Joint Optimization Method of Sensor Placement and Feedback Control Algorithm in Structural Vibration Control

Abstract

The present invention belongs to the technical field of vibration control. The present invention relates to a combined optimization method for sensor placement and feedback control algorithm in structural vibration control. A continuous-time recurrent neural network (CTRNN) is used as the basic neuron, and the NEAT algorithm is used to optimize the network structure to obtain the optimal network structure. The optimal network structure is used as the control strategy of the vibration control system to obtain the positions of the sensors. When the structure is subjected to an external excitation, the optimal control force is output by the optimal network structure based on the state responses measured by the sensors at the corresponding positions, and then the actuator is controlled to generate the corresponding control force. The present invention can accurately optimize the positions of the sensors, and at the same time evolve the simplest neural network architecture, improving the overall efficiency of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of vibration control, relates to structural vibration control technology, and specifically relates to a method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control. Background Art

[0002] Structural vibration may cause fatigue damage to engineering structures such as buildings, bridges, and aircraft, shorten the service life, and even cause structural failure or collapse in extreme cases, endangering safety. At present, structural vibration control methods are mainly divided into three categories: passive control, active control, and semi-active control. Passive control does not require external energy input and relies on passive elastic or damping elements to achieve vibration reduction effects, with the advantages of simple structure and high reliability. However, since the control force of passive control is generated passively with the vibration deformation of the structure and cannot actively adjust the control strategy to adapt to vibrations of different frequencies, its effect in broadband vibration suppression is relatively poor. Active control is based on modern control theory and includes sensors for collecting structural vibration information, a controller for controlling actuators according to a control algorithm, and actuators for applying control forces. Active control can drive the actuators through external energy to generate the optimal control force and suppress structural vibration, so it theoretically has a significantly better vibration reduction effect than passive control. Semi-active control adjusts its control strategy by using a small amount of energy to change the parameters of the control device (such as the damping coefficient). Although semi-active control solves the problem of high energy consumption of active control, its vibration reduction effect is between passive control and active control, and it does not achieve the vibration suppression performance of active control.

[0003] Although active control technology theoretically provides superior vibration suppression performance, it has not been widely applied in practical applications, mainly due to the complexity of the system. In an active control system, the linear quadratic regulator (hereinafter referred to as: LQR) algorithm is a commonly used control algorithm. However, the LQR algorithm relies on the full-state feedback of the system, and in practical applications, the number of degrees of freedom of large structures is huge, and it is often impossible to install a sufficient number of sensors in practice to achieve full-state feedback. Therefore, the active control method based on output feedback has become a feasible solution to meet the actual control requirements. However, the existing active control technologies based on output feedback have the following deficiencies:

[0004] (1) The problem of sensor placement: The placement of sensors usually relies on empirical design, making it difficult to ensure that they are in the optimal position and difficult to ensure effective matching with the feedback control algorithm.

[0005] (2) The problem of feedback control optimization: Artificial neural networks are often used as tools for output feedback optimization, but their initial configuration (such as the number of nodes and layers) is usually set based on experience and has a significant impact on the final optimization effect. Summary of the Invention

[0006] In view of the problems existing in the prior art, the present invention provides a method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control, which can improve the overall control efficiency and effect of the structural vibration control system.

[0007] In the first aspect of the present invention, a method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control is provided, and its steps include:

[0008] S1. Establish a simulation model of the vibration control system according to the set model parameters of the structural vibration control system. The input layer of each network structure of the simulation model is the state response of the structure, the first hidden layer is the sensor, and the output layer is the control force;

[0009] S2. Between the input layer and the first hidden layer of each network structure, a feedforward neural network is used as the basic neuron; between the first hidden layer and the output layer of each network structure, a continuous-time recurrent neural network CTRNN is used as the basic neuron;

[0010] S3. Use the NEAT algorithm to optimize the structure part from the first hidden layer to the output layer of each network structure to obtain the optimal network structure;

[0011] S4. Take the optimal network structure as the control strategy of the vibration control system and obtain the positions of the sensors. When the structure is subjected to an external excitation, through the state response measured by the sensors at the corresponding positions, the optimal network structure outputs the optimal control force, and then controls the actuator to generate the corresponding control force.

[0012] In some embodiments, in step S2, the input of the feedforward neural network is the entire state response of the structure, and the feedforward neural network randomly selects less than or equal to state responses from the entire state response as the input of the first hidden layer.

[0013] In some embodiments, in step S2, for the structure part from the first hidden layer to the output layer of each network structure, the nodes and connection weights of the structure part are initialized according to a Gaussian distribution , whose mean is , whose standard deviation is , and whose probability density function is:

[0014]

[0015] In the formula, is the probability density function, is the input variable;

[0016] The initial connection of the structural part is a full connection. The initial neural network type of the structural part is a continuous-time recurrent neural network (CTRNN). The input of the structural part is the input of the first hidden layer, and the output is the control force.

[0017] The behavioral equation of nodes in the continuous-time recurrent neural network (CTRNN) is as follows:

[0018]

[0019] In the formula, is the time constant of neuron , is the potential of neuron , is the activation function of neuron , is the bias of neuron , is the index set of neurons that provide input to neuron , is the connection weight from neuron to neuron .

[0020] In some embodiments, the specific method for optimizing the structure from the first hidden layer to the output layer of each network structure using the NEAT algorithm to obtain the optimal network structure is as follows:

[0021] S31: Take all network structures as a population, and each network structure is an individual in the population;

[0022] S32: Use the fitness function to evaluate the individuals in the population. If the fitness of an individual in the population exceeds the fitness threshold, stop the optimization; otherwise, proceed to step S33;

[0023] S33: Select individuals from the parent generation for crossover to generate offspring, and perform mutation operations on the offspring to form a new population;

[0024] S34: Divide the new population into different subpopulations according to the competition rule, and perform species division within each subpopulation;

[0025] S35: Repeat steps S32 to S34 until the preset stop condition is reached.

[0026] In some embodiments, in step S32, the negative of the quadratic cost function is set as the fitness function, and the fitness function is expressed as:

[0027]

[0028] In the formula, is the structural state response of the system, is the control force, state weight matrix, control force weight matrix, is the initial time, is the termination time.

[0029] In some embodiments, during the optimization process, a shared fitness is set within each species :

[0030]

[0031] In the formula, is the sharing function. When , its value is 0. When , its value is 1; is the th individual, is the th individual, is the distance threshold, is the number of each species.

[0032] In some embodiments, in step S33, during the crossover process, the parts with the same structure in different network structures are matched one by one, and the parts with different structures are selected from the parent with better fitness.

[0033] In some embodiments, in step S33, the mutation includes adding nodes and adding connections; the method of adding nodes is: adding new nodes to break the old connections and generating two new connections; the method of adding connections is: adding a new connection at the end of the current connection list; both the new nodes and the new connections mutations are assigned new gene codes.

[0034] In some embodiments, in step S34, all individuals are divided into subpopulations according to the layout positions of the sensors, and each subpopulation has an id.

[0035] In some embodiments, in step S34, when dividing species within each subpopulation, the species are divided according to the similarity distance between the neural networks of different individuals; the similarity distance is expressed as:

[0036]

[0037] In the formula, is the similarity distance between two individuals, is the number of different redundant genes, is the number of different missing genes, is the average difference of the connection weights in the matching genes, is the coding number of the larger genome in the two individuals, To adjust different numbers of redundant genes The coefficient of importance To adjust different numbers of deleted genes The coefficient of importance To adjust the average difference in connection weights among matching genes The coefficient of importance;

[0038] If the similarity distance of the current generation of individuals is less than the distance threshold , it means they belong to the same species. If the similarity distance of the current generation of individuals is greater than or equal to the distance threshold , it means they do not belong to the same species, and a new species is created with this individual as the representative.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0040] (1) The method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control provided by the present invention adopts an enhanced topology neural evolution algorithm to optimize the neural network structure. By using the local measurement output of the structure as the input of the control algorithm to replace the traditional full-state feedback-based control method, an output feedback optimal controller is designed. It not only optimizes the control strategy but also can optimize the placement position of sensors, and solves the optimal layout of sensors and its corresponding optimal control strategy. Compared with the traditional experience-based sensor placement, the present invention can accurately optimize the position of sensors, and at the same time evolve the simplest neural network architecture, improving the overall efficiency of the system.

[0041] (2) In traditional neural network control, the network structure needs to be predefined. Whether the structure is too simple or too complex will affect the optimization effect. The method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control provided by the present invention starts from the simplest network structure and uses an evolutionary algorithm to gradually optimize the network structure, solving the problem of performance degradation caused by improper initial network structure design.

[0042] (3) Traditional neural network control algorithms usually rely on experience to preset the installation position of sensors and cannot optimize the sensor layout. The method for jointly optimizing sensor placement and feedback control algorithm in structural vibration control provided by the present invention not only improves the control algorithm but also optimizes the layout position of sensors between the input layer and the first hidden layer, thereby improving the overall performance and control effect of the system.

[0043] (4)The joint optimization method of sensor layout and feedback control algorithm in the structural vibration control provided by the present invention. The evolution of the neural network structure and the search for weights greatly accelerate the speed of the neural network learning complex features such as non-linearity. The evolutionary algorithm enables the neural network individuals to compete intensively. Survival of the fittest eliminates the individuals with improper structure and weight evolution, further accelerating the optimization speed, and at the same time well avoiding local optima and having good global search ability. Description of the Drawings

[0044] Figure 1 It is a flowchart of the joint optimization method of sensor layout and feedback control algorithm in the structural vibration control described in the embodiment of the present invention;

[0045] Figure 2 It is a schematic diagram of the principle of the joint optimization method of sensor layout and feedback control algorithm in the structural vibration control described in the embodiment of the present invention;

[0046] Figure 3 It is a schematic diagram of adding nodes in the embodiment of the present invention;

[0047] Figure 4 It is a schematic diagram of adding connections in the embodiment of the present invention;

[0048] Figure 5 It is a structural schematic diagram of the three-story building described in the embodiment of the present invention;

[0049] Figure 6 It is a schematic diagram of the initial network structure of the CRNEAT algorithm for the three-story building described in the embodiment of the present invention;

[0050] Figure 7 It is a schematic diagram of the final network structure of the CRNEAT algorithm for the three-story building described in the embodiment of the present invention;

[0051] Figure 8 It is a schematic diagram of the evolution curve of the CRNEAT algorithm for the three-story building described in the embodiment of the present invention;

[0052] Figure 9 It is a displacement time history diagram of the first floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention;

[0053] Figure 10 It is a displacement time history diagram of the second floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention;

[0054] Figure 11 It is a displacement time history diagram of the third floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention;

[0055] Figure 12 It is a velocity time history diagram of the first floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention;

[0056] Figure 13 This is the velocity time history graph of the second floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention;

[0057] Figure 14 This is the velocity time history graph of the third floor under the control of the CRNEAT algorithm and the LQR algorithm of the present invention.

[0058] In the figure, 101 is the sensor, 102 is the controller, and 103 is the actuator. Specific embodiments

[0059] Next, the present invention will be specifically described by way of exemplary embodiments in conjunction with the accompanying drawings. However, it should be understood that, without further elaboration, the elements, structures, and features in one embodiment can also be beneficially incorporated into other embodiments.

[0060] See Figure 1 , 2 , an embodiment of the first aspect of the present invention provides a method for jointly optimizing sensor placement and feedback control algorithms in structural vibration control, and its steps include:

[0061] S1. Establish a simulation model of the vibration control system according to the set model parameters of the structural vibration control system. The input layer of each network structure of the simulation model is the state response of the structure, the first hidden layer is the sensor, and the output layer is the control force.

[0062] Specifically, the vibration control system parameters include simulation duration, simulation step size, external excitation, and system matrix.

[0063] S2. Between the input layer and the first hidden layer of each network structure, use a feedforward neural network as the basic neuron to optimize the position of the sensor; between the first hidden layer and the output layer of each network structure, use a continuous-time recurrent neural network CTRNN as the basic neuron to optimize through network structure evolution.

[0064] Specifically, the input of the feedforward neural network is the entire state response of the structure , is the number of structural state responses. The feedforward neural network randomly selects less than or equal to state responses from all state responses as the input of the first hidden layer.

[0065]

[0066] In the formula, is the desired number of sensors, is the input of the first hidden layer.

[0067] For the first hidden layer to the output layer structure part of each network structure, initialize the weights of the nodes and connections in the structure part according to the Gaussian distribution , whose mean is , whose standard deviation is , and whose probability density function is:[[]]

[0068]

[0069] In the formula, is the probability density function,[[]] is the input variable;[[]]

[0070] The initial connection of the structure part is a full connection, the initial neural network type of the structure part is a continuous-time recurrent neural network CTRNN, the input of the structure part is the input of the first hidden layer, and the output is the control force;[[]]

[0071] The behavior equation of the nodes in the continuous-time recurrent neural network CTRNN is:[[]]

[0072]

[0073] In the formula,[[]] is the time constant of the neuron , is the potential of the neuron , is the activation function of the neuron , is the bias of the neuron , is the index set of the neurons that provide input to the neuron , is the connection weight from the neuron to the neuron .[[]]

[0074] S3. Use the NEAT algorithm to optimize the first hidden layer to the output layer structure part of each network structure to obtain the optimal network structure.[[]]

[0075] Specifically, the specific method of using the NEAT algorithm to optimize the first hidden layer to the output layer structure part of each network structure to obtain the optimal network structure is:[[]]

[0076] S31. Regard all network structures as a population, and each network structure is an individual in the population.[[]]

[0077] S32. Use the fitness function to evaluate the individuals in the population. If the fitness of an individual in the population exceeds the fitness threshold, stop the optimization; otherwise, go to step S33.[[]]

[0078] Specifically, in some embodiments, the negative of the quadratic cost function is set as the fitness function, and the fitness function is expressed as:

[0079]

[0080] In the formula, is the structural state response of the system, is the control force, is the state weight matrix, is the control force weight matrix, is the initial time, is the termination time.

[0081] It should be noted that the smaller the structural state response and the control force of the structure within the time period , the higher the fitness. The fitness threshold is set to 0, and the evolution stops when the fitness reaches the threshold.

[0082] S33. Select individuals from the parent generation for crossover to generate offspring, and perform mutation operations on the offspring to form a new population.

[0083] Specifically, during the crossover process, the parts with the same structure in different network structures are matched one by one, and the parts with different structures are selected from the parent with better fitness.

[0084] Specifically, mutation includes adding nodes and adding connections. The method of adding nodes is: adding a new node to break the old connection and generate two new connections. For example: Refer to Figure 3 , adding a new node 6 breaks the old connection 3→5 and generates two new connections 3→6 and 6→5. The method of adding a connection is: adding a new connection at the end of the current connection list. For example: Refer to Figure 4 , adding a new connection 2→6 at the end of the current Connection list. Both the new node and the new connection mutations are assigned new gene encodings. It should be noted that as the network structure becomes more complex, its gene encoding will gradually expand. Since the new mutations do not change the old gene encoding, the historical origin of the network structure mutations is known throughout the evolution process.

[0085] S34. Divide the new population into different subpopulations according to the competition rules, and perform species division within each subpopulation.

[0086] Specifically, all individuals are divided into subpopulations according to the layout position of the sensors, and each subpopulation has an id. For example: The subpopulation with id 000011 means that all individuals within this subpopulation install sensors at the 5th and 6th positions. After dividing the subpopulations, species division is performed within the subpopulations.

[0087] Specifically, in some embodiments, when dividing species within each sub-population, species are divided according to the similarity distance between the neural networks of different individuals; the similarity distance is expressed as:

[0088]

[0089] In the formula, is the similarity distance between two individuals, is the number of different redundant genes, is the number of different missing genes, is the average difference in connection weights among the matching genes, is the number of codings in the larger genome of the two individuals, is the coefficient for adjusting the number of different redundant genes of importance, is the coefficient for adjusting the number of different missing genes of importance, is the coefficient for adjusting the average difference in connection weights among the matching genes of importance;

[0090] If the similarity distance of the current generation of individuals is less than the distance threshold , it means they belong to the same species. If the similarity distance of the current generation of individuals is greater than or equal to the distance threshold

[0091]

[0092] S35. Repeat steps S32 to S34 until a preset stop condition is reached.

[0093] Specifically, in some embodiments, during the optimization process, a shared fitness is set within each species:

[0094]

[0095] In the formula, is the sharing function, when , its value is 0, when , its value is 1; is the th individual, is the th individual, is the distance threshold, is the number of each species.

[0096] It should be noted that during the evolution process, in order to improve the evolution efficiency, it is necessary to prevent the number of individuals within a certain species from being too large. A shared fitness is set within each species. After each generation of training, a part of the individuals will be eliminated according to the shared fitness, and the remaining individuals will generate new individuals through crossover and mutation. It can be set that if the maximum fitness of a certain species does not increase within several generations, this species will not produce new offspring.

[0097] S4. Using the optimal network structure as the control strategy of the vibration control system and obtaining the positions of the sensors. When the structure is subjected to an external excitation, the state response measured by the corresponding position sensors , from the optimal network structure outputs the optimal control force , and then controls the actuator to generate the corresponding control force. The optimal network structure is the output feedback optimal controller.

[0098]

[0099] The optimization method of the present invention adopts the enhanced topological neural evolution algorithm to optimize the neural network structure. By using the local measurement output of the structure as the input of the control algorithm to replace the traditional full-state feedback-based control method, the problem of performance degradation caused by improper initial network structure design is solved. Through the joint optimization of the neural network, the present invention not only improves the control algorithm but also optimizes the layout position of the sensors between the input layer and the first hidden layer, thereby improving the overall performance and control effect of the system.

[0100] To verify the effectiveness of the joint optimization method of sensor layout and feedback control algorithm in the structural vibration control described in the above embodiments of the present invention, the following specific embodiments are combined for description.

[0101] Embodiment: Taking the structure of a three-story building as an example. The schematic diagram of the structure of the three-story building is shown in Figure 5 . In the figure, is the mass of the first-story building, is the stiffness of the first-story building, is the displacement of the first-story building relative to the ground, is the mass of the second-story building, is the stiffness of the second-story building, is the displacement of the second-story building relative to the ground, is the mass of the third-story building, The stiffness of the third-story building is the displacement of the third-story building relative to the ground, is the absolute ground displacement, and is the structural control force. The equation of motion of the third-story building in matrix form is expressed as:

[0102]

[0103] where, 、 、 are the structural mass matrix, damping matrix, and stiffness matrix, respectively; 、 、 are the relative displacement, velocity, and acceleration vectors, respectively; is the environmental disturbance location matrix; is the actuator layout matrix; is the control force, is the external load vector.

[0104] The equation of motion is transformed into state-space form and expressed as:

[0105]

[0106] where,

[0107]

[0108]

[0109]

[0110]

[0111] In the formula, is the state vector, is the state matrix, is the input matrix of the external load, is the input matrix of the control force.

[0112] In a structure with base isolation, the external load is the base excitation , and passive, semi-active, or active dampers used in the isolation system are installed between the ground and the first mass of the structure.

[0113] In this embodiment, the mass matrix 、damping matrix 、stiffness matrix 、environmental disturbance location matrix and actuator layout matrix of the three-story building are:

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] In the formula, kg represents kilogram, N represents newton, s represents second, and m represents meter.

[0120] Set the fitness function of CRNEAT as:

[0121]

[0122] where

[0123]

[0124] In the formula, is the identity matrix

[0125] Adopt the joint optimization method of sensor layout and feedback control algorithm in the structural vibration control described in the present invention (hereinafter referred to as: CRNEAT algorithm), take as the input of the neural network, and set the installation quantity of sensors not exceeding 2. The initial neural network is as Figure 6 shown, and the final neural network is as Figure 7 shown. In Figure 7 , In1, In2, In3, In4, In5, In6 are the displacements and velocities of the 1st - 3rd layers respectively, act1 is the final control force, and 256, 1429, 1397 are the evolved hidden layer nodes. The results show that velocity sensors should be adopted and installed on the 1st layer and the 3rd layer. Figure 8This is the evolutionary curve of the optimal individual in the sub-population of the CRNEAT algorithm of the present invention. The optimal sub-population adopts the configuration with speed sensors installed on the first and third floors. Sub-population 1 installs speed sensors only on the third floor; Sub-population 2 selects to install displacement sensors on the first floor and speed sensors on the second floor; The configuration of Sub-population 3 is to install displacement sensors on the third floor and speed sensors on the first floor. It can be seen from the figure that the fitness values of each sub-population are generally low in the initial stage, but as the number of evolutionary generations increases, the fitness of the optimal individuals in different sub-populations gradually improves. In particular, Sub-population 1 and Sub-population 2 become extinct after reaching the maximum stagnation generation, while Sub-population 3 and the optimal sub-population continue to evolve and finally converge to the local optimal solutions in their respective solution space domains. The results show that the CRNEAT algorithm adopting the species division strategy significantly improves the solution efficiency. By evolving the neural network in parallel in different solution space domains, the algorithm can effectively avoid local optimal solutions, improve the global search ability, and thus accelerate the optimization process.

[0126] The optimization method proposed by the present invention has excellent vibration control effect on earthquakes. To better illustrate the performance of this method, the optimal control LQR algorithm is introduced for comparison. The optimal control LQR algorithm is the most widely used control in the field of structural control, and its expression is:

[0127]

[0128] In the formula, is the optimal feedback gain matrix, is the parameter matrix of the optimal controller.

[0129] To compare the superiority of the CRNEAT algorithm of the present invention, the displacement and velocity time history diagrams of different floors under the control of the CRNEAT algorithm and the LQR algorithm of the present invention are compared, as Figures 9 to 14 shown. Combining the root mean square values and peak values of displacement, velocity and acceleration of the LQR and CRNEAT algorithms shown in Table 1, it can be seen that the control performance of the CRNEAT algorithm of the present invention is basically the same as that of the LQR control.

[0130] Table 1

[0131]

[0132] In summary, the optimization method of the present invention can optimize the sensor layout while optimizing the control strategy, solving the problem of possible degradation of model performance caused by the traditional installation of sensors based on experience. Through the evolutionary algorithm, neural network individuals conduct natural selection in competition, survival of the fittest, eliminating individuals with improper structural and weight evolution, accelerating the optimization process, effectively avoiding local optima. In addition, the optimization method of the present invention can also evolve the simplest neural network structure, further improving the efficiency and performance of the system and demonstrating powerful global search capabilities.

[0133] The above embodiments are used to explain the present invention rather than limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims of the present invention fall within the protection scope of the present invention.

Claims

1. A joint optimization method for sensor placement and feedback control algorithm in structural vibration control, characterized in that: The steps include: S1. Establishing a simulation model of the vibration control system according to the set model parameters of the structural vibration control system, wherein the input layer of each network structure of the simulation model is the state response of the structure, the first hidden layer is the sensor, and the output layer is the control force; S2. Between the input layer and the first hidden layer of each network structure, a feedforward neural network is used as the basic neuron; between the first hidden layer and the output layer of each network structure, a continuous time recurrent neural network CTRNN is used as the basic neuron; the input of the feedforward neural network is the total state response of the structure, and the feedforward neural network randomly selects less than or equal to from all state responses The state response is used as the input of the first hidden layer. is the desired number of sensors; S3. The NEAT algorithm is used to optimize the structure from the first hidden layer to the output layer of each network structure to obtain the optimal network structure. In the optimization process, all network structures are regarded as a population, each network structure is an individual in the population, and the individuals in the population are evaluated using the fitness function. The negative number of the quadratic cost function is set as the fitness function, and the fitness function is expressed as: In the formula, is the structural state response of the system, For control, The state weight matrix, Control weight matrix, is the initial time, is the end time; S4. Use the optimal network structure as the control strategy of the vibration control system and obtain the position of the sensor. When the structure is subjected to external excitation, the state response measured by the corresponding position sensor is used, and the optimal control force is output by the optimal network structure, thereby controlling the actuator to generate the corresponding control force.

2. The joint optimization method of sensor arrangement and feedback control algorithm in structural vibration control according to claim 1, characterized in that: In step S2, for each network structure from the first hidden layer to the output layer, the weights of the nodes and connections of the structure are initialized according to the Gaussian distribution. , whose mean is , and its standard deviation is , its probability density function is: In the formula, is the probability density function, is the input variable; The initial connection of the structure part is full connection, the initial neural network type of the structure part is continuous time recurrent neural network CTRNN, the input of the structure part is the input of the first hidden layer, and the output is control force; The behavior equation of the nodes in the continuous-time recurrent neural network CTRNN is: In the formula, It's a neuron The time constant, It's a neuron The potential, It's a neuron The activation function, It's a neuron The bias of It is to the neuron The set of indices of the neurons that provide input, It's a neuron To the neuron The connection weight of .

3. The joint optimization method of sensor arrangement and feedback control algorithm in structural vibration control according to claim 1, characterized in that: The specific method of using the NEAT algorithm to optimize the first hidden layer to the output layer structure of each network structure to obtain the optimal network structure is: S31. All network structures are regarded as a population, and each network structure is an individual in the population; S32, using the fitness function to evaluate the individuals in the population, if the fitness of an individual in the population exceeds the fitness threshold, then stop the optimization, otherwise go to step S33; S33, selecting individuals from the parent generation to perform crossover to generate offspring, and performing mutation operations on the offspring to form a new population; S34. Divide the new population into different sub-populations according to the competition rules, and divide species within each sub-population; S35. Repeat steps S32 to S34 until a preset stop condition is reached.

4. The method for joint optimization of sensor arrangement and feedback control algorithm in structural vibration control according to claim 3, characterized in that: During the optimization process, a shared fitness is set within each species. : In the formula, is a shared function, when , its value is 0, when , its value is 1; For the Individuals, For the Individuals, is the distance threshold, The number for each species.

5. The method for joint optimization of sensor arrangement and feedback control algorithm in structural vibration control according to claim 3, characterized in that: In step S33, during the crossover process, parts with the same structure in different network structures are matched one by one, and parts with different structures are selected from parents with better fitness.

6. The method for joint optimization of sensor placement and feedback control algorithm in structural vibration control according to claim 3, characterized in that: In step S33, mutations include adding nodes and adding connections; the method for adding nodes is: adding a new node to disconnect the old connection and generate two new connections; the method for adding connections is: adding a new connection at the end of the current connection list; new node and new connection mutations are both given new genetic codes.

7. The method for joint optimization of sensor arrangement and feedback control algorithm in structural vibration control according to claim 3, characterized in that: In step S34, all individuals are divided into sub-populations according to the layout positions of the sensors, and each sub-population has an id.

8. The method for joint optimization of sensor placement and feedback control algorithm in structural vibration control according to claim 3, characterized in that: In step S34, when dividing species within each subpopulation, the species are divided according to the similarity distance between the neural networks of different individuals; the similarity distance is expressed as: In the formula, is the similarity distance between two individuals, For different numbers of redundant genes, For different numbers of missing genes, is the average difference in connection weights among matched genes, is the number of codes in the larger genome of the two individuals, To regulate the number of different redundant genes The coefficient of importance, To adjust the number of missing genes The coefficient of importance, The average difference in connection weights among adjusted matched genes The coefficient of importance; If the similarity distance of contemporary individuals Less than distance threshold , it means they belong to the same species. If the similarity distance of contemporary individuals is greater than or equal to the distance threshold , it means that they do not belong to the same species, and a new species is created with this individual as the representative.

Images