Disturbance observation method for six-axis vibration table, and system and disturbance rejection control method

Abstract

Disclosed in the present invention is a disturbance observation method for a six-axis vibration table. The method comprises: acquiring data information of a target six-axis vibration table; constructing an extended state observer for the target six-axis vibration table; constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer for the target six-axis vibration table, setting start-stop rules, and performing training; and using a trained fuzzy neural disturbance observer to complete disturbance observation for the target six-axis vibration table. Further disclosed in the present invention are a system implementing the disturbance observation method for a six-axis vibration table, and a disturbance rejection control method comprising the disturbance observation method for a six-axis vibration table. In the present invention, by means of the design and implementation of an extended state observer, a fuzzy neural network identifier and a fuzzy neural disturbance observer, not only disturbance observation and disturbance control of a system are realized, thus improving the tracking performance of the system, but also higher reliability and higher accuracy are realized.

Description

Disturbance observation method, system and anti-disturbance control method of six-axis vibration table TECHNICAL FIELD

[0001] The present application belongs to the field of motion real-time control, and particularly relates to a disturbance observation method, system and anti-disturbance control method of a six-axis vibration table. BACKGROUND

[0002] Today, high-speed trains are developing continuously, and the research on the train-bridge coupling problem is increasingly important. In particular, under extreme conditions such as earthquakes, the interaction between trains and bridge structures has gradually become the focus of researchers. In order to improve the seismic performance of the structure and explore its solutions, real-time hybrid testing (RTHS) has emerged. The RTHS technology not only can simulate the vibration response of the train on the bridge, but also can provide real-time simulation of the interaction between the train and the track in a complex dynamic system, which has important guiding significance for improving the seismic performance of the structure under the action of train-bridge coupling.

[0003] In RTHS, the test object is usually divided into numerical substructures and test substructures. The numerical substructure usually includes parts that are easy to analyze numerically but difficult to test, such as track and bridge models. The test substructure is usually a train model. Taking a train-bridge coupling model as an example, RTHS measures the displacement, velocity, force and other information of the train model in real time, substitutes it into the numerical bridge model to calculate the expected response of the vibration table at the next time, and then gives the train model excitation through the vibration table in real time, thereby forming a closed loop. The stability and reliability of real-time hybrid testing are greatly affected by whether the vibration table can accurately and precisely track the input signal. Compared with a single-degree-of-freedom vibration table, a six-degree-of-freedom vibration table can simulate the multi-degree-of-freedom response of train motion and is suitable for a more realistic train motion model.

[0004] At present, the traditional control scheme for the six-degree-of-freedom vibration table does not fully consider the uncertainty and disturbance of the six-degree-of-freedom vibration table in the control process, and does not carry out corresponding control analysis and response, which makes the existing scheme still have the problems of poor control reliability and low control accuracy in the control process.

[0005] SUMMARY

[0006] One of the purposes of the present application is to provide a disturbance observation method for a six-axis vibration table with high reliability and good accuracy.

[0007] The second purpose of the present application is to provide a system for implementing the disturbance observation method of the six-axis vibration table.

[0008] The third purpose of the present application is to provide an anti-disturbance control method including the disturbance observation method of the six-axis vibration table.

[0009] The application provides the disturbance observation method of the six-axis vibration table, and the method comprises the following steps:

[0010] S1. obtaining data information of a target six-axis vibration table;

[0011] S2. regarding nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances, constructing an extended state observer of the target six-axis vibration table based on the data information obtained in the step S1, and realizing preliminary suppression of the total disturbances based on an output result of the extended state observer;

[0012] S3. constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the extended state observer and the fuzzy neural network constructed in the step S2, and setting start-stop rules of the fuzzy neural network identifier and the fuzzy neural disturbance observer;

[0013] S4. training the fuzzy neural network identifier and the fuzzy neural disturbance observer according to real-time data of the extended state observer of the target six-axis vibration table;

[0014] S5. completing disturbance observation of the target six-axis vibration table by using the trained fuzzy neural disturbance observer.

[0015] The step S2 of regarding the nonlinear factors, the uncertain factors and the external disturbances of the target six-axis vibration table as the total disturbances comprises the following steps:

[0016] The nonlinear factors, the uncertain factors and the external disturbances of the target six-axis vibration table are regarded as the total disturbances, so that a transfer function is determined as Y(s)=G'(s)(U(s)+g(s)), wherein Y(s) is an output of the system; G'(s) is an actual system model including the total disturbances, and G'(s)=G(s)(1+Δ(s)), G(s) is a numerical model of the system, and Δ(s) is a disturbance amount; U(s) is a control input of the system; and g(s) is uncertainty of the system;

[0017] The transfer function is converted into Y(s)=G(s)(U(s)+g'(s)), wherein g'(s) is the total disturbance of the system;

[0018] If g'(s) can be completely eliminated, the transfer function can be simplified as a nominal model.

[0019] The step S2 of constructing the extended state observer of the target six-axis vibration table comprises the following steps:

[0020] An extended state observer is designed based on a principle of active disturbance rejection, and is used for observing the total disturbance g'(s) of the system;

[0021] total disturbance of the system z f , the active disturbance rejection control law is designed as u = u0 - z f , where u0 is the input without disturbance rejection.

[0022] The fuzzy neural network identifier and the fuzzy neural disturbance observer of the target six-axis vibration table are constructed according to the following steps.

[0023] The fuzzy neural network identifier of the target six-axis vibration table is constructed based on the fuzzy neural network, and is used to fit the extended state observer.

[0024] The fuzzy neural disturbance observer of the target six-axis vibration table is constructed based on the fuzzy neural network, and is a multi-input single-output fuzzy neural network, which is used to replace the extended state observer to observe the disturbance and suppress the disturbance through a negative feedback adjustment mechanism after the disturbance is observed.

[0025] The parameters of the fuzzy neural disturbance observer are obtained through the training of the fuzzy neural network identifier.

[0026] The fuzzy neural network identifier and the fuzzy neural disturbance observer of the target six-axis vibration table are constructed according to the following steps.

[0027] The fuzzy neural network identifier of the target six-axis vibration table comprises an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer connected in sequence. The inputs of the input layer include the difference e d between the output value and the actual value of the extended state observer, the differential value d of the difference e and the control signal u. The output of the network output layer is the increment Δf z of the output of the fitted extended state observer. The processing process of the network output layer is represented as where w kiii is the output weight, is the normalized output of the lth fuzzy rule, and n represents the number of fuzzy rules.

[0028] The fuzzy neural disturbance observer of the target six-axis vibration table comprises an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer connected in sequence. The output of the network output layer is the observed increment Δf o of the total disturbance of the system. The processing process of the network output layer is represented as where w kmmm is the output weight, is the normalized output of the lth fuzzy rule.

[0029] Step S3, which involves setting the start and stop rules for the fuzzy neural network identifier and the fuzzy neural perturbation observer, specifically includes the following steps:

[0030] The difference e between the total perturbation output of the extended state observer and the output value of the fuzzy neural network identifier d Design the discriminant function f(e) d ); for the discriminant function f(e d The value of ) is used to determine the feedback gain K1 of the extended state observer and the feedback gain K2 of the fuzzy neural perturbation observer:

[0031] If f(e) d If α ≥ 0, then set K1 = 0 and K2 = 1. At this point, the output data of the extended state observer is used to train the fuzzy neural network identifier. After each training iteration, the difference e is calculated for the next round. d The judgment is made, and the input data in the fuzzy neural network identifier is used to train the network parameters and adjust them in the next round;

[0032] If f(e) d If α < 1, then set K1 = 1 and K2 = 0. At this time, the extended state observer is turned off, and the parameters in the fuzzy neural network identifier are passed to the fuzzy neural perturbation observer. The fuzzy neural perturbation observer is then used directly for perturbation observation. Subsequently, the fuzzy neural perturbation observer will undergo self-training.

[0033] The training described in step S4 specifically includes the following steps:

[0034] The following formula function is used as the loss function E1 of the fuzzy neural network identifier, and E1 is based on ESO, and training is performed on it: In the formula f I f is the output of the fuzzy neural network identifier. O This is the output of the extended state observer;

[0035] The following formula function is used as the loss function E of the fuzzy neural perturbation observer, and E is based on the nominal model and used for training: In the formula, y is the output of the nominal model; y1 is the output of the actual system.

[0036] The application further provides a system for realizing the disturbance observation method of the six-axis vibration table, comprising a data acquisition module, a first construction module, a second construction module, a training module and an observation module; the data acquisition module, the first construction module, the second construction module, the training module and the observation module are connected in sequence; the data acquisition module is used for acquiring data information of a target six-axis vibration table and uploading the data information to the first construction module; the first construction module is used for regarding nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances, constructing an extended state observer of the target six-axis vibration table according to the received data information, realizing preliminary suppression of the total disturbances based on an output result of the extended state observer, and uploading the data information to the second construction module; the second construction module is used for constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the constructed extended state observer and a fuzzy neural network according to the received data information, setting start-stop rules of the fuzzy neural network identifier and the fuzzy neural disturbance observer, and uploading the data information to the training module; the training module is used for training the fuzzy neural network identifier and the fuzzy neural disturbance observer according to real-time data of the extended state observer of the target six-axis vibration table according to the received data information, and uploading the data information to the control module; and the observation module is used for completing disturbance observation of the target six-axis vibration table by using the trained fuzzy neural disturbance observer according to the received data information.

[0037] The application further provides an anti-disturbance control method comprising the disturbance observation method of the six-axis vibration table, and further comprising the following steps:

[0038] S6. Real-time anti-disturbance control of the target six-axis vibration table is completed according to the disturbances of the target six-axis vibration table observed in real time in step S5.

[0039] The disturbance observation method, system and anti-disturbance control method of the six-axis vibration table provided by the application realize total disturbance observation and elimination of the system by the extended state observer, so that the system characteristics are consistent with the nominal model, then the fuzzy neural network identifier is trained by the output data of the extended state observer, when the system error is small enough, the trained fuzzy neural network identifier parameters are introduced into the fuzzy neural disturbance observer, the fuzzy neural disturbance observer is used to replace the extended state observer, and the fuzzy neural disturbance observer is further trained by the control error of the system, so as to realize disturbance observation and disturbance control of the system, and finally the tracking performance of the system is improved; therefore, the reliability of the application is higher, and the accuracy is better. BRIEF DESCRIPTION OF DRAWINGS

[0040] Fig. 1 is a method flow diagram of the observation method of the application.

[0041] Fig. 2 is a functional module diagram of the system of the application.

[0042] Figure 3 is a schematic diagram of the method flow of the control method of the present application. DETAILED DESCRIPTION

[0043] Figure 1 is a schematic diagram of the method flow of the method of the present application: the disturbance observation method of the six-axis vibration table disclosed in the present application comprises the following steps:

[0044] S1. Obtain the data information of the target six-axis vibration table;

[0045] S2. Consider the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances, construct an extended state observer of the target six-axis vibration table based on the data information obtained in step S1, and realize preliminary suppression of the total disturbances based on the output results of the extended state observer;

[0046] In the implementation, the consideration of the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances specifically comprises the following steps:

[0047] The nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table are considered as total disturbances, so that the transfer function is determined as:

[0048] Y(s) = G'(s)(U(s) + g(s))

[0049] In the formula, Y(s) is the output of the system; G'(s) is the actual system model including the total disturbances, and G'(s) = G(s)(1 + Δ(s)), G(s) is the numerical model of the system (directly established according to data or principles), and Δ(s) is the disturbance quantity (including the uncertainties and unmodeled factors in the model); U(s) is the control input of the system; and g(s) is the uncertainty of the system.

[0050] The transfer function is converted into the following form:

[0051] Y(s) = G(s)(U(s) + g'(s))

[0052] In the formula, g'(s) is the total disturbance of the system.

[0053] If g'(s) can be completely eliminated, the transfer function can be simplified as a nominal model; therefore, based on this idea, a three-order state space model containing disturbances is set, which is expressed as:

[0054] In the formula, x is a state vector, is the differential of the system state, x1 is displacement information, x2 is speed information, and x3 is acceleration information; A is a state matrix, and a1 is the first information of the model, a2 is the second information of the model, and a3 is the third information of the model; B is an input vector, and b0 is the fourth information of the model; u is an input signal; g(x1, x2, x3, h(t)) is a total disturbance, and there is a first derivative and a bound; h(t) is a disturbance information; y is an output signal; C is an output vector, and

[0055] The extended state is set as x f , and x f =g(x1, x2, x3, h(t)); define θ as a state vector of the extended system, and θ=[x1 x2 x3 x f ] T , then A, B and C are expressed in the following forms:

[0056] In the formula, is a differential of the state vector of the extended system; A m is a state matrix; B m is an input matrix; M is a disturbance matrix; C m is an output matrix;

[0057] The extended state observer for constructing the target six-axis vibration table specifically includes the following steps:

[0058] The extended state observer designed based on the active disturbance rejection principle is expressed as:

[0059] In the formula, z=[z1 z2 z3 z f ] T , z is a state vector, is a differential of the system state, z1 is a displacement signal, z2 is a speed signal, z3 is an acceleration signal, and z f is a disturbance signal; u is a control signal; is an estimated displacement of the extended state estimator; L is an observer gain matrix, and L=[l1 l2 l3 l4] T , l1 is a first observation gain, l2 is a second observation gain, l3 is a third observation gain, and l4 is a fourth observation gain; L can be adjusted by a bandwidth method; the extended state observer is used for observing the total disturbance g'(s) of the system;

[0060] Based on the observed total disturbance z f of the system, the active disturbance rejection control law is designed as:

[0061] u=u0-z f

[0062] In the formula, u0 is an input without disturbance suppression;

[0063] S3. Constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the expansion state observer and the fuzzy neural network constructed in step S2, and setting the start-stop rule of the fuzzy neural network identifier and the fuzzy neural disturbance observer;

[0064] In particular implementation, the construction of the fuzzy neural network identifier and the fuzzy neural disturbance observer of the target six-axis vibration table includes the following steps:

[0065] The fuzzy neural network identifier of the target six-axis vibration table is constructed based on the fuzzy neural network, and is used for fitting the expansion state observer;

[0066] The fuzzy neural disturbance observer of the target six-axis vibration table is constructed based on the fuzzy neural network, and is a multi-input single-output fuzzy neural network, which is used for disturbance observation instead of the expansion state observer, and inhibits the disturbance through a negative feedback adjustment mechanism after observing the disturbance;

[0067] The parameters of the fuzzy neural disturbance observer are obtained through the training of the fuzzy neural network identifier;

[0068] In particular implementation, the following steps can be adopted:

[0069] The fuzzy neural network identifier of the target six-axis vibration table includes an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer connected in sequence;

[0070] The inputs of the input layer include the difference e between the output value and the actual value of the expansion state observer d , the differential value d of the difference e and the control signal u;

[0071] The membership function layer adopts a Gaussian membership function as the membership function of the fuzzy neural network, and outputs the membership of the error, the membership of the differential value of the error and the membership of the control signal. The processing process of the membership function layer is represented as wherein is the output value of the i th input of the input layer at the j th membership function, x i is the i th input of the input layer, c ij is the center coefficient of the i th input of the input layer at the j th membership function, and σ ij is the width coefficient of the i th input of the input layer at the j th membership function;

[0072] The output of the fuzzy rule fitness layer is the cross product of the membership function of error, the membership function of error differential value and the membership function of control signal; the processing procedure of the fuzzy rule fitness layer is represented as wherein is the output of the lth fuzzy rule fitness in the fuzzy rule fitness layer, is the output value of the jth membership function of the first input in the input layer, is the output value of the jth membership function of the second input in the input layer, is the width coefficient of the jth membership function of the first input in the input layer, is the width coefficient of the jth membership function of the second input in the input layer, x1 is the first input in the input layer and x2 is the second input in the input layer, is the center coefficient of the jth membership function of the first input in the input layer, is the center coefficient of the jth membership function of the second input in the input layer;

[0073] The output of the normalization layer is the normalized value of the output of the fuzzy rule fitness layer; the processing procedure of the normalization layer is represented as wherein is the normalized value of the output of the lth fuzzy rule fitness in the normalization layer, is the output of the lth fuzzy rule fitness in the fuzzy rule fitness layer and n is the total number of the outputs of the fuzzy rule fitness layer;

[0074] The output of the network output layer is the increment Δf of the fitted extended state observer output z ; the processing procedure of the network output layer is represented as wherein w kiii is the output weight, is the normalized output of the lth fuzzy rule and n represents the number of fuzzy rules;

[0075] The fuzzy neural perturbation observer for the target six-axis vibration table comprises an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer connected in sequence;

[0076] The inputs of the input layer include the difference e d between the output value and the actual value of the extended state observer, the differential value d of the difference e and the control signal u;

[0077] The membership function layer adopts Gaussian membership function as the membership function of the fuzzy neural network, and the output is the membership of error, the membership of error differential value and the membership of control signal; the processing procedure of the membership function layer is represented as where is the output value of the nth membership function for the mth input of the input layer, x m is the mth input of the input layer, x mn is the center coefficient of the nth membership function for the mth input of the input layer, σ mn is the width coefficient of the nth membership function for the mth input of the input layer;

[0078] The output of the fuzzy rule fitness layer is the cross product of the membership functions of the error, the membership functions of the error differential value and the membership functions of the control signal; the processing procedure of the fuzzy rule fitness layer is represented as where is the output of the llth fuzzy rule fitness in the fuzzy rule fitness layer, is the output value of the jj 11 th membership function for the 11th input of the input layer, is the output value of the jj 22 th membership function for the 22nd input of the input layer, is the width coefficient of the jj 11 th membership function for the 11th input of the input layer, is the width coefficient of the vth membership function for the 22nd input of the input layer, x 11 is the 11th input of the input layer, x 22 is the 22nd input of the input layer, is the center coefficient of the jj 11 th membership function for the 11th input of the input layer, is the center coefficient of the jj 22 th membership function for the 22nd input of the input layer;

[0079] The output of the normalization layer is the normalized value of the output of the fuzzy rule fitness layer; the processing procedure of the normalization layer is represented as where is the normalized value of the output of the llth fuzzy rule fitness in the normalization layer, is the output of the llth fuzzy rule fitness in the fuzzy rule fitness layer, nn is the total number of the outputs of the fuzzy rule fitness layer;

[0080] The output of the network output layer is the total disturbance increment Δf o observed by the system; the processing procedure of the network output layer is represented as where w kmmm is the output weight, is the output of the normalization layer.

[0081] Further, the start-stop rule of setting the fuzzy neural network identifier and the fuzzy neural disturbance observer comprises the following steps.

[0082] The difference e between the total disturbance output by the extended state observer and the output value of the fuzzy neural network identifier d Design a discriminant function f(e d ); judge the value of the discriminant function f(e d ), and then set the feedback gain K1 of the extended state observer and the feedback gain K2 of the fuzzy neural disturbance observer:

[0083] If f(e d )≥α, set K1=0 and K2=1, at this time, the output data of the extended state observer is used to train the fuzzy neural network identifier, and after each training, the difference e d is judged for the next round, and the data is input into the fuzzy neural network identifier for training network parameters and adjustment in the next round;

[0084] If f(e d )<α, set K1=1 and K2=0, at this time, the extended state observer is closed, the parameters in the fuzzy neural network identifier are transmitted to the fuzzy neural disturbance observer, and the fuzzy neural disturbance observer is directly used for disturbance observation, and the fuzzy neural disturbance observer will be trained itself subsequently.

[0085] As a preferred solution, the discriminant function f(e d ) can be set as f(e d )=e d .

[0086] Through the setting of the above start-stop rule, the problem of difficult parameter setting of the conventional extended state observer and the stability problem in practice are effectively avoided.

[0087] S4. According to the real-time data of the extended state observer of the target six-axis vibration table, the fuzzy neural network identifier and the fuzzy neural disturbance observer are trained.

[0088] In specific implementation, the training comprises the following steps:

[0089] The following formula function is used as the loss function E1 of the fuzzy neural network identifier and is trained:

[0090] In the formula, f I is the output of the fuzzy neural network identifier; f O is the output of the extended state observer.

[0091] The following formula function is used as the loss function E of the fuzzy neural disturbance observer, and training is performed:

[0092] In the formula, y is the output of the nominal model; y1 is the output of the actual system.

[0093] S5. The trained fuzzy neural disturbance observer is used to complete disturbance observation of the target six-axis vibration table.

[0094] Since the parameters of the extended state observer are difficult to set, the stability of the system cannot be guaranteed in actual application, the relatively conservative bandwidth of the extended state observer is selected for preliminary control to ensure smooth operation of the system, and then the fuzzy neural disturbance observer is used to replace the extended state observer for disturbance control, which has the advantages of rapidness, accuracy and stability compared with directly using the extended state observer. Since the initial parameters of the fuzzy neural disturbance observer have a great influence on the performance of disturbance estimation and the training process, the fuzzy neural network identifier is used to fit the extended state observer in the present application, so as to quickly determine the appropriate initial value of the fuzzy neural disturbance observer, and then the fuzzy neural disturbance observer is trained in real time based on online data, so as to ensure that the performance of the fuzzy neural disturbance observer is better than that of the extended state observer, and at the same time, the fuzzy neural disturbance observer is prevented from falling into local optimization.

[0095] As shown in Figure 2, the functional module diagram of the system of the present application is shown: the system for implementing the disturbance observation method of the six-axis vibration table disclosed in the present application comprises a data acquisition module, a first construction module, a second construction module, a training module and an observation module; the data acquisition module, the first construction module, the second construction module, the training module and the observation module are connected in series; the data acquisition module is used to acquire data information of the target six-axis vibration table, and upload the data information to the first construction module; the first construction module is used to regard the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances based on the received data information, construct an extended state observer of the target six-axis vibration table, and realize preliminary suppression of the total disturbances based on the output results of the extended state observer, and upload the data information to the second construction module; the second construction module is used to construct a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the constructed extended state observer and fuzzy neural network based on the received data information, set the start-stop rules of the fuzzy neural network identifier and the fuzzy neural disturbance observer, and upload the data information to the training module; the training module is used to train the fuzzy neural network identifier and the fuzzy neural disturbance observer based on the real-time data of the extended state observer of the target six-axis vibration table based on the received data information, and upload the data information to the control module; the observation module is used to complete disturbance observation of the target six-axis vibration table by using the trained fuzzy neural disturbance observer based on the received data information.

[0096] As shown in Figure 3 is a method flow chart of the control method of the present application: the anti-disturbance control method of the present application including the disturbance observation method of the six-axis vibration table, comprising the following steps:

[0097] S1. Obtain the data information of the target six-axis vibration table;

[0098] S2. Take the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as the total disturbance, based on the data information obtained in step S1, construct an extended state observer of the target six-axis vibration table, and realize preliminary suppression of the total disturbance based on the output results of the extended state observer;

[0099] S3. Based on the extended state observer constructed in step S2 and the fuzzy neural network, construct a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table, and set the start-stop rules of the fuzzy neural network identifier and the fuzzy neural disturbance observer;

[0100] S4. According to the real-time data of the extended state observer of the target six-axis vibration table, train the fuzzy neural network identifier and the fuzzy neural disturbance observer;

[0101] S5. Use the trained fuzzy neural disturbance observer to complete the disturbance observation of the target six-axis vibration table;

[0102] S6. According to the disturbance of the target six-axis vibration table observed in real time in step S5, complete the anti-disturbance control of the target six-axis vibration table in real time; Since steps S1-S5 have realized the observation of the disturbance amount of the control target (the target six-axis vibration table), the existing classic sliding mode control scheme, model predictive control scheme and state feedback control scheme can better realize the anti-disturbance control of the target six-axis vibration table.

Claims

1. A method for disturbance observation of a six-axis vibration table, characterized in that The method comprises the following steps: S1. obtaining data information of a target six-axis vibration table; S2. regarding the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances, constructing an extended state observer of the target six-axis vibration table based on the data information obtained in step S1, and realizing preliminary suppression of the total disturbances based on the output results of the extended state observer; S3. constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the extended state observer and the fuzzy neural network constructed in step S2, and setting the start-stop rules of the fuzzy neural network identifier and the fuzzy neural disturbance observer; S4. training the fuzzy neural network identifier and the fuzzy neural disturbance observer according to the real-time data of the extended state observer of the target six-axis vibration table; S5. completing disturbance observation of the target six-axis vibration table by using the trained fuzzy neural disturbance observer.

2. The method of claim 1, wherein The step S2 of regarding the nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table as total disturbances comprises the following steps: The nonlinear factors, uncertain factors and external disturbances of the target six-axis vibration table are regarded as total disturbances, so that the transfer function is determined as Y(s)=G′(s)(U(s)+g(s)), wherein Y(s) is the output of the system; G′(s) is the actual system model including the total disturbances, and G′(s)=G(s)(1+Δ(s)), G(s) is the numerical model of the system, and Δ(s) is the disturbance amount; U(s) is the control input of the system; g(s) is the uncertainty of the system; The transfer function is converted into Y(s)=G(s)(U(s)+g′(s)), wherein g′(s) is the total disturbance of the system; If g′(s) can be completely eliminated, the transfer function can be simplified as a nominal model; 3. The method of disturbance observation of a six-axis shaker of claim 2, wherein The step S2 of constructing the extended state observer of the target six-axis vibration table comprises the following steps: Based on the active disturbance rejection principle, the extended state observer is designed to observe the total disturbance g′(s) of the system; Based on the observed total disturbance of the system z f , the active disturbance rejection control law is designed as u = u0 - z f , where u0 is the input without disturbance rejection.

4. The method of disturbance observation of a six-axis shaker according to claim 3, characterized in that The step S3 of constructing the fuzzy neural network identifier and the fuzzy neural disturbance observer of the target six-axis vibration table comprises the following steps: Based on the fuzzy neural network, the fuzzy neural network identifier of the target six-axis vibration table is constructed, and the fuzzy neural network identifier of the target six-axis vibration table is used to fit the extended state observer; Based on the fuzzy neural network, the fuzzy neural disturbance observer of the target six-axis vibration table is constructed, and the fuzzy neural disturbance observer is a multi-input single-output fuzzy neural network, which is used to replace the extended state observer to observe the disturbance, and suppresses the disturbance through a negative feedback adjustment mechanism after observing the disturbance; The parameters of the fuzzy neural disturbance observer are obtained through the training of the fuzzy neural network identifier.

5. The method of disturbance observation of a six-axis shaker of claim 4, wherein The step S3 of constructing the fuzzy neural network identifier and the fuzzy neural disturbance observer of the target six-axis vibration table comprises the following steps: The constructed fuzzy neural network identifier of the target six-axis vibration table comprises an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer which are sequentially connected; the input of the input layer includes the difference value e between the output value of the extended state observer and the actual value d , the differential value of the difference value e d ​ and control signal u; the output of the network output layer is the increment Af of the fitted extended state observer output z ; the process of the network output layer is represented as where w kiii is an output weight, is the output of the lth fuzzy rule after normalization, and n represents the number of fuzzy rules; The constructed fuzzy neural disturbance observer of the target six-axis vibration table comprises an input layer, a membership function layer, a fuzzy rule fitness layer, a normalization layer and a network output layer connected in sequence; the output of the network output layer is the observed system total disturbance increment Δf o ; the processing process of the network output layer is represented as where w kmmm is an output weight, is the output of the lth fuzzy rule after normalization.

6. The method of disturbance observation of a six-axis shaker of claim 5, wherein The step S3 sets the start-stop rule of the fuzzy neural network identifier and the fuzzy neural disturbance observer, and specifically includes the following steps. The difference e between the total disturbance of the extended state observer output and the output value of the fuzzy neural network identifier d The design discriminant function f(e d ); The value of the discriminant function f(e d ) is judged, so as to set the feedback gain K1 of the extended state observer and the feedback gain K2 of the fuzzy neural disturbance observer: If f(e d )≥α, set K1=0 and K2=1, at this time, the output data of the extended state observer is used to train the fuzzy neural network identifier, after each training, the difference e d of the next round is judged, and the data is input into the fuzzy neural network identifier for training network parameters and adjustment of the next round. If f(e d ) < a, set K1 = 1 and K2 = 0, at this time, the expansion state observer is closed, the parameters in the fuzzy neural network identifier are transmitted to the fuzzy neural disturbance observer, and the fuzzy neural disturbance observer is directly used for disturbance observation, and the subsequent fuzzy neural disturbance observer will be self-trained.

7. The method of disturbance observation of a six-axis shaker of claim 6, wherein The training of step S4 specifically includes the following steps. The following equation function is used as the loss function E1 of the fuzzy neural network identifier, and E1 is based on ESO and is trained: where f I is the output of the fuzzy neural network identifier; f O is the output of the extended state observer; The following equation function is used as the loss function E of the fuzzy neural disturbance observer, and E is based on the nominal model and is trained: In the formula, y is the output of the nominal model; y1 is the output of the actual system.

8. A system for implementing the perturbation observation method of the six-axis vibration table according to any one of claims 1 to 7, characterized by The data acquisition module, the first construction module, the second construction module, the training module and the observation module are connected in series; the data acquisition module is used for acquiring data information of the target six-axis vibration table and uploading the data information to the first construction module; the first construction module is used for regarding the nonlinear factors, the uncertain factors and the external interference of the target six-axis vibration table as total disturbance according to the received data information, constructing an extended state observer of the target six-axis vibration table, and realizing preliminary suppression of the total disturbance based on an output result of the extended state observer, and uploading the data information to the second construction module; The second construction module is used for constructing a fuzzy neural network identifier and a fuzzy neural disturbance observer of the target six-axis vibration table based on the constructed extended state observer and the fuzzy neural network according to the received data information, setting a start-stop rule of the fuzzy neural network identifier and the fuzzy neural disturbance observer, and uploading the data information to the training module; The training module is used for training the fuzzy neural network identifier and the fuzzy neural disturbance observer according to real-time data of the extended state observer of the target six-axis vibration table according to the received data information, and uploading the data information to the control module; and the observation module is used for completing disturbance observation of the target six-axis vibration table by using the trained fuzzy neural disturbance observer according to the received data information.

9. A disturbance observer method for anti-disturbance control including the six-axis vibration table according to any one of claims 1 to 7, characterized by The following steps are further included: S6. Real-time anti-interference control of the target six-axis vibration table is completed according to the disturbance of the target six-axis vibration table obtained by real-time observation of step S5.

Images