sidedly.Following formula is the mathematic(al) representation of LuGre model, and wherein T is surface in contact normal pressure one friction force or a moment of friction regularly, and z is the average deformation amount of surface in contact bristle, ω _lBe relative motion angular velocity; Stiffness coefficient σ ₀, viscous damping coefficient σ ₁, Coulomb moment of friction F _C, static friction torque F _S, critical Stribeck angular velocity omega _S, the viscous friction factor sigma ₂Be 6 friction parameters.

$\left\{\begin{array}{c}\frac{\mathrm{dz}}{\mathrm{dt}}={\mathrm{\ω}}_l-{\mathrm{\σ}}_0\frac{\left|{\mathrm{\ω}}_l\right|}{g\left({\mathrm{\ω}}_l\right)}z\\ {\mathrm{\σ}}_{0}g\left({\mathrm{\ω}}_l\right)=F_C+(F_S-F_C)e^{-{({\mathrm{\ω}}_l/{\mathrm{\ω}}_s)}^2}\\ T={\mathrm{\σ}}_{0}z+{\mathrm{\σ}}_1\frac{\mathrm{dz}}{\mathrm{dt}}+{\mathrm{\σ}}_2{\mathrm{\ω}}_l\end{array}\right.---\left(1\right)$

In order in engineering, to use this model, just need measure the friction between the actual contact surface, yet the average deformation amount z of bristle can't measure, 6 model parameter F _C, F _S, ω _s, σ ₀, σ ₁, σ ₂Also be highly to be coupled, therefore can not have only the optimum estimate of adopting parameters identification method just can obtain these 6 parameters by the value of simple experiment acquisition relevant parameter.

Domestic and international existing friction parameter testing apparatus has at present:

(1) the FT friction testing system of Britain RAY-RAN company

This test macro can be measured coefficient of static friction and the kinetic friction coefficient between surface in contact automatically, and can carry out data storage, demonstration, and this coefficient of static friction is equivalent to the Fs of LuGre model, and this kinetic friction coefficient is equivalent to the Fc in the LuGre model, σ ₂Because the model that uses is simple, both can't embody the Stribeck effect in the measurement data, these dynamic friction effects of stiction that can not describe the friction memory again, increase.

(2) the MHK-500 friction wear testing machine of Jinan experimental machine factory

The method of testing machine employing manual loading, manual shift counterweight and rider is measured the friction factor between surface of contact, be Fs and the Fc in the LuGre model, can't measuring speed and friction force between relation, energy measurement dynamic friction parameter more not is as the σ in the LuGre model ₀, σ ₁Existing at present on its basis by increasing the research report that the electronic installation realization is measured automatically, be published in " sealing is with lubricated " 2007 as Yu Jianwei, 32 (2) paper: " development of MHK-500 ring piece friction wear testing machine intelligent observing and controlling system ", but still can not test the dynamic friction parameter.

The domestic and international at present measuring method at the LuGre friction parameter mainly contains two step identification methods, this method is at first proposed by the paper " Adaptive friction compensation with partially known dynamic frictionmodel " that C.Candus etc. is published in " Internatinal Journal of Adaptive Control and Signal Processing " 1997 the 11st volumes at first, its thinking is that the LuGre Model parameter is divided into two classes, i.e. 4 static parameter F _C, F _S, ω _S, σ ₂With 2 dynamic parameter σ ₀, σ ₁Suppose that when uniform motion the average deformation amount z of bristle does not change, promptly $\stackrel{\·}{z}=0,$ Above-mentioned like this LuGre model just can be reduced to

$T=[F_C+(F_S-F_C)e^{-{\left(\frac{{\mathrm{\ω}}_l}{{\mathrm{\ω}}_S}\right)}^2}]\mathrm{sgn}\left({\mathrm{\ω}}_l\right)+{\mathrm{\σ}}_2{\mathrm{\ω}}_2$

Be about to dynamic LuGre model and become static Stribeck model.So the identification step is:

The first step: adopt at the uniform velocity control method, measure speed of related movement or angular velocity and corresponding friction force or moment of friction respectively, just can use the method for curve fitting to determine 4 static parameter F _C, F _S, ω _S, σ ₂

Second step: under the prerequisite of determining static parameter, identification dynamic parameter σ ₀, σ ₁

For dynamic parameter, following several discrimination method is arranged again at present:

(1) linearization technique: give less power u of systemic effect, make it pre-slip micro-corner θ occur, directly utilize the corner of measuring, by formula ${\widehat{\mathrm{\σ}}}_0=\frac{u}{\mathrm{\θ}}$ Pick out stiffness coefficient σ ₀

Friction model when stagnating sliding movement can get in the linearization of equilibrium point place:

$\frac{\mathrm{dz}}{\mathrm{dt}}=\stackrel{\·}{\mathrm{\θ}}$

$J\stackrel{\·\·}{\mathrm{\θ}}+({\mathrm{\σ}}_1+{\mathrm{\σ}}_2)\stackrel{\·}{\mathrm{\θ}}{+\mathrm{\σ}}_0\mathrm{\θ}=u$

Wherein, J is the motion parts moment of inertia.

The method is a second-order system with system simplification, makes it be in the critical damping state, utilizes expression formula ${\mathrm{\σ}}_1=2\mathrm{\ξ}\sqrt{{\mathrm{\σ}}_{0}m}-{\mathrm{\σ}}_2$ Obtain σ ₁For example C.Candus etc. is published in the paper of " International Journall of AdaptiveControl and Signal Processing " 1997 the 11st volumes " Adaptive friction compensationwith partially known dynamic friction model just makes in this way.

Because the identification result of this linearization technique depends on choosing of parameter initial value, so identification precision and convergence cannot say for sure card, and the unusual difficulty of the measurement of the micro-displacement that slides in advance.

(2) optimization method: drive load and make it carry out the low speed periods motion, with site error as optimization aim, two dynamic parameter σ ₀, σ ₁Be optimized calculating as design variable, the final σ that obtains satisfaction ₀, σ ₁Value.What be published in that the paper " Research on Parameter Identification of FrictionModel for Servo System Based on Genetic Algorithms " of " Proceedings of the Fourth International Conference onMachine Learning and Cybernetics " adopts in 2005 as DE-PENG LIU is exactly this method.

(3) frequency domain identification method: adopt the method for local linearization that original system is reduced to second-order linear system, adopt frequency domain method identification dynamic parameter.Make system be in pre-sliding phase, import any Noise Excitation, just can measure the frequency domain response function to system.Frequencies omega is near in 0 the zone, according to | G (j0) |=C _m/ σ ₀, directly pick out stiffness coefficient σ ₀, according to the frequency response function of measuring, estimate the coefficient of second-order linear system, can pick out σ ₁+ σ ₂What for example people such as RonH.A.Hensen was published in that the papers " Frequency Domain Identification of Dynamic Friction ModelParameters " of " IEEE Transactions on Control systems Technology " 2002 the 10th the 2nd phases of volume describes is exactly this method.

All there is following deficiency in above-mentioned discrimination method:

1. experiment obtains static parameter because these methods all will adopt at the uniform velocity, and at the uniform velocity experiment wants pointwise to carry out, so workload is big, consuming time, also requires system not produce creeping phenomenon simultaneously when slow running, is difficult for realizing.In fact one of purpose of testing friction is exactly to obtain good slow running stationarity in servo-drive system, will reach this purpose when unknown friction parameter, and difficulty is very big.

2. must cause its accuracy to descend because linearization technique and frequency domain method have all carried out approximate processing to master mould, thereby for the measurement of dynamic parameter.

Therefore 3. because dynamic parameter and static parameter itself in the former LuGre model are coupled, by the hypothesis of " uniform motion does not change to the average deformation amount z of bristle ", system decoupling there is not theoretical foundation yet.

In sum, both do not had at present simple and direct, the method for identification LuGre friction parameter effectively of a kind of energy, not can be used for obtaining the instrument and equipment of related data yet.

The content of invention

The objective of the invention is to avoid above-mentioned the deficiencies in the prior art, a kind of system and method for measuring the dynamic friction parameter is provided,, improve the performance of servo-drive system to realize high-acruracy survey easily to dynamic friction parameter between different materials.

The technical thought that realizes the object of the invention is: adopt high precision moment sensor and high precision absolute encoder design friction torque test device, adopt adjustable variable period sinusoidal drive mode dead time, the dynamic and static characteristic that rubs between surface of contact is demonstrated fully, and be aided with real-time measurement, storage that controller is realized moment of friction, rotating speed, corner and driving moment; Test data is carried out automatic analyzing and processing, thus the disposable identifier that obtains 6 whole parameters.

Test macro of the present invention comprises: test specimen and counterweight stationary installation, motor, driving shaft, driving shaft, latch mechanism and controller, wherein, be fixed with absolute encoder below the motor, and be used to detect the relative position and the speed on test specimen surface; Be coaxially installed with torque sensor by stationary shaft between test specimen and counterweight stationary installation and the latch mechanism, be used to detect moment of friction.

Described test specimen and counterweight stationary installation are divided two parts up and down, and bottom and driving shaft are fixed as one, and top and torque sensor are connected; The surface of this bottom is provided with foursquare groove, is used to install down test block, and top is equipped with test block.

Described motor adopts torque motor, and with the coaxial bottom that is installed in driving shaft of absolute encoder, directly drive load operation.

Be provided with automatic test and the automatic identification program of parameter in the described controller.This controller is connected with absolute encoder, torque sensor and torque motor respectively, constitutes close loop control circuit, with relative velocity, measurement real time position, speed and the moment of friction on control test block surface.

The middle part of described driving shaft is provided with two layers of horizontal stand, and this support and frame are fixed as one.

Method of testing of the present invention adopts a step identification method, and its know-why is:

When each estimated parameter connects when approaching true value, the error minimum of the sum of errors alliance response of moment of friction so just can be converted into the parameter identification problem and find the solution following optimization problem:

find σ ₀、σ ₁，F _C、F _S、ω _S、σ ₂

$\min D={\mathrm{\λ}}_1\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{[{\widehat{\mathrm{\θ}}}_i-{\mathrm{\θ}}_i]}^2+{\mathrm{\λ}}_2\max |{\widehat{\mathrm{\θ}}}_i-{\mathrm{\θ}}_i|+{\mathrm{\λ}}_3\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{[{\hat{T}}_{\mathrm{fi}}-T_{\mathrm{fi}}]}^2+{\mathrm{\λ}}_4\max |{\hat{T}}_{\mathrm{fi}}-T_{\mathrm{fi}}|,$

$s.t.J\stackrel{\·\·}{\mathrm{\θ}}=T_u-T_f$

$\left\{\begin{array}{c}\frac{\mathrm{dz}}{\mathrm{dt}}={\mathrm{\ω}}_l-{\mathrm{\σ}}_0\frac{\left|{\mathrm{\ω}}_l\right|}{g\left({\mathrm{\ω}}_l\right)}z\\ {\mathrm{\σ}}_{0}g\left({\mathrm{\ω}}_l\right)=F_C+(F_S-F_C)e^{-{({\mathrm{\ω}}_l/{\mathrm{\ω}}_s)}^2}\\ T_f={\mathrm{\σ}}_{0}z+{\mathrm{\σ}}_1\frac{\mathrm{dz}}{\mathrm{dt}}+{\mathrm{\σ}}_2{\mathrm{\ω}}_l\end{array}\right.$

Wherein J is the moment of inertia of system, and Tu is a driving moment,

θ _i,

And T _FiBe respectively corner discreet value, actual measurement corner, the discreet value of moment of friction and the measured value of moment of friction of i sampled point, λ ₁For characterizing the weights that the accumulative total angular errors requires, its span is 0～1, λ ₂For characterizing the weights of hard-over error requirements, its span is 0～1, λ ₃For characterizing the weights of accumulative total moment of friction error requirements, its span is 0～1, λ ₄Be the weights for sign maximum friction moment error requirements, its span is 0～1.

According to above-mentioned principle method of testing of the present invention, comprise following process:

(1) installation testing spare and test parameter is set;

(2) the start-up control device is according to the test parameter control moment motor-driven load running that is provided with, and synchronous recording sampling instant, moment of friction, rotating speed, corner and motor windings current parameters;

(3) set current friction parameter, i.e. stiffness coefficient σ ₀, viscous damping coefficient σ ₁, Coulomb moment of friction F _C, static friction torque F _S, critical Stribeck angular velocity omega _S, the viscous friction factor sigma ₂Corresponding estimated value And according to described motor windings electric current by the solving system differential equation of motion, obtain each torque motor corner estimated value constantly

(4) calculate each the torque motor corner estimated value constantly and the deviation accumulation value E of the corner measured value in this moment of record _A, and find out maximum torque motor angular errors value E _MA

(5) calculate each the moment of friction estimated value constantly and the deviation accumulation value E of the moment of friction measured value in this moment of record _T, and find out maximum moment of friction error value E _MT

(6) whether determining step (4), (5) resulting parameter satisfy the test parameter that step (1) sets, if satisfy then export current friction parameter

If do not satisfy by genetic algorithm correction friction parameter Repeating step (3) finishes until satisfying end condition.

The present invention compared with prior art has following advantage:

1). because test macro of the present invention is provided with controller, can realize the automatic test of measured material and the automatic identification of parameter; Because system of the present invention is provided with absolute encoder, can effectively detect the relative rotation and the angular velocity on measured material surface simultaneously; Because system of the present invention is equipped with torque sensor, can realize moment of friction is carried out high-acruracy survey in addition.

2) because method of testing of the present invention adopts a step identification method, need not at the uniform velocity to test, thereby test is simple, and an identification of all friction parameters is finished; Solved the problem of supposing not have theoretical foundation by " uniform motion does not change the average deformation amount z of bristle " simultaneously.

Simulation result shows, adopts dynamic friction testing method of the present invention, and the average error of 6 friction parameters is 0.601%.

Description of drawings

Fig. 1 is the structural representation of test macro of the present invention;

Fig. 2 is a controller principle block diagram of the present invention;

Fig. 3 is that the present invention distinguished knowledge method test flow chart in one step;

Fig. 4 is the The simulation experiment result figure of the present invention's one step discrimination method, and wherein (a) is the identifier presentation graphs of 1000 each friction parameters of iteration, (b) is the identifier presentation graphs of 50000 each friction parameters of iteration.

Below in conjunction with accompanying drawing the present invention is described in further detail.

Embodiment

With reference to Fig. 1, test macro of the present invention comprises: the physical construction part is made up of handle 1, latch mechanism 2, torque sensor 4, test block and counterweight stationary installation 6, test specimen to be measured 7, driving shaft 8, frame 10, horizontal stand 11, torque motor 13, absolute encoder 14 and controller 16.This test block and counterweight stationary installation 6 are divided into two-layer up and down, the surface of the 6a of lower floor is provided with square indentations, upper strata 6b is a multidiameter, be fixed with lower flange 4a and 4b on the torque sensor, frame 10 is a cast-in-block rectangular parallelepiped framed structure, is convenient to operation, and the centre is provided with two layers of horizontal stand 11 to strengthen rigidity, driving shaft 8 is installed on the horizontal stand 11 by bearing 9, can be around its axis rotation.Absolute encoder 14 is fixing with torque motor 13, and be coaxially installed on the lower end of driving shaft 8 by shaft joint 12, the 6a of lower floor of the upper end of this driving shaft and test block and counterweight stationary installation fixes, the upper strata 6b of test block and counterweight stationary installation is connected with torque sensor 4 by lower flange 4a, counterweight 5 is placed on the both sides of this upper strata 6b, be test block imposed load up and down, torque sensor 4 is connected with stationary shaft 3 lower ends by upper flange plate, stationary shaft 3 is connected with frame 10 by latch mechanism 2, and the upper end of stationary shaft 3 and handle 1 are connected.Test block 7 places the levels of test block and counterweight stationary installation respectively, wherein descends test block 7a to place in the groove of the 6a of lower floor of test block and counterweight stationary installation, and last test block 7 is fixed on the 6b surface, upper strata of test block and counterweight stationary installation by screw 15.Controller 16 is connected with torque motor 13, absolute encoder 14 and torque sensor 4 respectively by cable, constitutes close loop control circuit, with relative velocity, measurement real time position, speed and the moment of friction on control test block surface.

With reference to Fig. 2, controller 16 is a core with the digital signal processor DSP, is made up of user interface, Electric Machine Control part and moment of friction test section, has functions such as motion control, data acquisition, information processing.

(1)DSP

Described DSP is used to receive detecting information that the user is provided with, test block is tested automatically and the automatic identification of parameter, test result sent to PC, be solidified with automatic test and the automatic identification program of parameter among this DSP, and expand the static RAM (SRAM) that 64KB is arranged, with store test data.The DSP of this example adopts the TMS320LF2407A model, but is not limited thereto model.

(2) user interface

Described user interface mainly is made up of 320 * 240 dot matrix LCD display, keyboard and USB2.0 interface circuit.

The general parallel I of 320 * 240 dot matrix LCD display/O port and DSP are by linking to each other, and DSP can send to LCD display by parallel I/O mouth with control command and character information, demonstrates information, duty and the parameter identification result of setting.

Keyboard adopts 16 bond moment configuration keyboards, but is not limited to 16 keys, links to each other with the universaling I/O port of DSP by 8 data lines, and DSP obtains key information by the state that scans corresponding I/O mouth.Parameter by keyboard input comprises: operational speed range, the cycle of operation, identification precision, dead time scope.

The USB2.0 interface is made of interface controller PHILIPS ISP1581, links to each other with DSP by system bus, simultaneously adopts the prior USB cable to be connected between USB2.0 interface controller and the PC.Measure after identification finishes, can will measure and identification result sends on the PC and stores.

(3) Electric Machine Control part

Described Electric Machine Control part is made up of D/A converter, servoamplifier, motor windings current detection circuit, absolute type encoder.

D/A converter adopts 12 high-speed d/a chip MX7541, slewing rate 0.6 μ s, and its digital input end links to each other with DSP by system bus, and its analog output links to each other with the analog input end of servoamplifier.

The analog input end of servoamplifier links to each other with the output terminal of D/A converter, and output terminal is connected with torque motor.

The winding current testing circuit is made up of precision resistance and A/D converter, precision resistance is series in the motor-driven loop, the pressure drop at its two ends has reflected the size of electric current in the motor windings, adopt 12 A/D converter MAX120 at a high speed, slewing rate 1.6 μ s are converted to this voltage signal digital quantity and send DSP to by system bus.

Absolute type encoder links to each other with DSP by the RS485 interface, and this absolute type encoder is used to detect the corner of torque motor.

This Electric Machine Control part can constitute electric current loop, speed ring and position ring three closed loop position control loops, also can work in open loop approach.When carrying out the friction parameter test, the work of use open loop approach, DSP provides the control signal that is sinusoidal variations according to the operational speed range of input, send servoamplifier to produce corresponding driving voltage after the D/A conversion, and drive motor drives load and follows the reference signal motion.

(4) moment of friction test section

Described moment of friction test section is by high-precision locked rotor torque sensor Staiger Mohilo 0510RD, range 10Nm, and precision ± 0.02Nm and 12 high-speed a/d converter MAX120 constitute.

The simulating signal of locked rotor torque sensor output is stored among the RAM of expansion after the A/D conversion.A/D converter MAX120 all links to each other with DSP by system bus with static RAM (SRAM).

During test, torque motor 13 passes through transmission shaft 8 and drives test block 7a running down, and it is static that last test block 7b keeps, the moment of friction about just can measuring by torque sensor 4 and controller between test block 7b and 7a.

With reference to Fig. 3, the present invention comprises mainly that to the measurement of dynamic friction parameter test is prepared, test and three processes of the automatic identification of parameter automatically, and concrete steps are as follows:

The first step: test is prepared, and comprises installation testing spare and test parameter is set.

1) installation testing spare: at first unclamp latch mechanism, upwards mention handle, make the friction testing emaciated face from, set lever makes its unlikely whereabouts again; Secondly, test block is installed in respectively up and down on two measurement faces; At last, unclamp latch mechanism 2, making up and down, the test specimen surface contacts fully;

2) set interface pressure:, select suitable counterweight to load set lever according to different operating modes;

3) parameter setting: by keyboard input operational speed range, parameter identification precision, test period number, dead time range parameter.

Second step: test automatically

Press " test " button on the lower keyboard, torque motor starts under the effect of DSP control circuit, its rotating speed in the range of operation that requires by sinusoidal variation, simultaneously with the moment of friction in every 1ms sampling period, rotating speed, corner, these reference records of motor windings electric current in static RAM (SRAM), after finishing the test period of setting, end of test (EOT).

The 3rd step: the automatic identification of parameter

1) sets current friction parameter

Promptly set 6 friction parameters, i.e. stiffness coefficient σ ₀, viscous damping coefficient σ ₁, Coulomb moment of friction F _C, static friction torque F _S, critical Stribeck angular velocity omega _S, the viscous friction factor sigma ₂Corresponding estimated value

2) estimated value of factored moment motor corner

The estimated value of this corner can be found the solution acquisition to following differential equation of motion by the quadravalence runge kutta method

$J\stackrel{\·\·}{\widehat{\mathrm{\θ}}}\left(t\right)K_{a}I\left(t\right)-{\hat{T}}_f\left(t\right)---\left(2\right)$

Wherein, For the estimated value of t moment moment of friction, according to the estimated value of current each parameter

Mathematic(al) representation (1) by the LuGre model can calculate; J is known system's moment of inertia, and the bottom that contains rotor, transmission shaft and test specimen, counterweight stationary installation reaches test block down; K _aBe known motor torque constant; I (t) the is the t measured value of the electric current of motor windings constantly, and I (t) derives from the experimental data of i sampled point of record, establishes t _pBe the sampling period, then t=it _p

Be t corner constantly.

3) calculating accumulative total corner valuation error is also selected hard-over valuation error

If experimentation obtains n group sampled data altogether,

And θ _iBe respectively the discreet value and the measured value of i sampled point corner, wherein ${\widehat{\mathrm{\θ}}}_i=\widehat{\mathrm{\θ}}(i\·t_p),$

The accumulative total angular errors calculates according to following formula:

$E_A=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{[{\widehat{\mathrm{\θ}}}_i-{\mathrm{\θ}}_i]}^2$

Hard-over valuation error is selected according to following formula:

$E_{\mathrm{MA}}=\max |{\widehat{\mathrm{\θ}}}_i-{\widehat{\mathrm{\θ}}}_i|.$

4) calculating accumulative total moment of friction valuation error is also selected maximum friction moment valuation error

If

And T _FiBe respectively the discreet value and the measured value of i sampled point moment of friction, wherein ${\hat{T}}_{\mathrm{fi}}={\hat{T}}_{\mathrm{fi}}(i\·t_p),$ Accumulative total moment of friction valuation error is calculated according to following formula:

$E_T=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{[{\hat{T}}_{\mathrm{fi}}-T_{\mathrm{fi}}]}^2$

Maximum friction moment valuation error is selected according to following formula:

$E_{\mathrm{MT}}=\max |{\hat{T}}_{\mathrm{fi}}-T_{\mathrm{fi}}|.$

5) calculating target function, judge whether it reaches given estimation accuracy objective function and be calculated as follows:

J＝λ ₁E _A+λ ₂E _MA+λ ₃E _T+λ ₄E _MT

Wherein, λ ₁For characterizing the weights that the accumulative total angular errors requires, its span is 0～1; λ ₂For characterizing the weights of hard-over error requirements, its span is 0～1; λ ₃For characterizing the weights of accumulative total moment of friction error requirements, its span is 0～1; λ ₄For characterizing the weights of maximum friction moment error requirements, its span is 0～1.

If the friction parameter valuation reaches given precision, then on LCD display, export the result, and test data of experiment and identification result sent on the PC by USB interface store.

As do not meet the requirements of precision, then according to the estimated value of each friction parameter of genetic algorithm correction

Return step 1), until satisfying accuracy requirement.

Effect of the present invention can further specify by following emulated data:

Simulated conditions: get J=1Kgm ², given one group of friction parameter is exported as identification with optimum individual as the actual value of parameter, adopts genetic algorithm to calculate the identifier in 100 generations, 1000 generations and 50000 generations.

Simulation result: shown in Fig. 4 and table 1, in the time of can finding out that by Fig. 4 (a) genetic algorithm moved for 1000 generations, the optimum individual value of each parameter is that identifier has leveled off to its true value, and in the time of can finding out that by Fig. 4 (b) genetic algorithm moved for 50000 generations, the identifier of each parameter has satisfied error requirements.The average identification error of 6 parameters when genetic algorithm moved for 1000 generations as shown in Table 1 is 2.612%, and maximum identification error is 9.4%; When genetic algorithm moved for 50000 generations, parameter σ ₀, σ ₁, F _S, F _C, ω _S, σ ₂The identification error be respectively 0.000%, 0.030%, 0.596%, 0.007%, 0.030%, 2.940%, the maximum identification error of these 6 parameters is 2.940%, average identification error is 0.601%, thereby the validity of the inventive method has been described.

Table 1

Parameter	Actual value	Identifier (100 generation)	Identifier (1000 generation)	Identifier (50000 generation)
					σ 0	0.5	0.48721	0.50000	0.50000
σ 1	0.4	0.4073	0.39980	0.40012
					F S	2.4	2.55469	2.62620	2.38570
F C	0.6	0.73354	0.59981	0.60004
					ω S	0.01	0.019	0.009773	0.010003
σ 2	0.005	0.00539	0.0051964	0.004853

Claims (10)

1. dynamic friction parameter test system, comprise test block and counterweight stationary installation, motor, driving shaft, latch mechanism and controller, it is characterized in that, be fixed with absolute encoder (14) below the motor (13), be used to measure the relative rotation and the angular velocity on test block surface; Be coaxially installed with torque sensor (4) by stationary shaft (3) between test block and counterweight stationary installation (6) and the latch mechanism (2), be used to detect moment of friction.

2. test macro according to claim 1 is characterized in that, test block and counterweight stationary installation (6) are divided two-layer (6b and 6a) up and down, and lower floor (6a) is fixed as one with driving shaft (8), and upper strata (6b) is connected with torque sensor (4); The surface of this lower floor is provided with foursquare groove, is used to install down test block (7a), and the upper strata is equipped with test block (7b).

3. test macro according to claim 1 is characterized in that, motor (13) adopts torque motor, and with the coaxial bottom that is installed in driving shaft (8) of absolute encoder (14), directly drive load operation.

4. test macro according to claim 1 is characterized in that, is provided with automatic test and the automatic identification program of parameter in the controller (16).

5. according to claim 1 or 4 described test macros, it is characterized in that, controller (16) is connected with absolute encoder (14), torque sensor (4) and torque motor (13) respectively, constitute close loop control circuit, with the relative rotation speed on control test block surface, measure real-time corner, rotating speed and moment of friction.

6. test macro according to claim 1 is characterized in that, the middle part of driving shaft (8) is provided with two layers of horizontal stand (11), and this support and frame (10) are fixed as one.

7. dynamic friction parameter test method comprises following process:

(1) installation testing spare and test parameter is set;

(2) the start-up control device is according to the test parameter control moment motor-driven load running that is provided with, and synchronous recording sampling instant, moment of friction, rotating speed, corner and motor windings current parameters;

(3) set current friction parameter, i.e. stiffness coefficient σ ₀, viscous damping coefficient σ ₁, Coulomb moment of friction F _C, static friction torque F _S, critical Stribeck angular velocity omega _S, the viscous friction factor sigma ₂Corresponding estimated value

And according to described motor windings electric current by the solving system differential equation of motion, obtain each torque motor corner estimated value constantly

(4) calculate each the torque motor corner estimated value constantly and the deviation accumulation value E of the corner measured value in this moment of record _A, and find out maximum torque motor angular errors value E _MA

(5) calculate each the moment of friction estimated value constantly and the deviation accumulation value E of the moment of friction measured value in this moment of record _T, and find out maximum moment of friction error value E _MT

(6) whether determining step (4), (5) resulting parameter satisfy the test parameter that step (1) sets, if satisfy then export current friction parameter

If do not satisfy by genetic algorithm correction friction parameter

Repeating step (3) finishes until satisfying end condition.

8. method of testing according to claim 7 is characterized in that the described test parameter that is provided with of step (1), comprises operational speed range, the cycle of operation, the dead time scope and friction parameter identification precision of torque motor.

9. method of testing according to claim 7 is characterized in that the described system motion differential equation of step (3) is:

$J\stackrel{\·\·}{\widehat{\mathrm{\θ}}}\left(t\right)=K_{a}I\left(t\right)-{\hat{T}}_f\left(t\right)$

In the formula,

Be the estimated value of t moment moment of friction, J is known system's moment of inertia, and I (t) is the t measured value of the electric current of motor windings constantly, K _aBe known motor torque constant,

Estimated value for t corner constantly.

10. method of testing according to claim 7 is characterized in that, whether the described determining step of step (6) (4), (5) resulting parameter satisfy the test parameter that step (1) sets, and carry out according to the following procedure:

(6a). by following formula step (4), (5) resulting parameter are sued for peace:

J＝λ ₁E _A+λ ₂E _MA+λ ₃E _T+λ ₄E _MT

In the formula, λ ₁For characterizing the weights that the accumulative total angular errors requires, its span is 0～1, λ ₂For characterizing the weights of hard-over error requirements, its span is 0～1, λ ₃For characterizing the weights of accumulative total moment of friction error requirements, its span is 0～1, λ ₄Be the weights for sign maximum friction moment error requirements, its span is 0～1;

(6b). use the value J that sued for peace and the identification precision parameter of setting to compare, less than this set the identification precision parameter as identifier output, otherwise revise again.

Images