CN115794644B - A real-time hybrid test method based on single-specimen restart and multi-task loading - Google Patents

Abstract

The invention provides a real-time mixed test method based on restarting multi-task loading of a single test piece. The present invention uses a test piece to sequentially load the calculated load targets at multiple initial moments through the test loading system until all the test substructures are loaded, and feeds back the test data measured after the loading to the numerical simulation system for calculation to obtain real-time loading. After comparing the accuracy of the target and/or updating the numerical test substructure model to obtain the truncated loading target, send the real-time loading target and/or the truncated loading target to the test loading system, load the specimen in sequence, reset the test object, until the test object is completed The process of judging the accuracy.

Description

A real-time hybrid test method based on single-specimen restart and multi-task loading

technical field

The invention specifically relates to a real-time mixed test method based on restarting multi-task loading of a single test piece.

Background technique

The mixed test is to accurately simulate the numerical simulation part that is easy to simulate simply and accurately by numerical method and the non-numerical part that is not easy to simulate simply and accurately by numerical method. It can save test cost, improve efficiency and accurately Simulate structural vibration or dynamic performance, and on this basis, for structures with dynamic characteristics that need to be studied, the numerical substructure and the experimental substructure need to be coupled with the same time step and online synchronous real-time coupling to complete the dynamic analysis or performance evaluation of the research object . However, in order to ensure the accuracy, the numerical substructure often requires more refined calculations, which is prone to the problem of time lag. Therefore, the restart method can not only apply mixed experiments but also perfectly solve the problem of time lag. However, restarting Loading from the first step of the experiment every time can undoubtedly guarantee the accuracy of the experiment, but when the total loading time is long, when the loading is in the middle and late stages, the loading target in the initial stage has very little impact on the loading in the current stage. Therefore, loading from the first step of the experiment every time will increase unnecessary test costs and waste of manpower and material resources. At the same time, when the substructures that need to be tested are the same but the stress conditions are different, using multiple test substructures to load at the same time will greatly increase the test cost, and the laboratory loading conditions are difficult to meet the requirements. A series of unfavorable conditions, such as Yes, for complex structures, traditional hybrid tests usually take the nonlinear part as the test substructure for real physical loading tests, and the rest of the linear part as the numerical substructure for modeling analysis. When there are mixed tests of multiple nonlinear components in the structure, there are usually two test methods, which are:

One is the conventional real-time mixed test, which requires multiple test pieces to be loaded synchronously in real time. Since the time lags in each test loading system are different, it is inevitable that the loading will not be synchronized during the loading process, and eventually may causing the test to fail.

The second is the hybrid experiment of model updating, in which one of the nonlinear components is taken as the test substructure, and the rest are taken as the numerical substructure. The difference is that the numerical nonlinear part and the test substructure have similar restoring force characteristics, and can be described by the same constitutive model, and the real-time loading feedback data of the test substructure is used to update its constitutive model online. A lot of research has been done in the model update method. However, due to the inaccuracy of the constitutive model, there are still data defects in the test during the test process, which cannot be conducive to subsequent accurate calculations.

In short, the problems existing in the current hybrid test are mainly reflected in the fact that it is difficult to guarantee the accuracy of the test by simplifying the model for numerical simulation; and dividing more test sub-domains in the overall research object, although the test can reflect the dynamic response of the structure more accurately, but the corresponding The cost, workload, and laboratory test environment requirements of the test are high, and higher requirements are put forward; the numerical model of the overall structure and the numerical model of the physical substructure are established based on the hybrid test of model update, although the constitutive parameters are continuously corrected online according to the test data. , but the problem of asynchronous time-delay is still unavoidable, and the results of numerical simulation inevitably have large errors.

Contents of the invention

In order to overcome the shortcomings of the existing technology, a real-time mixed test method based on single-specimen restart and multi-task loading is provided to solve the above problems.

A real-time hybrid test method based on restarting multi-task loading of a single specimen. The real-time hybrid test method is to load the calculated loading targets at multiple initial moments through a test specimen sequentially through a test loading system until the test substructures are uniform. After the loading is completed, feed back the test data measured after the loading to the numerical simulation system to calculate the real-time loading target and/or update the numerical test substructure model for accuracy comparison to obtain the truncated loading target, then real-time loading target and/or truncated loading target The target is sent to the test loading system, which loads the specimen in turn, resets the test object, until the process of judging the accuracy of the test object is completed.

As an optimal solution: through a test piece, the calculated multiple initial moment loading targets are sequentially loaded through the test loading system until the test substructures are all loaded. The process is to load multiple test substructures of the target at a single initial moment, by the same The set of loading devices sequentially load according to their respective loading targets to realize the process of multi-task loading.

As an optimal solution: the process of feeding back the test data measured after loading to the numerical simulation system to calculate the real-time loading target is a single loading target at the initial moment. The accuracy of the realization of the loading target. When the accuracy of the test substructure of the current step loading target does not meet the test requirements, reload the test substructure of the current step loading target that does not meet the test requirements. After loading, the test loading system And the reset of the test object, and so on, until all the test substructures meet the precision requirements.

As an optimal solution: the process of updating the numerical test substructure model for accuracy comparison to obtain the truncated loading target is a single initial moment loading target. After each step of loading is completed, the simulation results of the numerical model of the test substructure learned by the long short-term memory network are compared with the current step. The test results of the test substructure of the loading target are compared. When the test results of the test substructure of the loading target in the current step meet the accuracy requirements, the loading target is continuously truncated until it does not affect the calculation result of the reaction force at the current moment, and the truncated value is output. When the target is loaded, when the test results of the test substructure of the current loading target do not meet the accuracy requirements, the new test data is used to continue updating the model and then repeat the accuracy comparison process.

As a preferred solution: the realization process of the truncated loading target is to first establish a long-short-term memory network architecture with m long-short-term memory network layers and multiple fully connected layers, and then establish a numerical model of the test substructure through finite elements, input Different incentives get corresponding counterforce responses, with different incentives as input and counterforce responses under different incentives as output, the process of establishing a training data set to train the long-short-term memory network architecture.

As a preferred version: the concrete steps of the real-time mixing test method are as follows:

Step 1. At the initial moment, establish the numerical model of the numerical substructure of the prototype structure and the numerical model of the test substructure, and obtain the loading target at the initial moment through the numerical simulation system;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the test substructure according to the loading target of the first test substructure, resets the test loading system and the test object after loading, and then presses the loading target of the second test substructure on the test substructure Loading, reset the test loading system and test objects after loading, and so on until all required test substructures are loaded, and the test loading system and test objects are reset;

Step 4. Feedback the test data measured by the test loading system to the numerical simulation system, compare the simulation results of the numerical model of the test substructure with the test results and update the numerical model of the test substructure;

Step 5. When the comparison between the simulation results of the numerical model of the test substructure and the test results meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target is output. When the test substructure When the simulation results of the numerical model are compared with the test results and the accuracy requirements are not met, the model is continued to be updated with new test data;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all required test substructures are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure respectively. If it does not meet, reload the current step of the unqualified test substructure. After loading, reset the test loading system and the test object. And so on, until all the accuracy requirements are met;

Repeat steps four to eight until the test is complete.

As a preferred solution: the real-time mixing test method is a real-time mixing test based on restarting the multi-task loaded high-rise damping structure as a prototype structure, and the steps of the real-time mixing test are:

Step 1. Establish the finite element model of the seven-story frame structure. Assuming that the reaction force of the test substructure is 0 at the initial moment, the finite element model calculates the loading target of the 6 seismic isolation bearings under the external load and initial conditions;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the isolation support according to the loading target of the first test substructure, resets the test loading system and the test object after loading, and then loads the support according to the loading target of the second test substructure , reset the test loading system and test objects after loading, and so on until all required test substructures are loaded, and the test loading system and test objects are reset;

Step 4. Feedback the test data measured by the test loading system to the numerical simulation system, compare the simulation results of the numerical model of the test substructure with the test results and update the numerical model of the test substructure;

Step 5. When the comparison between the simulation results of the numerical model of the test substructure and the test results meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target is output. When the test substructure When the simulation results of the numerical model are compared with the test results and the accuracy requirements are not met, the model is continued to be updated with new test data;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all required test substructures are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure respectively. If it does not meet, reload the current step of the unqualified test substructure. After loading, reset the test loading system and the test object. And so on, until all the accuracy requirements are met;

Repeat steps four to eight until the test is complete.

As a preferred version: when the real-time mixing test method is based on the real-time mixing test of the viscous damper-bridge coupling of restarting multi-task loading, the concrete steps of the real-time mixing test method are:

Step 1. Establish the finite element model of the bridge, set the initial position of the viscous damper on the bridge, and obtain the initial state such as the deflection of the bridge at the initial moment through calculation. The finite element model calculates the total viscosity under the external load and initial conditions The loading target of the damper;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the test substructure according to the loading target of the first test substructure, resets the test loading system and the test object after loading, and then loads the test substructure according to the loading target of the second test substructure , reset the test loading system and test objects after loading, and so on until all required test substructures are loaded, and the test loading system and test objects are reset;

Step 4. Feedback the test data measured by the test loading system to the numerical simulation system, compare the simulation results of the numerical model of the test substructure with the test results and update the numerical model of the test substructure;

Step 5. When the comparison between the simulation results of the numerical model of the test substructure and the test results meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target is output. When the test substructure When the comparison between the simulation results of the numerical model and the test results does not meet the accuracy requirements, the new test data is used to continue to update the model;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all required test substructures are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure respectively. If it does not meet, reload the current step of the unqualified test substructure. After loading, reset the test loading system and the test object. And so on, until all the accuracy requirements are met;

Repeat steps four to eight until the test is complete.

The beneficial effects of the present invention are:

1. The present invention uses a test piece to load the calculated multiple initial moment loading targets sequentially through the test loading system until the test substructures are all loaded, and feeds back the test data measured after the loading to the numerical simulation system to calculate After the real-time loading target and/or update the numerical test substructure model for precision comparison to obtain the truncated loading target, send the real-time loading target and/or the truncated loading target to the test loading system, load the specimen in sequence, and reset the test object until the truncated loading target is completed. The process of judging the accuracy of the test object. The whole process of the present invention is a mixed test process to ensure the test accuracy and synchronization through a single test piece. It solves the problem of unavoidable loading of multiple specimens out of synchronization, which directly saves costs and reduces the complex conditions for the laboratory. Require.

2. The present invention proposes a real-time mixed test method based on restarting multi-task loading, and adopts a single test piece and restarting loading mode to replace loading of multiple test substructures, which reduces test costs and relieves pressure on laboratory equipment. It directly solves the unavoidable problem of asynchrony when multiple specimens are loaded simultaneously.

3. The present invention reduces unnecessary loading by setting up a numerical model of the test substructure and simulating the test process, thereby reducing unnecessary test costs and time of the test by truncating the loading target. Specifically, by truncating the loading target, by The method of model checking cuts off part of the loading target in the previous part, which greatly reduces the overall time cost of the test.

4. The numerical simulation system of the present invention is realized by using a finite element numerical simulation system, which reduces the error caused by the simplified model calculation of the numerical substructure.

5. The present invention adopts the restart loading mode, which directly excludes the time lag difference caused by the simultaneous synchronization of the numerical substructure and the experimental substructure under the finite element calculation, and breaks the requirement of the mixed test that the numerical substructure and the experimental substructure are within the prescribed The conventional processing method completed within a time step length of 10000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 were used were the conventional processing methods which completed within the time step, therefore had great restriction to the numerical substructure, because the calculation time of the numerical substructure was relatively short, required the model to be simplified enough, which had a great influence on the test accuracy. It directly solves the difficult synchronous loading process of the specimen.

6. The present invention is applicable to the fields of civil engineering, transportation, bridges, aerospace, machinery and other related fields.

Description of drawings

Fig. 1 is a schematic flow sheet of the present invention;

Fig. 2 is a schematic diagram of the principle of the relationship between m LSTM (long short-term memory network) layers and multiple fully connected layers in the truncated loading target;

Figure 3 is a schematic diagram of the structure of a single cell in the LSTM architecture when the loading target is truncated;

Fig. 4 is the working principle schematic diagram when the present invention is used in high-rise damping structure;

Fig. 5 is the working principle schematic diagram when the present invention is used in viscous damper-bridge coupling model;

Fig. 6 is the structural representation of viscous damper-high-rise frame model;

Fig. 7 is a comparison diagram of the curve formed by the mixing test data of the present invention and the curve formed by the standard theoretical value.

1-prototype structure; 2-numerical substructure; 3-experimental substructure; 4-actuator; 5-isolation support; 6-force sensor.

Detailed ways

Embodiments of the present invention are described below through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation modes, and various modifications or changes can be made to the details in this specification based on different viewpoints and applications without departing from the spirit of the present invention.

Specific embodiment 1: This embodiment is described in conjunction with Fig. 1 to Fig. 7. The real-time mixing test method described in this embodiment is to use a test piece to load the calculated multiple initial moment loading targets sequentially through the test loading system until the test The loading of substructure 3 is completed, and the test data measured after loading is fed back to the numerical simulation system to calculate the real-time loading target. The process of judging the accuracy of the test object.

Among them, the process of sequentially loading the calculated multiple initial time loading targets through the test loading system until the test substructures 3 are all loaded is to load multiple test substructures 3 of the target at a single initial time, and the same set of loading devices sequentially Load according to their respective loading targets to realize the process of multi-task loading.

The process of feeding back the test data measured after loading to the numerical simulation system to calculate the real-time loading target is a single initial moment loading target. After each step of loading is completed, the loading target of the current step is verified through the loading target of each test substructure 3 respectively. When the accuracy achieved by the test substructure 3 of the loading target of the current step does not meet the test requirements, reload the test substructure 3 of the loading target of the current step that does not meet the test requirements. After loading, the test loading system and The test object is reset, and so on, until all test substructures 3 meet the precision requirements.

The specific method adopted by the present invention to ensure the test accuracy is to judge the accuracy of the two data curves of the loading target of the current step and the loading target of the previous step, so as to ensure that the coincidence of the two curves can reach more than 98%, and then the next step can be carried out. Since the loading commands of the current step and the previous step are basically the same except for the last item, the reason for the error in the loading process of the two is mainly because the loading time lag of the actuator is not fixed, resulting in two basically the same The command deviated during loading. To solve this problem, time lag compensation can be added. The time lag compensation methods used include but are not limited to: polynomial extrapolation method, adaptive compensation method, etc., so as to ensure that the accuracy of the two loading targets can meet the above requirements.

Specific embodiment two: This embodiment is described in conjunction with Figures 1 to 7. The real-time hybrid test method described in this embodiment is to load the calculated multiple initial moment loading targets sequentially through a test loading system through a test piece until the test Substructure 3 has been loaded, and the test data measured after loading is fed back to the numerical simulation system to calculate and update the model of numerical test substructure 3 for accuracy comparison to obtain the truncated loading target, and then send the truncated loading target to the test loading system , sequentially load the test piece, reset the test object, until the process of judging the accuracy of the test object is completed.

Among them, the process of loading the calculated single initial time target through the test loading system for multiple times until all test substructures 3 are loaded is to load multiple test substructures 3 of the target at a single initial time, and the same set of loading devices sequentially Load according to their respective loading targets to realize the process of multi-task loading.

Among them, the process of updating the numerical model of the experimental substructure 3 and comparing the accuracy to obtain the truncated loading target is a single loading target at the initial moment. After each step of loading of the loading target, the simulation results of the numerical model of the experimental substructure 3 learned by using the long short-term memory network are compared with the current The comparison of the test results of the test substructure 3 of the step loading target, when the test result of the test substructure 3 of the current step loading target meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment and When the truncated loading target is output, when the test result of the test substructure 3 of the current step loading target does not meet the accuracy requirement, the new test data is used to continue to update the model and then repeat the accuracy comparison process.

Further, the realization process of the truncated loading target is to first establish a long-term short-term memory network architecture with m long-term short-term memory network layers and multiple fully connected layers, and then establish a numerical model of the experimental substructure 3 through finite elements, input Different incentives get corresponding counterforce responses, with different incentives as input and counterforce responses under different incentives as output, the process of establishing a training data set to train the long-short-term memory network architecture.

Specific embodiment three: This embodiment is described in conjunction with Fig. 1 to Fig. 5. The real-time mixing test method described in this embodiment is to use a test piece to load the calculated multiple initial moment loading targets sequentially through the test loading system until the test Substructure 3 has been loaded and processed in two ways in parallel. One way is to feed back the test data measured after the loading is completed to the numerical simulation system to calculate the real-time loading target. The obtained test data is fed back to the numerical simulation system by using the updated numerical test substructure 3 model for accuracy comparison to obtain the truncated loading target, and the real-time loading target and the truncated loading target are respectively obtained in two ways, and the real-time loading target and the truncated loading target are obtained Send it to the test loading system, load the test piece in turn, reset the test object, until the process of judging the accuracy of the test object is completed.

Furthermore, the process of loading the calculated single initial moment loading target through the test loading system for multiple times until all the test substructures 3 are loaded is to load multiple test substructures 3 of the target at a single initial moment, and the same set of loading device Load according to the respective loading targets in turn to realize the process of multi-task loading.

Further, the process of feeding back the test data measured after loading to the numerical simulation system to calculate the real-time loading target is a single loading target at the initial moment. After each step of loading is completed, the current step is verified through the loading target of each test substructure 3 in the previous step. Accuracy of the realization of the loading target. When the accuracy achieved by the test substructure 3 of the loading target of the current step does not meet the test requirements, reload the test substructure 3 of the loading target of the current step that does not meet the test requirements. After loading, the test substructure 3 The loading system and the test object are reset, and so on, until all the test substructures 3 meet the accuracy requirements.

Furthermore, the process of updating the numerical model of the experimental substructure 3 and comparing the accuracy to obtain the truncated loading target is a single loading target at the initial moment. After each step of loading is completed, the simulation results of the numerical model of the experimental substructure 3 learned by using the long short-term memory network are compared with Comparing the test results of the test substructure 3 of the loading target in the current step, when the test results of the test substructure 3 of the loading target in the current step meet the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment And until the truncated loading target is output, when the test result of the test substructure 3 of the current step loading target does not meet the accuracy requirement, the new test data is used to continue updating the model and then repeat the accuracy comparison process.

Further, the realization process of the truncated loading target is to first establish a long-term short-term memory network architecture with m long-term short-term memory network layers and multiple fully connected layers, and then establish a numerical model of the experimental substructure 3 through finite elements, input Different incentives get corresponding counterforce responses, with different incentives as input and counterforce responses under different incentives as output, the process of establishing a training data set to train the long-short-term memory network architecture.

Among them, the long-short-term memory network layer is the LSTM layer, and the long-short-term memory network architecture is the LSTM architecture,

As shown in Figure 2 and Figure 3, where D _n is the displacement input at the nth moment (the rest is the same), F _n is the reaction force output at the nth moment (the rest is the same), C ⁽¹⁾ _n-1 It is the influence of the internal state vector at the previous moment of the nth moment (which can be understood as the memory state vector Memory of LSTM) on the nth moment, h ⁽¹⁾ _n-1 is the output of the previous moment at the nth moment The influence of the vector on the nth moment, similarly C ⁽¹⁾ _n , h ⁽¹⁾ _n is the influence of the nth moment on the next moment.

Establish an LSTM architecture with m LSTM (long short-term memory network) layers and multiple fully connected layers (FC Layers), where the forget gate acts on the LSTM state vector from above to control the memory c _t-1 of the previous moment For the influence of the current moment, the control variable _gf of the forget gate is generated by formula (1):

g _f ＝σ(W _f [h _t-1 ,x _t ]+b _f ) (1)

由式(2)产生：The input gate is used to control the degree of acceptance of the input by the LSTM. First, the new input amount is obtained by linearly transforming the input x _t at the current moment and the output h _t-1 at the previous moment.

Produced by formula (2):

会通过输入门控制接受输入的量，输入门的控制变量同样来自输入x_t和输出h_t-1，见式(3)：but

The amount of input received will be controlled through the input gate, and the control variable of the input gate also comes from the input x _t and output h _t-1 , see formula (3):

g _i ＝σ(W _i [h _t-1 ,x _t ]+b _i ) (3)

Under the control of the forget gate and the input gate, the refresh method of the state vector c _t is shown in formula (4):

Inside the LSTM, the state vectors are not all output, but selectively output under the action of the output gate. The gating variable g _o of the output gate is generated by formula (5), and the output of LSTM is generated by formula (6).

g _o ＝σ(W _o [h _t-1 ,x _t ]+b _o ) (5)

h _t ＝g _o *tanh(c _t ) (6)

Among them, W _f and b _f are the parameter tensors of the forget gate, which are automatically optimized by the back propagation algorithm; σ is the activation function; W _c and b _c are the parameters of the input gate, which are automatically optimized by the back propagation algorithm; tanh is the activation function ; W _i and b _i are the parameters of the input gate, which are automatically optimized by the back propagation algorithm; W _o and b _o are the parameters of the input gate, which are automatically optimized by the back propagation algorithm.

Then, the numerical model of the experimental substructure 3 is established through finite elements, and different excitations are input to obtain the corresponding reaction force responses. Different excitations are used as inputs, and the reaction responses of different excitations are used as outputs. A training data set is established for this The LSTM architecture is trained, and after the training effect is achieved, the loading target of the current step of the test is input. The simulation results of the numerical model of the test substructure 3 are compared with the test results through the deep learning system. If the simulation results of the numerical model of the test substructure 3 are the same as When the test results are compared and the accuracy requirements are met, the loading target will be truncated until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target will be output, otherwise, the method of modifying the loading target will be abandoned and learning will continue with new data.

In this embodiment, after each test loading system is loaded, it is necessary to restore the test loading system to the initial state, and then carry out the next loading, repeating a cycle to reproduce multiple test substructures 3 with a single test substructure 3 The dynamic response will lay the foundation for the next experiment.

Embodiment 4: This embodiment is a further limitation of Embodiment 1, 2 or 3. In this embodiment, the numerical simulation system is implemented by a finite element numerical simulation system.

Embodiment 5: This embodiment is a further limitation of Embodiment 1, 2 or 3. The test loading system uses a high-performance electro-hydraulic servo loading system to complete real-time loading.

Embodiment 6: This embodiment is a further limitation of Embodiment 1, 2, 3, 4 or 5. This embodiment is to use the working principle of the real-time mixing test method, that is, the real-time mixing test method is through a The specimen loads the calculated loading targets at multiple initial moments sequentially through the test loading system until the loading of the test substructure 3 is completed. After the loading is completed, the measured test data is fed back to the numerical simulation system to calculate the real-time loading targets. The real-time loading target is sent to the test loading system, the test piece is loaded sequentially, the test object is reset, and the specific implementation steps of the process until the accuracy judgment of the test object is completed are as follows:

Step 1. At the initial moment, the numerical models of the numerical substructure 2 and the experimental substructure 3 of the prototype structure 1 are established, and the loading target at the initial moment is obtained through the numerical simulation system;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the test substructure 3 according to the loading target of the first test substructure, resets the test loading system and the test object after loading, and then loads the test substructure 3 according to the second test substructure Target loading, after loading, reset the test loading system and test objects, and so on, until all the required test substructures 3 are loaded, and reset the test loading system and test objects;

Step 4, feeding back the test data measured by the test loading system to the numerical simulation system, comparing the simulation results of the numerical model of the test substructure 3 with the test results and updating the numerical model of the test substructure 3;

Step 5. When the comparison between the simulation results of the numerical model of the test substructure 3 and the test results meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target is output. When the test substructure When the simulation results of the numerical model of structure 3 are compared with the test results and the accuracy requirements are not met, the new test data is used to continue to update the model;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all the required test substructures 3 are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure 3, and reload the current step of the unqualified test substructure 3 if it does not meet the requirements, and then load the system and test objects into the test after loading Reset, and so on, until all accuracy requirements are met;

Repeat steps four to eight until the test is complete.

Embodiment 7: As shown in Figure 1 and Figure 4, this embodiment is a further limitation of Embodiment 6. In this embodiment, when the prototype structure 1 is a high-rise damping structure, the real-time hybrid test method is based on weight Start the real-time hybrid test of the multi-task loading seismic isolation bearing 5 of the high-rise vibration-absorbing structure. The seat is used as the test substructure 3 for test loading, and the interactive data is used for mixed tests during the test. In the test, a single isolator 5 and the restart loading mode were used to replace the loading of all the isolators 5 .

The specific steps of the real-time hybrid test of the rubber isolation bearing of the high-rise vibration-damping structure based on restarting multi-task loading are as follows:

Step 1. Establish the finite element model of the seven-story frame structure. Assuming that the reaction force of the test substructure 3 is 0 at the initial moment, the finite element model calculates the loading target of the 6 seismic isolation supports 5 under the external load and the initial condition;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the isolation support 5 of the high-rise vibration-absorbing structure according to the loading target of the first test substructure. After loading, reset the test loading system and the test object, and then press the second test on the support The loading target of the substructure is loaded, and after loading, the test loading system and test objects are reset, and so on, until all the required test substructures are loaded, and the test loading system and test objects are reset;

Step 4. Feedback the test data measured by the test loading system to the numerical simulation system, compare the simulation results of the numerical model of the test substructure with the test results and update the numerical model of the test substructure;

Step 5. When the comparison between the simulation results of the numerical model of the test substructure and the test results meets the accuracy requirements, the loading target is continuously cut off until it does not affect the calculation result of the reaction force at the current moment, and the truncated loading target is output. When the test substructure When the simulation results of the numerical model are compared with the test results and the accuracy requirements are not met, the model is continued to be updated with new test data;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading and reset the specimen. Specifically, the sequential loading process of the specimen is realized through the cooperation of the actuator 4 acting on the isolation support 5 until The required test substructures are all loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure respectively. If it does not meet, reload the current step of the unqualified test substructure. After loading, reset the test loading system and the test object. And so on, until all the accuracy requirements are met;

Repeat steps four to eight until the test is complete.

In the real-time hybrid test, the seismic isolation bearing 5 is used for test loading, and the high-rise frame is used for numerical simulation. The key difficulty of the test process that the present invention solves is:

1) The present invention can accurately reproduce the dynamic response of the high-level framework;

2) If a plurality of seismic isolation bearings 5 are used for test loading in the present invention, multiple test loading systems and multiple test pieces will be required, which will undoubtedly increase the cost of the test, and will also impose higher requirements on the laboratory. Correspondingly, the present invention only uses one shock-isolation bearing 5 to directly avoid the occurrence of the above-mentioned situation;

3) The present invention does not need to use the test method of the mixed test of model updating, and saves the constitutive parameter according to the continuous online correction process according to the test data, so that the error of the numerical simulation result is reduced.

Embodiment 8: This embodiment is a further limitation of Embodiment 6. In the real-time hybrid test of the viscous damper-bridge coupling model, when the prototype structure 1 is a viscous damper, the viscous damper is used for testing Loading, the bridge structure is numerically simulated.

The key difficulty that the present invention solves in carrying out test process is:

1) The present invention can accurately reproduce the dynamic response of the viscous damper-bridge coupling model;

2) The present invention does not need to carry out the test loading of a plurality of viscous dampers, only one repeated loading is required;

3) The present invention does not need to use the test method of the mixed test of model updating, and saves the constitutive parameter according to the continuous online correction process according to the test data, so that the error of the numerical simulation result is reduced.

The schematic diagram of the viscous damper-bridge coupling model is shown in Fig. 5:

The bridge is taken as the numerical substructure 2 for numerical simulation, and all the viscous dampers are used as the experimental substructure 3 for test loading. During the test process, interactive data are used for mixed tests. In the test, a single viscous damper and a restart loading mode were used to replace the loading of all viscous dampers.

The real-time hybrid test method in this embodiment is based on the real-time hybrid test method of the viscous damper-bridge coupling model of restarting multi-task loading, which specifically includes the following steps:

Step 1. Establish the finite element model of the bridge, set the initial position of the viscous damper on the bridge, and obtain the initial state such as the deflection of the bridge at the initial moment through calculation. The finite element model calculates the total viscosity under the external load and initial conditions The loading target of the damper;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the test substructure 3 according to the loading target of the first test substructure 3, resets the test loading system and the test object after loading, and then presses the test substructure 3 according to the second test substructure 3 After loading, reset the test loading system and test objects, and so on, until all the required test substructures 3 are loaded, and reset the test loading system and test objects;

Step 4, feeding back the test data measured by the test loading system to the numerical simulation system, comparing the simulation results of the numerical model of the test substructure 3 with the test results and updating the numerical model of the test substructure 3;

Step 5. If the comparison between the simulation results of the numerical model of the test substructure 3 and the test results meets the accuracy requirements, then continuously truncate the loading target until it does not affect the calculation result of the reaction force at the current moment and output the truncated loading target, otherwise use the new The test data continues to update the model;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all the required test substructures 3 are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure 3, and reload the current step of the unqualified test substructure 3 if it does not meet the requirements, and then load the system and test objects into the test after loading Reset, and so on, until all accuracy requirements are met;

Repeat steps four to eight until the test is complete.

In this embodiment, a force sensor 6 is arranged between the viscous damper and the actuator 4 to transmit the load generated between them in real time.

Specific Embodiment 9: This embodiment is a further limitation of specific embodiments 1, 2, 3, 4, 5, 6 or 8. Through this embodiment, it is explained that the test accuracy of the present invention can meet the detailed process of corresponding requirements, specifically as follows The real-time hybrid test of the multi-viscous damper high-rise damping structure illustrates the basic principle and key technology of the method of the present invention. In order to effectively study the dynamic response of multi-story frames under the action of earthquakes and dampers, it is necessary to carry out experimental research on the overall structure. Building a real multi-story frame structure in the test and purchasing a corresponding number of viscous dampers will lead to high test costs. It is one of the efficient and economical means to study the high-rise damping structures with multi-viscous dampers by using real-time hybrid experiments.

In the real-time hybrid test of the multi-viscous damper high-rise damping structure, the viscous damper is used for test loading, and the high-rise frame is used for numerical simulation. The key difficulties in solving the test are:

1) This real-time hybrid test can accurately reproduce the dynamic response of the high-level framework;

2) The present invention does not need to carry out the test loading of a plurality of viscous dampers, only one repeated loading is required;

3) The present invention does not need to use the test method of the mixed test of model updating, and saves the constitutive parameters according to the continuous online correction process according to the test data, so that the error of the numerical simulation result is effectively reduced, thereby improving the accuracy.

The method of this embodiment performs a real-time hybrid test of a multi-viscous damper high-rise damping structure based on restarting multi-task loading, which specifically includes the following steps:

Step 1. Establish the finite element model of the three-story frame structure. Assuming that the reaction force of the test substructure is 0 at the initial moment, the finite element model calculates the loading targets of the three dampers under the external load and initial conditions;

Step 2. Send the loading target obtained in step 1 to the test loading system;

Step 3: The test loading system loads the test substructure according to the loading target of the first test substructure, resets the test loading system and the test object after loading, and then loads the test substructure according to the loading target of the second test substructure , reset the test loading system and test objects after loading, and so on until all required test substructures are loaded, and the test loading system and test objects are reset;

Step 4. Feedback the test data measured by the test loading system to the numerical simulation system, compare the simulation results of the numerical model of the test substructure with the test results and update the numerical model of the test substructure;

Step 5. If the comparison between the simulation results of the numerical model of the test substructure and the test results meets the accuracy requirements, then continuously truncate the loading target until it does not affect the calculation result of the reaction force at the current moment and output the truncated loading target, otherwise use a new test The data continues to update the model;

Step 6: Entering the next moment, the numerical simulation system combines the received test data to obtain the loading target at this moment;

Step 7. Send the received truncated loading target to the test loading system for sequential loading, and reset the specimen until all required test substructures are loaded;

Step 8. Verify the accuracy of the loading target of the current step through the loading target of the previous step of each test substructure respectively. If it does not meet, reload the current step of the test substructure that does not meet the test. After loading, reset the test loading system and the test object. And so on, until all the accuracy requirements are met;

Repeat steps four to eight until the test is complete.

The working condition used in the test in this embodiment is the El Centro seismic record, and the results of the displacement of the top floor in the viscous damper-high-rise frame model are shown in Figure 6: The main parameters of the layer frame structure are: m ₁ =400.06kg, m ₂ =355.30kg, m ₃ =357.50kg, k ₁ =k ₂ =k ₃ =586500N/m, c ₁₁ =413.64N·s/m, c ₁₂ =413.64N·s/m, c ₁₃ =-26.71N·s/m, c ₂₂ =367.41N·s/m, c ₂₃ =-139.42N·s/m, c ₃₃ =252.36N·s/m .

As shown in Figure 7, the curve formed by the data obtained after the loading test of the displacement of the top frame in this embodiment is compared with the standard theoretical value curve, and the red line is the curve formed by the data obtained after the loading test of the displacement of the top frame. The black line is the standard theoretical value curve, where only the black line is shown is where the black line and the red line do not overlap, where the red line is shown is where the black line and the red line overlap, and the area where the red line is located is the same as the area where the black line is The overlapping parts of the regions are all finally reflected in the red region. As can be seen from Figure 7, the overlapping degree of the curve formed by the data obtained after the loading test on the displacement of the top floor frame and the standard theoretical value curve reaches more than 98%, which shows that the present invention carries out the hybrid test method The accuracy can be guaranteed, the obtained data is stable and reliable, and can be used for subsequent use, which is conducive to the accurate derivation of subsequent effective calculation results.

Claims (7)

1. A real-time hybrid test method based on single test piece restarting multi-task loading is characterized in that: the real-time mixing test method comprises the following specific steps:

firstly, at an initial moment, establishing a numerical model of a numerical substructure (2) of a prototype structure (1) and a numerical model of a test substructure (3), and obtaining a loading target at the initial moment through a numerical simulation system;

step two, the loading target obtained in the step one is sent to a test loading system;

thirdly, loading the test sub-structure (3) according to the loading target of the first test sub-structure (3), resetting the test loading system and the test object after loading, loading the test sub-structure (3) according to the loading target of the second test sub-structure (3), resetting the test loading system and the test object after loading, and so on until all the required test sub-structures (3) are loaded, and resetting the test loading system and the test object;

step four, test data measured by a test loading system are fed back to a numerical simulation system, a simulation result of a numerical model of the test substructure (3) is compared with the test result, and the numerical model of the test substructure (3) is updated;

Step five, when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) meets the precision requirement, continuously cutting off the loading target until the calculation result of the counter force at the current moment is not influenced and outputting the cutting-off loading target, and when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) does not meet the precision requirement, continuously updating the model by using new test data;

step six, entering the next moment, and combining the received test data by the numerical simulation system to obtain a loading target at the moment;

step seven, the received cut-off loading targets are sent to a test loading system to be loaded in sequence, and the test piece is reset until all the required test substructures (3) are loaded;

step eight, verifying the accuracy of the realization of the current step of loading targets through the previous step of loading targets of each test substructure (3), reloading the current step of the test substructure (3) which is not met if the current step of loading targets are not met, resetting the test loading system and the test objects after loading, and the like until all the requirements on precision are met;

repeating the fourth step to the eighth step until the test is completed.

2. The real-time hybrid testing method based on single-test-piece restarting multi-task loading according to claim 1, wherein the method is characterized by comprising the following steps: the process of sequentially loading the calculated multiple initial time loading targets through the test loading system until the test substructures (3) are loaded is that the multiple test substructures (3) of the single initial time loading target are loaded by the same loading device according to the respective loading targets, so that the multi-task loading process is realized.

3. The real-time hybrid testing method based on single-test-piece restarting multi-task loading according to claim 1 or 2, wherein the method comprises the following steps: and feeding back the test data measured after the loading is finished to a numerical simulation system to calculate that the process of obtaining the real-time loading target is that after each step of loading of the loading target is finished at a single initial moment, the accuracy of the realization of the current step of loading target is verified through the previous step of loading target of each test substructure (3), when the accuracy of the realization of the test substructure (3) of the current step of loading target does not meet the test requirement, the test substructure (3) of the current step of loading target which does not meet the test requirement is reloaded, the test loading system and the test object are reset after the loading, and so on until all the test substructures (3) meet the precision requirement.

4. The real-time hybrid testing method based on single-test-piece restarting multi-task loading according to claim 1 or 2, wherein the method comprises the following steps: and after the process of obtaining the cut-off loading target after the precision comparison is carried out on the updated numerical test substructure (3) model, after each step of loading of the single initial moment loading target is completed, comparing the simulation result of the numerical model of the test substructure (3) learned by the long-short-term memory network with the test result of the test substructure (3) of the current step loading target, when the test result of the test substructure (3) of the current step loading target meets the precision requirement, continuously cutting off the loading target until the calculation result of the counter force at the current moment is not influenced and outputting the cut-off loading target, and when the test result of the test substructure (3) of the current step loading target does not meet the precision requirement, continuously updating the model by using new test data and then carrying out the precision comparison process again.

5. The real-time hybrid testing method based on single-test piece restarting multi-task loading according to claim 4, wherein the method is characterized by comprising the following steps: the implementation process of the cut-off loading target comprises the steps of firstly establishing a long-period memory network system structure with m long-period memory network layers and a plurality of full-connection layers, then establishing a numerical model of a test substructure (3) through finite elements, inputting different excitation actions to obtain corresponding counter-force responses, taking the different excitation actions as input, taking the counter-force responses of the different excitation actions as output, and establishing a training data set to train the long-period memory network system structure.

6. The real-time hybrid testing method based on single-test-piece restarting multi-task loading according to claim 1, wherein the method is characterized by comprising the following steps: the real-time hybrid test method is a real-time hybrid test based on a high-rise vibration reduction structure loaded by restarting a multitask as a prototype structure (1), and comprises the following steps of:

step one, establishing a finite element model of a seven-layer frame structure, and calculating to obtain loading targets of 6 shock insulation supports (5) under external load and initial conditions by assuming that counter forces of the test substructures (3) at initial moments are all 0;

Step two, the loading target obtained in the step one is sent to a test loading system;

thirdly, the test loading system loads the shock insulation support (5) according to the loading target of the first test substructure (3), resets the test loading system and the test object after loading, loads the support according to the loading target of the second test substructure (3), resets the test loading system and the test object after loading, and so on until all the required test substructure (3) is loaded, and resets the test loading system and the test object;

step four, test data measured by a test loading system are fed back to a numerical simulation system, a simulation result of a numerical model of the test substructure (3) is compared with the test result, and the numerical model of the test substructure (3) is updated;

step five, when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) meets the precision requirement, continuously cutting off the loading target until the calculation result of the counter force at the current moment is not influenced and outputting the cutting-off loading target, and when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) does not meet the precision requirement, continuously updating the model by using new test data;

Step six, entering the next moment, and combining the received test data by the numerical simulation system to obtain a loading target at the moment;

step seven, the received cut-off loading targets are sent to a test loading system to be loaded in sequence, and the test piece is reset until all the required test substructures (3) are loaded;

step eight, verifying the accuracy of the realization of the current step of loading targets through the previous step of loading targets of each test substructure (3), reloading the current step of the test substructure (3) which is not met if the current step of loading targets are not met, resetting the test loading system and the test objects after loading, and the like until all the requirements on precision are met;

repeating the fourth step to the eighth step until the test is completed.

7. The real-time hybrid test method based on the single-test-piece restarting multi-task loading according to claim 1, wherein the real-time hybrid test method is based on the restarting multi-task loading viscous damper-bridge coupling, and comprises the following specific steps:

step one, establishing a finite element model of a bridge, setting an initial position of a viscous damper on the bridge, and calculating to obtain initial states such as deflection of the bridge at an initial moment by calculation, wherein the finite element model calculates to obtain loading targets of all the viscous dampers under external load and initial conditions;

Step two, the loading target obtained in the step one is sent to a test loading system;

thirdly, the test loading system loads the test substructure (3) according to the loading target of the first test substructure (3), resets the test loading system and the test object after loading, loads the test substructure (3) according to the loading target of the second test substructure (3), resets the test loading system and the test object after loading, and so on until all the required test substructure (3) is loaded, and resets the test loading system and the test object;

step four, test data measured by a test loading system are fed back to a numerical simulation system, a simulation result of a numerical model of the test substructure (3) is compared with the test result, and the numerical model of the test substructure (3) is updated;

step five, when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) meets the precision requirement, continuously cutting off the loading target until the calculation result of the counter force at the current moment is not influenced and outputting the cutting-off loading target, and when the comparison of the simulation result and the test result of the numerical model of the test substructure (3) does not meet the precision requirement, continuously updating the model by using new test data;

Step six, entering the next moment, and combining the received test data by the numerical simulation system to obtain a loading target at the moment;

step seven, the received cut-off loading targets are sent to a test loading system to be loaded in sequence, and the test piece is reset until all the required test substructures (3) are loaded;

step eight, verifying the accuracy of the realization of the current step of loading targets through the previous step of loading targets of each test substructure (3), reloading the current step of the test substructure (3) which is not met if the current step of loading targets are not met, resetting the test loading system and the test objects after loading, and the like until all the requirements on precision are met;

repeating the fourth step to the eighth step until the test is completed.

Images