Control method and system of vibration test system based on data driving

Abstract

The invention discloses a control method and a control system of a vibration test system based on data driving. The method comprises the following steps: acquiring test data; iterative decision evaluation processing is carried out on the test data based on a reinforcement learning algorithm, and the iterative decision evaluation processing comprises the steps of inputting the test data into a decision evaluation network, outputting an action instruction of a next state, inputting the action instruction of the next state into a vibration test system control model, and outputting the next state, namely a control decision result; receiving and backing up test data through a database module, receiving previous three rounds of iteration data in an iteration decision evaluation process through an experience pool module, receiving historical data of a database, judging the historical data received from the database through a clustering objective function, and feeding back a judgment result to a decision evaluation network; and judging whether the control decision result meets test requirements or not by combining a preset test input condition. The method is high in adaptability, and real data can be effectively utilized.

Description

Technical Field

[0001] The present invention relates to the technical field of vibration test system control, and in particular to a control method and system of a vibration test system based on data drive. Background Art

[0002] The optimization technology of the test system control algorithm can be divided into online optimization and offline optimization. The essential difference lies in whether the optimization process occurs during the use of the test system. The offline optimization method can be adjusted and iterated by analyzing and designing the control model in combination with the test data results previously carried out. Its advantage is that there is no limit on the calculation time, and data information can be more fully utilized for comparative analysis. However, there is also a problem that the controller parameters and configuration are fixed, that is, the offline corrected model is specific, which makes it unable to adapt to the dynamic changes of the environment and the interference in the new environmental state, and thus cannot generate the optimal control solution. The online optimization method is to perform synchronous correction and iteration based on real-time data and prior data in the database during the test, and its overall control model is uncertain. Summary of the invention

[0003] The present invention proposes a control method and system for a vibration test system based on data drive, so as to solve the problem that the control algorithm of the existing test system has weak adaptability and cannot effectively use real data.

[0004] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0005] The present invention provides a control method for a vibration test system based on data drive, comprising:

[0006] Acquire test data, the test data including first data collected and recorded by the sensor module and second data recorded by the environmental data recording module, the first data including data of the vibration test system during the test process, and the second data including information about people, machines, objects, and materials at the test site;

[0007] Performing iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, the iterative decision evaluation processing comprising inputting the test data into a decision evaluation network, outputting an action instruction for a next state, then inputting the action instruction for the next state into a vibration test system control model, and outputting a next state, i.e., a control decision result;

[0008] Receive and back up the test data through the database module, receive the first three rounds of iterative data in the iterative decision evaluation process through the experience pool module, receive the historical data of the database, and judge the historical data received from the database through the clustering objective function, and feed back the judgment result to the decision evaluation network;

[0009] Combined with the preset test input conditions, determine whether the control decision result meets the test requirements. If not, enter the next round of iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

[0010] Furthermore, before obtaining the test data, it also includes:

[0011] Setting test parameters of a vibration test system, wherein the vibration test system is provided with a self-checking system;

[0012] The self-test system output result obtained by the vibration test system through self-testing the self-test system is obtained. If the self-test system output result is normal, the step of obtaining test data is started, otherwise, it ends.

[0013] Furthermore, the method further includes processing the control decision result through a feedback function to obtain a feedback value.

[0014] Specifically, the formula of the cumulative reward function of the reinforcement learning algorithm is as follows:

[0015]

[0016] In the formula, α represents the attenuation factor, , , ... Denotes the discrete time step , , The reward function of , i represents the moment value.

[0017] Specifically, the Q table update formula of the reinforcement learning algorithm is:

[0018]

[0019] In the formula, represents the return reward, β represents the learning rate, α represents the decay factor, Indicates the current state. Indicates action.

[0020] Specifically, the formula of the decision evaluation function O of the reinforcement learning algorithm is as follows:

[0021]

[0022]

[0023] ,

[0024] Where D m represents the clustering objective function of the prior database, d iRepresents historical feature data, v represents cluster center, U represents membership matrix, α (0<α<1) is a regulating factor used to control the influence between current round data and historical data in decision evaluation function, N represents the number of dimensions, C represents the number of cluster centers, and t, k, and m are all used as counts.

[0025] The present invention also provides a control system for the data-driven vibration test system, comprising:

[0026] An environment perception module, the environment module is used to obtain test data, the test data includes first data collected and recorded by the sensor module and second data recorded by the environment data recording module, the first data includes data of the vibration test system during the test process, and the second data includes information about people, machines, objects and materials at the test site;

[0027] A control decision module, the control decision module is connected to the environment perception module, and the control decision module is used to perform iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, wherein the iterative decision evaluation processing includes inputting the test data into a decision evaluation network control model and outputting a next state, i.e., a control decision result;

[0028] A data feedback module, the data feedback module includes a database module and an experience pool module, the experience pool module is connected to the control decision module, the experience pool module is connected to the network, outputs the action instruction of the next state, and then inputs the action instruction of the next state into the vibration test system; the database module is connected to the environment perception module, the database module is used to receive and back up the test data, the experience pool module is used to receive the first three rounds of iteration data of the control decision module, and receive the historical data of the database, and judge the historical data received from the database through the clustering objective function, and feed back the judgment result to the decision evaluation network;

[0029] The first judgment module is used to judge whether the control decision result meets the test requirements in combination with the preset test input conditions. If not, enter the next round of loop iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

[0030] Furthermore, the control system for the data-driven vibration test system also includes a feedback value module, which is connected to the control decision module. The feedback value module is used to process the control decision result through a feedback function to obtain a feedback value.

[0031] Furthermore, the data-driven vibration test system control system also includes a second judgment module, which is used to judge whether the output result of the vibration test self-test system is normal before acquiring the test data. If it is normal, the step of acquiring the test data is started, otherwise it ends.

[0032] Furthermore, the environmental perception module includes a data acquisition subsystem, and the data acquisition subsystem is connected to the control decision module.

[0033] The beneficial effects of the present invention are:

[0034] The data-driven vibration test system control method and system proposed in the present invention have strong adaptability and can effectively utilize real data, thereby improving the control effect of the vibration test system and effectively improving the effectiveness and accuracy of the test system control. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a module flow chart of the data-driven vibration test system control method and system of this application. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0037] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0039] In the description of the present invention, it should be understood that the terms "upper", "lower", "inside", "outside", "left", "right", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are directions or positional relationships in which the product of the invention is usually placed when in use, or are directions or positional relationships commonly understood by those skilled in the art. These directions or positional relationships are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0040] Furthermore, the terms “first”, “second”, etc. are merely used for distinguishing descriptions and should not be understood as indicating or implying relative importance.

[0041] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms such as "setting" and "connection" should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0042] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0043] like Figure 1 As shown, the present invention provides a control method for a vibration test system based on data drive, including:

[0044] Acquire test data, the test data including first data collected and recorded by the sensor module and second data recorded by the environmental data recording module, the first data including data of the vibration test system during the test process, and the second data including information about people, machines, objects, and materials at the test site;

[0045] Specifically, environmental perception mainly consists of three parts: the layout of multiple sensors, environmental data recording and data acquisition subsystem. Among them: In order to better obtain the environmental data of the test site and facilitate the judgment and evaluation of the subsequent control system, different types of (force, heat, temperature, etc.) sensors must be arranged on the equipment and ancillary facilities and instruments to record the test process data. In addition, information such as people, machines, objects, and materials at the test site must also be recorded synchronously. The recorded multi-source data is transmitted to the data acquisition subsystem, and after data preprocessing and preliminary analysis, it is passed to the database in the data feedback module and the input state in the control decision module respectively.

[0046] Performing iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, the iterative decision evaluation processing comprising inputting the test data into a decision evaluation network, outputting an action instruction for a next state, then inputting the action instruction for the next state into a vibration test system control model, and outputting a next state, i.e., a control decision result;

[0047] For iterative decision evaluation processing of the test data based on the reinforcement learning algorithm, specifically:

[0048] The input state receives the data information transmitted by the data acquisition subsystem, calculates through the optimized control algorithm, combines the historical data in the priori database, and jointly makes decisions and evaluates to obtain the action instructions for the next state, which are sent to the control model of the test system to complete the control iteration of the test system.

[0049] Since reasonable vibration control needs to rely on accurate algorithm design, and dynamic environmental changes will have uncertain interference effects on control, based on the above analysis, the present invention proposes a vibration control method based on data-driven adaptive learning strategy, the specific contents of which are as follows:

[0050] The optimization iteration of the control algorithm is in the process of solving the interaction between the environment and the test system. It is necessary to achieve the optimal control effect through continuous and rapid learning. The test system T makes corresponding execution operations A according to the environmental state set H. The test system T responds accordingly according to the corresponding operations, and feeds back the new state and calculation results to the test system T for analysis and judgment, and then makes new execution operations again. In the iterative process, it is necessary to formulate corresponding quantitative indicators, clearly define a decision evaluation function on the environmental state and action, and then use this evaluation function to support the optimization of the control strategy. Repeat the iterative cycle and reuse the test data drive to promote the system to obtain richer feature data information, and improve its own control strategy through feedback.

[0051] Since the entire learning process will obtain new test data information as operations are executed, decision evaluation functions are updated, and states are updated, that is, the next state of the system as a whole is only related to the current state and has nothing to do with the previous state. Therefore, the entire process can be represented by a Markov decision process, which consists of a quaternion array (S, A, P, R), where S represents the state space; A represents the action space; P represents the probability of uncertain state transition; and R represents the reward value function. In the modified iterative process, the overall goal is to maximize the cumulative reward value, and then obtain the optimal control strategy. In the entire Markov decision process, the changes in rewards and states are mainly based on the current state and action, and its reward function r is:

[0052]

[0053]

[0054]

[0055]

[0056] In the formula, k1, k2, k3 represent weight coefficients; r1 represents the reward function for walking distance, where This is the difference between the control signal amplitude and the reference amplitude. is the reference signal amplitude; r2 represents the error feedback reward function, where is the overall error value under this cycle, c is a constant to prevent singularity; r3 represents the feedback reward function of the environment, where, is a discrete time step, and by observing the current state , and select an action After that, the environment responds with the reward.

[0057] In the learning process, the main goal of the test system is to optimize the strategy, specify the execution operation of the next stage, and transform it into the control algorithm design, that is, a cumulative reward function that can characterize the value of different execution operations is required. The formula of the cumulative reward function of the reinforcement learning algorithm is as follows:

[0058]

[0059] In the formula, α represents the attenuation factor, , , ... Denotes the discrete time step , , The reward function of , i represents the moment value.

[0060] By designing a Q table for storing state-action, and evaluating the environmental adaptability of the pre-selected execution operations of the test system according to the reward function: if the state-action pair receives positive feedback rewards from the environment, its corresponding Q value will continue to increase; conversely, if the state-action pair receives negative feedback rewards from the environment, its corresponding Q value will continue to decrease. The Q table update formula of the reinforcement learning algorithm is:

[0061]

[0062] In the formula, represents the return reward, β represents the learning rate, α represents the decay factor, Indicates the current state. represents an action, and a represents the action at the current moment.

[0063] According to the experimental training, the corresponding Q value is obtained, and the mapping relationship between action and state is established to construct the corresponding network model. The network model mainly includes two core neural networks, the current network and the target value network. The network is trained under the same structural parameters and the square of the network difference is used as the error function to achieve the reverse regulation of the neural network. Then, the decision evaluation function is used to select the experimental system to perform the operation, so as to obtain the best strategy to gradually converge the value function. The formula of the decision evaluation function O of the reinforcement learning algorithm is as follows:

[0064]

[0065]

[0066] ,

[0067] Where D m represents the clustering objective function of the prior database, d i Represents historical feature data, v represents clustering center, U represents membership matrix, α (0<α<1) is a regulating factor used to control the degree of influence between current round data and historical data in decision evaluation function, N represents the number of dimensions, and C represents the number of clustering centers.

[0068] The test data is received and backed up through a database module, the first three rounds of iterative data in the iterative decision evaluation process are received through an experience pool module, the historical data of the database is received, and the historical data received from the database is judged through a clustering objective function, and the judgment result is fed back to the decision evaluation network, thereby providing support for control iteration;

[0069] Combined with the preset test input conditions, determine whether the control decision result meets the test requirements. If not, enter the next round of iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

[0070] In some embodiments, when the test is completed, clicking Stop on the software operation interface of the software based on the present invention can stop the host computer program and return to the non-logged-in state. This operation does not affect the real-time program running of the lower computer. Before stopping, please confirm that the conditions for the operation of the lower computer are currently required, otherwise "clear all configurations" and "confirm update configuration" to the lower computer, and then stop running the host computer program.

[0071] Before obtaining test data, it also includes:

[0072] Setting test parameters of a vibration test system, wherein the vibration test system is provided with a self-checking system;

[0073] The self-test system output result obtained by the vibration test system through self-testing the self-test system is obtained. If the self-test system output result is normal, the step of obtaining test data is started, otherwise, it ends.

[0074] Specifically, according to the specific test conditions, set the relevant test name, reference spectrum, tolerance line, control input and output channels, overall test magnitude, and octave required by the test. After confirmation, the relevant information is sent to the control system host computer to facilitate the subsequent iteration and judgment design of the control system. According to the self-test starting voltage, the reference spectrum signal is sent and gradually increased to reach the self-test magnitude. By detecting the output of the system, it is determined whether the system is working normally. If it is normal, it will proceed to the next step, otherwise it will end.

[0075] The method further includes processing the control decision result through a feedback function to obtain a feedback value. Specifically, after each round of test state update, the control decision result and the reference value are input into the feedback function to obtain a corresponding feedback value.

[0076] The present invention also provides a control system for the data-driven vibration test system, comprising:

[0077] An environment perception module, the environment module is used to obtain test data, the test data includes first data collected and recorded by the sensor module and second data recorded by the environment data recording module, the first data includes data of the vibration test system during the test process, and the second data includes information about people, machines, objects and materials at the test site;

[0078] A control decision module, the control decision module is connected to the environment perception module, and the control decision module is used to perform iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, wherein the iterative decision evaluation processing includes inputting the test data into a decision evaluation network control model and outputting a next state, i.e., a control decision result;

[0079] A data feedback module, the data feedback module includes a database module and an experience pool module, the experience pool module is connected to the control decision module, the experience pool module is connected to the network, outputs the action instruction of the next state, and then inputs the action instruction of the next state into the vibration test system; the database module is connected to the environment perception module, the database module is used to receive and back up the test data, the experience pool module is used to receive the first three rounds of iteration data of the control decision module, and receive the historical data of the database, and judge the historical data received from the database through the clustering objective function, and feed back the judgment result to the decision evaluation network;

[0080] The first judgment module is used to judge whether the control decision result meets the test requirements in combination with the preset test input conditions. If not, enter the next round of loop iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

[0081] The control system for the data-driven vibration test system further includes a feedback value module, which is connected to the control decision module. The feedback value module is used to process the control decision result through a feedback function to obtain a feedback value.

[0082] The control system for the vibration test system based on data drive also includes a second judgment module, which is used to judge whether the output result of the vibration test self-checking system is normal before acquiring the test data, and if it is normal, the step of acquiring the test data is started, otherwise it ends. The control method and system of the vibration test system based on data drive proposed by the present invention solves the problem.

[0083] The environment perception module includes a data acquisition subsystem, and the data acquisition subsystem is connected to the control decision module.

[0084] The advantages of the present invention compared to the prior art are:

[0085] Making full use of the test data of the entire test cycle and the previous data in the priori database can effectively solve the problems of poor adaptability of the vibration control environment state and local deviations in the control results caused by factors such as nonlinearity and local resonance of the product test pieces, and effectively improve the control capability of the vibration environment test;

[0086] The idea of ​​online learning is adopted, and the collected test process data is input into the decision network. It is also proposed to combine the historical data in the database to form a joint decision function based on the synchronous correction and iteration of real-time data from multiple sources and the prior data in the database. The limited data in the sample space is fully utilized to construct an adaptive control network with certain generalization ability. The model parameter generation and iteration are driven by test data, which effectively overcomes the shortcomings of weak adaptability and inability to effectively utilize real data, thereby improving the control effect.

[0087] The adaptability improvement and optimization design of the overall test system have been improved. Mainly through the design of environmental perception module, data feedback module and control decision module, the control algorithm is effectively integrated into the actual vibration test system, realizing the overall design of intelligent control vibration test system, which can effectively improve the effectiveness and accuracy of test system control.

[0088] Under the premise of ensuring that the test equipment and test efficiency are not affected, the present invention optimizes the design of the control algorithm, combines the real-time data with the historical data in the database, inputs them into the decision-making evaluation network for joint judgment, iterates the action instructions suitable for the current environmental conditions of the test system, and optimizes the control of the test system. In addition, based on the intelligent control algorithm, a corresponding improved test system design is proposed, which effectively improves the control capability of the vibration environment test.

[0089] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A control method for a vibration test system based on data drive, characterized in that: include: Acquire test data, the test data including first data collected and recorded by the sensor module and second data recorded by the environmental data recording module, the first data including data of the vibration test system during the test process, and the second data including information about people, machines, objects, and materials at the test site; Performing iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, the iterative decision evaluation processing comprising inputting the test data into a decision evaluation network, outputting an action instruction for a next state, then inputting the action instruction for the next state into a vibration test system control model, and outputting a next state, i.e., a control decision result; Receive and back up the test data through the database module, receive the first three rounds of iterative data in the iterative decision evaluation process through the experience pool module, receive the historical data of the database, and judge the historical data received from the database through the clustering objective function, and feed back the judgment result to the decision evaluation network; Combined with the preset test input conditions, determine whether the control decision result meets the test requirements. If not, enter the next round of iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

2. The control method of a vibration test system based on data drive according to claim 1, characterized in that: Before obtaining test data, it also includes: Setting test parameters of a vibration test system, wherein the vibration test system is provided with a self-checking system; The self-test system output result obtained by the vibration test system through self-testing the self-test system is obtained. If the self-test system output result is normal, the step of obtaining test data is started, otherwise, it ends.

3. The control method of a vibration test system based on data drive according to claim 1, characterized in that: The method also includes processing the control decision result through a feedback function to obtain a feedback value.

4. The control method of a vibration test system based on data drive according to claim 1, characterized in that: The formula for the cumulative reward function of the reinforcement learning algorithm is as follows: ， In the formula, α represents the attenuation factor, , , ... Denotes the discrete time step , , The reward function of , i represents the moment value.

5. The control method of a vibration test system based on data drive according to claim 4, characterized in that: The Q table update formula of the reinforcement learning algorithm is: ， In the formula, represents the return reward, β represents the learning rate, α represents the decay factor, Indicates the current state. Represents the current action, and a represents the action at the current moment.

6. The control method of a vibration test system based on data drive according to claim 5, characterized in that: The formula of the decision evaluation function O of the reinforcement learning algorithm is as follows: ， ， ， ， Where D m represents the clustering objective function of the prior database, d i Represents historical feature data, v represents clustering center, U represents membership matrix, α (0<α<1) is a regulating factor used to control the degree of influence between current round data and historical data in decision evaluation function, N represents the number of dimensions, and C represents the number of clustering centers.

7. A control system for a data-driven vibration test system according to any one of claims 1 to 6, characterized in that: include: An environment perception module, the environment module is used to obtain test data, the test data includes first data collected and recorded by the sensor module and second data recorded by the environment data recording module, the first data includes data of the vibration test system during the test process, and the second data includes information about people, machines, objects and materials at the test site; A control decision module, the control decision module is connected to the environment perception module, and the control decision module is used to perform iterative decision evaluation processing on the test data based on a reinforcement learning algorithm, wherein the iterative decision evaluation processing includes inputting the test data into a decision evaluation network control model and outputting a next state, i.e., a control decision result; A data feedback module, the data feedback module includes a database module and an experience pool module, the experience pool module is connected to the control decision module, the experience pool module is connected to the network, outputs the action instruction of the next state, and then inputs the action instruction of the next state into the vibration test system; the database module is connected to the environment perception module, the database module is used to receive and back up the test data, the experience pool module is used to receive the first three rounds of iteration data of the control decision module, and receive the historical data of the database, and judge the historical data received from the database through the clustering objective function, and feed back the judgment result to the decision evaluation network; The first judgment module is used to judge whether the control decision result meets the test requirements in combination with the preset test input conditions. If not, enter the next round of loop iteration; if satisfied, output the corresponding parameters and curves, and the test ends.

8. The control system of a vibration test system based on data drive according to claim 7, characterized in that: It also includes a feedback value module, which is connected to the control decision module. The feedback value module is used to process the control decision result through a feedback function to obtain a feedback value.

9. The control system of a vibration test system based on data drive according to claim 7, characterized in that: It also includes a second judgment module, which is used to judge whether the output result of the vibration test self-test system is normal before acquiring the test data. If it is normal, the step of acquiring the test data begins, otherwise it ends.

10. The control system of a vibration test system based on data drive according to claim 7, characterized in that: The environment perception module includes a data acquisition subsystem, and the data acquisition subsystem is connected to the control decision module.

Images