Automobile parts vibration test method and system based on H-infinity robust control

Abstract

The application provides a kind of automobile parts vibration test method and system based on H ∞ Robust Control.The method comprises: arranging sensors on a vibration table and a to-be-tested automobile part to synchronously collect actual state signals and strain signals; constructing a generalized controlled object comprising a vibration table dynamics model and each mapping channel; designing a multi-objective performance weighting function and converting it into an H ∞ norm optimization target, obtaining an optimal control strategy to configure an H ∞ Robust Controller by solving a linear matrix inequality; filtering and baseline correcting the collected signals, and based on the deviation of the reference waveform and the accurate feedback signal and the optimal control strategy, the H ∞ Robust Controller dynamically outputs a control quantity to control the vibration table to achieve high-precision waveform tracking.The application can maintain the adaptive adjustment capability and waveform fidelity of the test system under the working conditions with disturbance and parameter drift.

Description

Technical Field

[0001] This invention relates to the field of vibration testing and closed-loop control technology, and in particular to a vibration testing method and system for automotive components based on H∞ robust control. Background Technology

[0002] Automotive components typically undergo vibration testing during the research and development and production stages to evaluate their structural fatigue resistance and other physical properties. Existing conventional testing systems often use a vibration table to apply specific waveform excitation forces to the components under test, and collect data such as displacement and acceleration of the component response through a sensor network to form feedback control.

[0003] During actual operation, the nominal mass and stiffness of the vibration table may be slightly adjusted due to changes in the test load. At the same time, the complex test environment is often accompanied by external objective disturbances such as foundation vibration and fluctuations in the equipment power supply network. In addition, the inherent thermal noise and quantization error of the sensor components themselves will have a certain objective impact on the purity of the acquisition and feedback link.

[0004] Furthermore, some automotive parts made of viscoelastic materials experience hysteretic energy loss due to internal friction under continuous alternating vibration loads, leading to heat generation and consequently changes in dynamic stiffness. Long-cycle alternating vibration can also cause the adhesive interface of strain sensors attached to the surface of the test component to loosen and detach. These external multi-source physical disturbances and the dynamic drift of the test object's own parameters present certain technical challenges to maintaining stable waveform tracking control of the testing equipment. Summary of the Invention

[0005] The purpose of this invention is to provide a vibration testing method and system for automotive components based on H∞ robust control, in order to solve the problems mentioned in the background art.

[0006] This invention provides a vibration testing method for automotive components based on H∞ robust control, comprising the following steps: S1. Fix the automotive parts to be tested on a vibration table, and arrange sensors on the vibration table and the automotive parts to be tested respectively, so as to synchronously collect the actual state signals reflecting the operating status of the equipment and the strain signals reflecting the response of the parts. S2. Set a reference waveform that matches the actual working conditions of the automotive parts under test, and construct a generalized controlled object for the physical properties of the vibration table; wherein, the generalized controlled object specifically includes: a vibration table dynamic model based on the correction parameters of the initial acquired data, and an external disturbance channel, a sensor noise channel, and a control input weighting channel for constraint drive module overload that are mapped in parallel with the vibration table dynamic model; S3. Design a multi-objective performance weighting function that works in conjunction with the generalized controlled object, specifically including: a low-frequency tracking weighting function with low-pass filtering characteristics to reduce low-frequency tracking lag; a high-frequency noise suppression weighting function with high-pass filtering characteristics to suppress waveform distortion caused by high-frequency interference; and a control amplitude limiting weighting function; and transform the multi-objective performance weighting function into an H∞ norm optimization objective, and obtain the optimal control strategy by solving linear matrix inequalities to configure an H∞ robust controller; S4. During the closed-loop test, the real-time acquired actual state signal and strain signal are simultaneously subjected to a digital low-pass filter with a preset cutoff frequency to filter out external high-frequency interference, and baseline correction is performed to remove signal drift. Subsequently, the denoised and corrected accurate feedback signal is directly input into the H∞ robust controller, so that it forms a closed-loop collaborative suppression mechanism with the high-frequency noise suppression weighting function. Based on the deviation between the reference waveform and the accurate feedback signal and the optimal control strategy, the H∞ robust controller dynamically outputs control quantities to the drive module to control the vibration table to achieve high-precision waveform tracking.

[0007] Optionally, in step S3, the step of obtaining the optimal control strategy by solving linear matrix inequalities specifically includes: Under the boundary conditions of uncertainty, external disturbance and control constraints of the vibration table dynamic model, the minimization problem that satisfies the H∞ norm optimization objective is solved in real time, and the optimal control quantity is extracted and sent to the H∞ robust controller to achieve dynamic adaptive adjustment in response to load changes within the control cycle.

[0008] Optionally, step S4 may further include a sensor response discrimination step based on multi-source temporal correlation: Within a set sliding time window, the displacement time sequence in the actual state signal and the strain time sequence in the strain signal are extracted synchronously. The cross-correlation function between the displacement time series and the strain time series is calculated, and the delay time corresponding to the maximum cross-correlation coefficient is extracted to calculate the dynamic phase difference characteristic value at the current excitation frequency.

[0009] Optional, The sensor response discrimination step also includes: The discrete dynamic phase difference characteristic values ​​are continuously recorded, and their phase difference change rate with respect to time is calculated; When the phase difference change rate exceeds the preset non-continuous step boundary threshold within the set continuous sampling period, and the corresponding cross-correlation coefficient decreases by more than a set proportion, it is determined that a non-dynamic constraint abrupt change has occurred and it is confirmed that the corresponding sensor adhesive interface has delamination state, thereby generating a first state signal. When generating the first state signal, the H∞ robust controller performs extreme value penalty amplification on the performance weighting factor of the information feedback channel corresponding to the abnormal sensor in the generalized controlled object matrix, so as to force the optimization solver to minimize the feedback gain of the failed channel, and re-solve the linear matrix inequality based on the updated generalized controlled object matrix to output a new optimal control quantity.

[0010] Optionally, step S4 may further include a feedforward adjustment step for the drift of dynamic parameters of the automotive component under test: During the closed-loop vibration test, the instantaneous displacement data stream within each complete vibration cycle is extracted, and the corresponding instantaneous force data stream is extracted based on the acceleration of the vibration table and the set equivalent mass. The discrete numerical integration algorithm is used to calculate the hysteresis energy loss in a single cycle. The calculation logic is to obtain the sum of the products of the instantaneous force data stream and the corresponding displacement increment at each sampling point in a single cycle.

[0011] Optionally, the feedforward adjustment step further includes: The hysteresis energy loss over multiple vibration cycles is accumulated and calculated to generate a cumulative heat accumulation index. When the cumulative heat accumulation index reaches or exceeds a preset first critical threshold, it is determined that the dynamic stiffness of the automotive component under test has decayed, and a second state signal is generated.

[0012] Optionally, the feedforward adjustment step further includes: In response to the second state signal, the corresponding stiffness reduction coefficient is obtained according to the preset energy-stiffness reduction mapping table; In the dynamic model of the vibration table-tested automotive component coupled system in the generalized controlled object, the nominal stiffness coefficient submatrix and damping coefficient submatrix corresponding to the body dimension of the tested automotive component are multiplied by the corresponding stiffness reduction coefficients to generate the updated dynamic model. The H∞ robust controller re-solves for the optimal control quantity that satisfies the H∞ norm optimization objective based on the updated dynamic model, so as to achieve feedforward pre-compensation of the control quantity by reducing the stiffness parameter setpoint in advance. Optionally, in step S1, the actual state signal is collected by an acceleration sensor and a displacement sensor fixed at the center of the vibration table surface, and the strain signal is collected by a strain sensor attached to the key test position of the automotive component under test. In step S4, before performing the digital low-pass filtering, the step of amplifying the weak original acquisition signal by a signal amplification unit is included, and the cutoff frequency of the digital low-pass filtering is configured to be higher than the highest frequency of the reference waveform to completely filter out inherent sensor noise.

[0013] Optionally, the method further includes the following auxiliary processes: Throughout the closed-loop vibration test, the operating parameters of the vibration table and the response signals of the automotive parts under test are monitored synchronously in real time. When the load, displacement, or strain exceeds the preset safety threshold, the emergency stop unit is triggered and the power supply to the vibration table and drive module is cut off. After the test, the measured transfer function is compared with the nominal model in the generalized controlled object. Based on the vibration fatigue cumulative damage theory, the fatigue life of the tested automotive parts is calculated and a standard test report is generated.

[0014] In a second aspect, the present invention provides a vibration testing system for automotive components based on H∞ robust control, for implementing the method described in any one of the first aspects, the system comprising: The host computer monitoring module is used to issue test commands and configure reference waveform parameters; the reference waveform setting unit is connected to the host computer monitoring module and is used to output preset sinusoidal sweep frequency, random vibration or impact vibration reference waveforms. The H∞ robust control module has its reference input terminal connected to the output terminal of the reference waveform setting unit, and its internal processor integrates a linear matrix inequality solving unit and a performance weighting function design unit. The vibration execution and drive module includes a drive control module whose input terminal is connected to the output terminal of the H∞ robust control module, and a vibration table electrically connected to the output terminal of the drive control module for fixing the automotive parts to be tested. The closed-loop feedback and signal processing module includes a signal acquisition module arranged on the vibration table and the surface of the automotive parts under test, and a filtering preprocessing module connected in series thereafter; the output end of the signal acquisition module is connected to the input end of the filtering preprocessing module, and the output end of the filtering preprocessing module is connected to the feedback input end of the H∞ robust control module, so as to form a closed-loop control loop that transmits accurate feedback signals to the controller.

[0015] The present invention has achieved the following beneficial effects: This invention constructs a generalized controlled object that includes a mapping channel between model uncertainty and external environmental disturbances, and combines it with an H∞ robust control algorithm based on linear matrix inequality solving. This enables stable closed-loop output control in test environments with objective sensor noise. The system utilizes a sensor response discrimination mechanism based on multi-source temporal correlation. When a non-dynamically constrained delamination mutation is detected at a certain measurement point, the system adaptively applies extreme value penalty amplification to the performance weighting factor of the abnormal feedback channel. This forces the optimization solver to minimize the feedback gain of the failed channel, effectively preventing single-local signal distortion from disrupting the overall vibration test process. Furthermore, by continuously calculating the hysteresis energy loss within a single cycle and obtaining the accumulated heat accumulation state of the tested component using a discrete numerical integration algorithm, the system can dynamically update the nominal stiffness coefficient corresponding to the body dimension of the tested automotive component within the dynamic model of the vibration table-tested automotive component coupling system based on an internal mapping table. This achieves feedforward pre-compensation for the drift of the tested component's own parameters, further ensuring the dynamic adaptive adjustment level and waveform fidelity capability of the test system during long-term alternating load cycles.

[0016] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 A flowchart of a vibration testing method for automotive components based on H∞ robust control is provided for an embodiment of the present invention. Figure 2 A block diagram illustrating the construction principle of a generalized controlled object provided in this embodiment of the invention; Figure 3 This is a flowchart of the multi-objective performance weighted and controller optimal strategy solution provided in an embodiment of the present invention; Figure 4 The following is a flowchart of sensor response discrimination and adaptive reconstruction control provided in an embodiment of the present invention; Figure 5 This is a flowchart of feedforward adjustment for dynamic parameter drift of automotive parts under test, provided in an embodiment of the present invention. Figure 6This invention provides a hardware architecture and connection diagram for an automotive component vibration testing system based on H∞ robust control, as shown in the embodiment of the invention. Detailed Implementation

[0019] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] This application provides a vibration testing method and system for automotive components based on H∞ robust control. It is applied to a closed-loop control architecture testing system to form a complete testing link of setting, driving, acquisition, processing, feedback, and adjustment.

[0021] This application discloses a vibration testing method for automotive components based on H∞ robust control, such as... Figure 1 As shown, the method specifically includes the following steps: S1. Fix the automotive component to be tested on a vibration table, and arrange sensors on the vibration table and the automotive component to be tested respectively, so as to synchronously collect the actual status signal reflecting the operating status of the equipment and the strain signal reflecting the response of the component.

[0022] Specifically, the automotive component under test is placed above the vibration table and fixed using an adjustable mechanical clamp. The adjustable mechanical clamp is fixed to the preset mounting holes on the vibration table surface using fasteners, and the clamping position and clamping force parameters are adjusted to achieve a set standard value, thus constraining the automotive component under test within the clamp. After mechanical fixing, the sensor hardware is arranged. An accelerometer and a displacement sensor are positioned at the center of the vibration table surface. The spindle direction of the accelerometer is aligned with the excitation direction of the vibration table, used to acquire the acceleration signal of the vibration table surface in real time; the measuring probe or measuring optical path of the displacement sensor is aligned with the reference measuring surface of the table surface, used to acquire the displacement signal of the vibration table surface in real time. The above acceleration and displacement signals together constitute the actual state signal reflecting the operating status of the equipment. Simultaneously, after surface treatment of the surface of the automotive component under test at the preset test position, a strain sensor is attached to acquire the surface strain signal of the component body during the excited vibration process in real time. Accelerometers, displacement sensors, and strain sensors are all physically and electrically connected to the input port of an external filtering and preprocessing module via shielded signal transmission cables. The sampling clocks of all sensors are connected to the same synchronous clock source to ensure strict synchronization of the actual state signal and strain signal in the time domain.

[0023] like Figure 2As shown, S2, a reference waveform matching the actual working conditions of the automotive parts under test is set, and a generalized controlled object for the physical properties of the vibration table is constructed; wherein, the generalized controlled object specifically includes: a vibration table dynamic model based on the correction parameters of the initial acquisition data, and an external disturbance channel, a sensor noise channel, and a control input weighting channel for constraint drive module overload that are mapped in parallel with the vibration table dynamic model.

[0024] Specifically, the system obtains preset vibration test parameters through the reference waveform setting unit and generates corresponding reference waveform signals. These reference waveform signals are discrete-time data sequences, configured as one or a combination of sinusoidal swept-frequency waves, random vibration waves, or impact vibration waves. Next, a generalized controlled object matrix is ​​constructed in the controller's processing unit. This generalized controlled object is an augmented state-space model containing the parameters of the controlled physical entity and its associated disturbance factors.

[0025] Further, the steps for constructing the generalized controlled object include: establishing a shaking table dynamic model. The system pre-stores a nominal dynamic model in the form of second-order ordinary differential equations characterizing the mechanical properties of the shaking table foundation, which includes a nominal mass matrix, a nominal stiffness matrix, and a nominal damping matrix. During the system initialization phase, a low-amplitude broadband excitation signal is applied to the shaking table, initial state data from the sensors are collected, and a system identification algorithm is used to estimate and correct the parameters of each matrix in the nominal dynamic model, resulting in a shaking table dynamic model with corrected parameters based on the initial acquired data. In this dynamic model, a parameter uncertainty perturbation matrix characterizing the variation range of mass, stiffness, and damping parameters is introduced to cover model errors caused by load changes.

[0026] Specifically, the boundary range of the parameter uncertainty perturbation matrix is ​​strictly obtained through physical limit condition calibration: for the vibration table system, a basic sinusoidal frequency sweep excitation is applied under two extreme physical boundary conditions: no load (only the basic test fixture is installed) and full load (the fixture and the automotive parts under test with the largest mass are installed), and the corresponding mass boundary intervals are extracted. and stiffness boundary interval .

[0027] The exact physical definitions and calculation methods for each parameter and boundary in the formula are as follows: and These represent the absolute lower and upper limits of the equivalent dynamic mass obtained by the system identification under the aforementioned no-load and full-load conditions, respectively, in units of [unit missing]. ; and These represent the measured lower and upper limits of the system's equivalent dynamic stiffness during calibration testing, corresponding to the differences in fixture assembly tolerances and the materials of the softest / hardest components. The units are... ; The nominal mass and nominal stiffness of the nominal dynamic model are set to the arithmetic mean of the corresponding intervals mentioned above, and the ratio of the interval half-width to its mean is used as the normalized bounded norm constraint (e.g., mass perturbation norm bound) of the corresponding elements in the uncertainty perturbation matrix. ).

[0028] The symbol derivation is defined as follows: Represents the perturbation matrix of the mass parameters of The norm (i.e., the maximum singular value) is used in control theory models to quantitatively characterize the maximum dynamic range of the deviation between the actual mass and the nominal mass of a system. The formula for deriving the upper bound of the normalized mass perturbation is as follows: Similarly, the upper limit of stiffness perturbation. .

[0029] This calibration step transforms the abstract model error norm into a bounded value that directly maps to the actual bench load weight and fixture stiffness, ensuring that the mathematically feasible domain for solving the linear matrix inequality (LMI) is completely encompassed by the real physical limit.

[0030] Furthermore, in the augmented state equation of the generalized controlled object, an external disturbance channel, a sensor noise channel, and a control input weighting channel are established in parallel with the shaking table dynamics model. The external disturbance channel adds an independent disturbance input vector to the input of the state equation to map the additional forces generated by ground foundation vibration and equipment power supply network fluctuations on the system state derivative term. The sensor noise channel adds a measurement noise vector to the observation output of the state equation to map the thermal noise of the sensor elements themselves and the quantization error during analog-to-digital conversion. The control input weighting channel connects a gain transfer matrix in series at the control input of the state equation to set the physical saturation threshold of the control input signal, thereby constraining the maximum amplitude of the control command at the model level and preventing the drive control module and the shaking table electromagnetic coil from operating under overload conditions. The above model, perturbation matrix, and various mapping channels together form the standard transfer function matrix form of the generalized controlled object.

[0031] like Figure 3As shown in step S3, design a multi-objective performance weighting function that works in conjunction with the generalized controlled object. Specifically, this includes: a low-frequency tracking weighting function with low-pass filtering characteristics to reduce low-frequency tracking lag; a high-frequency noise suppression weighting function with high-pass filtering characteristics to suppress waveform distortion caused by high-frequency interference; and a control amplitude limiting weighting function. The multi-objective performance weighting function is then transformed into an H∞ norm optimization objective, and the optimal control strategy is obtained by solving linear matrix inequalities to configure an H∞ robust controller.

[0032] Specifically, to achieve multivariable frequency domain shaping control, the system is configured with three sets of performance weighting function matrices in the frequency domain. The low-frequency tracking weighting function is connected in series at the output of the system tracking error signal. Its transfer function structure is configured as a first-order or higher-order low-pass filter, which has the amplitude-frequency characteristics of providing high gain in the low-frequency band and rapidly attenuating gain in the high-frequency band. It is used to amplify and penalize the system's low-frequency tracking deviation, thereby forcibly reducing the waveform tracking error and phase lag under low-frequency conditions.

[0033] Specifically, the low-frequency band tracking weighting function adopts the standard first-order transfer function form, and its mathematical expression is as follows: .

[0034] The specific physical definitions and values ​​of each parameter in the formula are as follows: The Laplace operator (complex frequency variable) is used to characterize the dynamic transfer characteristics of the weighting function in the continuous complex frequency domain; The dimensionless coefficient representing the maximum allowable steady-state waveform tracking error of the system is directly adopted from the steady-state error threshold required in the vibration testing standards and specifications for automotive components (e.g., set to...). or That is, respectively corresponding to or (error margin) The desired minimum tracking bandwidth of the closed-loop system, in units of The specific calculation formula is as follows: ,in Set the highest effective operating frequency of the reference waveform in step S2 (in units of...). ),coefficient The value range is configured as follows to This parameter setting mechanism is designed to ensure that the controller has high gain and attenuation-free tracking capability throughout the entire effective operating frequency band; This is a dimensionless high-frequency gain limiting constant. Its function is to constrain the excitation amplitude of high-frequency control quantities outside the effective test frequency band. To prevent high-frequency amplification from causing overcurrent saturation of the vibration table's electromagnetic coil, its value range is strictly limited. to .

[0035] Through the aforementioned explicit parameter solution and mapping rules, this invention completely transforms the abstract frequency domain weight parameters in H∞ robust control into deterministic engineering indicators that are strongly correlated with the device waveform fidelity and the actual test frequency band.

[0036] A high-frequency noise suppression weighting function is connected in series at the observation output of the generalized controlled object. Its transfer function structure is configured as a high-pass filter, and its cutoff frequency is configured to be higher than the highest effective frequency set by the reference waveform. It is used to penalize high-frequency measurement noise signals in the feedback link, reducing the system's sensitivity in the high-frequency band. A control quantity amplitude limiting weighting function is connected in series at the control output of the controller. It is a constant gain matrix or a dynamic matrix with low-pass characteristics, used to penalize and limit the amplitude of the controller's output voltage or current.

[0037] Furthermore, after obtaining the aforementioned performance weighting functions, a fractional linear transformation (LFT) is performed on the low-frequency tracking weighting function, the high-frequency noise suppression weighting function, the control amplitude limiting weighting function, and the transfer function matrix of the generalized controlled object to construct the augmented weighted open-loop transfer function matrix of the closed-loop system. Subsequently, the system's multiple performance requirements are transformed into a standard H∞ norm optimization objective, namely, finding a controller that minimizes the H∞ norm of the closed-loop transfer function matrix from external disturbance input to evaluation signal output.

[0038] It is understood that in step S3, the step of obtaining the optimal control strategy by solving linear matrix inequalities specifically includes: based on the state space realization of the augmented weighted open-loop transfer function matrix, and based on the bounded real lemma, the H∞ norm constraint is equivalently transformed into a linear matrix inequality (LMI) system containing a series of symmetric positive definite matrix variables. Under the boundary conditions of the uncertainty bounds of the shaking table dynamics model, the norm bounds of the external disturbance vectors, and the control quantity constraint thresholds, the processor inside the controller calls the interior-point solver to perform real-time numerical solution of the linear matrix inequality system. After obtaining the optimal symmetric positive definite matrix solution that satisfies the H∞ norm optimization objective, the state space matrix parameters of the H∞ robust controller are solved by matrix algebra operations. The optimal control quantity is extracted, issued, and overwritten into the running memory of the H∞ robust controller to achieve dynamic adaptive adjustment in response to load changes within continuous control cycles.

[0039] S4. During the closed-loop test, the real-time acquired actual state signal and strain signal are simultaneously subjected to a digital low-pass filter with a preset cutoff frequency to filter out external high-frequency interference, and baseline correction is performed to remove signal drift. Subsequently, the denoised and corrected accurate feedback signal is directly input into the H∞ robust controller, so that it forms a closed-loop collaborative suppression mechanism with the high-frequency noise suppression weighting function. Based on the deviation between the reference waveform and the accurate feedback signal and the optimal control strategy, the H∞ robust controller dynamically outputs control quantities to the drive module to control the vibration table to achieve high-precision waveform tracking.

[0040] Specifically, during the vibration table operation phase, the system continuously acquires the raw time-domain signal streams from each sensor at a set sampling frequency. Since some raw signals have low amplitudes, before performing the digital low-pass filtering, the system amplifies the weak raw acquisition signals through a signal amplification unit, adjusting the signal amplitude to the optimal input range of the analog-to-digital converter. Subsequently, the discrete digital signal after analog-to-digital conversion is input to the digital low-pass filtering algorithm module. The cutoff frequency of this digital low-pass filter is configured to be higher than the highest frequency component of the reference waveform, ensuring that signal components within the effective test frequency band are retained while filtering out sensor inherent noise and high-frequency mechanical interference components above this cutoff frequency. After filtering, the signal data stream enters the baseline correction unit, which uses a polynomial fitting or moving average algorithm to calculate the low-frequency slowly varying trend term of the signal sequence. This trend term is then subtracted from the filtered signal to remove the DC bias component caused by temperature changes or sensor zero-point drift, ultimately outputting a denoised and corrected accurate feedback signal.

[0041] Furthermore, the precise feedback signal is directly input to the feedback input port of the H∞ robust controller via the data bus. Within the controller logic, since a high-frequency noise suppression weighting function has been introduced into the algorithm framework, when the precise feedback signal is used as a state observation input for control law calculation, its residual high-frequency fluctuations will be further penalized and suppressed in the frequency domain by the high-frequency noise suppression weighting function. This forms a closed-loop collaborative suppression mechanism against high-frequency interference in both the physical hardware filtering process and the algorithm's frequency domain penalty.

[0042] Furthermore, within each control cycle, the H∞ robust controller performs differential calculations on the reference waveform signal at the current moment and the actual acquired precise feedback signal to obtain a tracking deviation vector. The controller inputs this tracking deviation vector and the system's current state variables into the state-space equation of the optimal control strategy obtained according to step S3, performs matrix multiplication and addition operations, and dynamically calculates the output control quantity for the current control cycle. This output control quantity is transmitted as a digital signal to a digital-to-analog converter, converted into an analog voltage or current signal, and sent to the drive control module. The power amplifier inside the drive control module modulates the AC output according to this control quantity, driving the electromagnetic exciter of the vibration table to generate the corresponding mechanical excitation force, controlling the vibration table surface to move according to the trajectory of the reference waveform, thereby achieving waveform tracking.

[0043] Furthermore, in vibration testing of automotive components, the connection status of strain sensors affects the accuracy of the feedback link. Therefore, this application also includes a sensor response discrimination step based on multi-source temporal correlation in step S4.

[0044] like Figure 4 As shown, specifically, the sensor response discrimination step includes: during system operation, setting a fixed-length sliding time window in memory. As the sampling clock advances, within this sliding time window, synchronously extracting the displacement time series data from the pre-processed actual state signal and the strain time series data from the strain signal. Due to the inherent delay between displacement excitation and structural strain response based on the mechanical system's transmission characteristics, the controller calculates the discrete cross-correlation function of the displacement time series and the strain time series. By traversing the set time delay vector, calculating the cross-correlation coefficient sequence under different delay times, and searching for the peak value in this sequence. Extracting the delay time corresponding to the maximum cross-correlation coefficient, and combining it with the current excitation frequency or fundamental period parameter, calculating the dynamic phase difference characteristic value between displacement and strain under the current operating condition and frequency.

[0045] Based on this, the sensor response discrimination step further includes: the system establishes a first-in-first-out (FIFO) queue and continuously records the dynamic phase difference characteristic values ​​that evolve discretely over time. The phase difference change rate with respect to time is calculated by performing a first-order forward difference operation on the dynamic phase difference characteristic values ​​of adjacent sampling periods. The system internally presets continuous evolution boundary thresholds and discontinuous step boundary thresholds. During the test, the system continuously performs conditional judgments on the phase difference change rate. When the phase difference change rate exceeds the preset discontinuous step boundary threshold within the set continuous sampling period, and the value of the corresponding maximum cross-correlation coefficient calculated in the above steps decreases by more than a set proportion relative to the cross-correlation coefficient in the previous time window due to signal anomalies, it indicates that the strain signal has lost its physical correlation with the displacement signal. At this time, the system determines that a non-dynamic constraint abrupt change has occurred, confirms that the strain sensor bonding interface at the corresponding location has delamination or connection failure, and generates a first state signal in the control logic accordingly.

[0046] Furthermore, when the first state signal is generated, to ensure the test is not interrupted, the H∞ robust controller performs adaptive reconstruction processing. The controller performs extreme value penalty amplification processing on the performance weighting factor of the information feedback observation channel corresponding to the abnormal strain sensor in the generalized controlled object matrix. Specifically, the specific acquisition and setting method of the discontinuous step boundary threshold is as follows: during the calibration stage when the test is initialized and the sensor is confirmed to be properly attached, the vibration table is controlled to output a low-amplitude frequency broadband excitation, and the reference fluctuation peaks of the phase difference change rate within multiple sliding time windows under this normal excited state are recorded. The maximum reference fluctuation peak is extracted, and it is multiplied by a margin coefficient of 1.5 to 2.0 as the discontinuous step boundary threshold for judging sensor abnormality; if the decrease in the cross-correlation coefficient exceeds the set proportion in the set proportion, its physical value range is limited to 25% to 30%. Furthermore, the specific execution logic of the extreme value penalty amplification processing of the performance weighting factor of the abnormal feedback channel in the generalized controlled object matrix is ​​as follows: the nominal gain coefficient of the high-frequency noise suppression weighting function of the observation channel corresponding to the abnormal strain sensor is used. Real-time updates Among them, the extreme value penalty amplification factor The range of values ​​is set to to This penalty amplification process, without altering the fundamental dimensions of the state space matrix of the generalized controlled object, makes the abnormal channel... The objective function is considered to contain significant observation noise, forcing the controller to actively compress the elements of the feedback control gain matrix of the corresponding channel to near zero during the solver optimization process. This allows for matrix reconstruction within an extremely short control cycle and complete isolation of feedback signal distortion caused by localized delamination. After parameter updates, the controller processor uses the updated generalized controlled object matrix to re-invoke the linear matrix inequality solver within an extremely short control cycle, outputting a new optimal control strategy and optimal control quantity. This enables the control system to continue vibration testing based on main feedback signals such as acceleration and displacement, ignoring feedback interference from the delamination sensor.

[0047] Furthermore, in response to the heating and dynamic stiffness reduction phenomena that occur in non-metallic automotive parts such as viscoelastic materials when subjected to continuous alternating vibration loads, this application also includes a feedforward adjustment step for the drift of dynamic parameters of the automotive parts under test in step S4.

[0048] like Figure 5 As shown, specifically, during the closed-loop vibration test, the system extracts data segments for each complete mechanical vibration cycle through time-domain zero-crossing detection. Within this complete vibration cycle, the instantaneous displacement data stream of each sampling point is extracted. Simultaneously, the system retrieves the synchronously acquired vibration table acceleration data sequence, and according to Newton's second law, multiplies the acceleration data of each sampling point with the equivalent mass parameter set in the initialization phase to calculate the instantaneous force data stream acting on the components.

[0049] Subsequently, a discrete numerical integration algorithm is used to calculate the hysteresis energy loss within a single cycle. The calculation logic is as follows: within the time scale of a single cycle, the difference between the displacement value of each sampling point and the displacement value of the previous sampling point is calculated to obtain the corresponding displacement increment. The instantaneous force data stream value at the current sampling point is multiplied by this displacement increment to obtain the infinitesimal mechanical work within a single step. The infinitesimal mechanical work of all sampling points within a single cycle is summed, and the resulting scalar value is the energy dissipated by the components due to internal friction and hysteresis characteristics within the complete vibration cycle, i.e., the hysteresis energy loss.

[0050] Furthermore, the feedforward adjustment step also includes: the system sets up an accumulation register to continuously accumulate the scalar value of the hysteresis energy loss over multiple consecutive vibration cycles, which increases over time, to generate a cumulative heat accumulation index characterizing the degree of heat accumulation inside the component. The controller compares this cumulative heat accumulation index with a preset first critical threshold in real time. When the cumulative heat accumulation index reaches or exceeds the preset first critical threshold, the system determines that the dynamic stiffness of the tested automotive component has significantly decreased due to continuous heat absorption, and then generates a second state signal in the system.

[0051] Furthermore, the feedforward adjustment step further includes: in response to the second state signal, the system accesses a preset energy-stiffness reduction mapping table in the storage unit. This mapping table records a one-to-one correspondence between the heat accumulation index range and the stiffness reduction coefficient. The system queries the mapping table based on the current accumulated heat accumulation index value to obtain the corresponding stiffness reduction coefficient. This coefficient is typically a positive real number less than 1.

[0052] Further details the specific physical definition and offline acquisition steps of the first critical threshold and energy-stiffness reduction mapping table: Before the formal closed-loop vibration test, samples of automotive parts from the same batch are selected for offline temperature-dynamic stiffness correlation calibration tests. Under simulated alternating loads, the cumulative hysteresis energy loss due to internal friction within the material is continuously recorded, and the measured dynamic stiffness value is simultaneously measured using a dynamic tensile-compression testing machine. The decrease in measured dynamic stiffness of the parts compared to the initial nominal stiffness is then calculated. The cumulative hysteresis energy loss node corresponding to the time is defined and fixed as the first critical threshold.

[0053] The energy-stiffness reduction mapping table is a one-dimensional discrete lookup table model generated based on the aforementioned offline calibration data. The table contains multiple sets of cumulative heat accumulation index nodes that increase with a fixed energy step size, and stiffness reduction coefficients corresponding to each node. .in, The value is equal to the measured dynamic stiffness corresponding to the current node divided by the nominal stiffness under the initial environmental conditions. In the real-time feedforward adjustment of the closed-loop test, in order to prevent the controller stiffness parameter from changing abruptly due to discrete table lookup, the system further adopts a first-order linear interpolation algorithm. Based on the currently calculated cumulative heat accumulation index, interpolation calculation is performed between two adjacent nodes in the mapping table to smoothly output a high-precision stiffness reduction coefficient, thereby ensuring the stability and waveform fidelity of the control output under feedforward compensation.

[0054] After obtaining the coefficients, the system performs scalar multiplication operations on the nominal stiffness coefficient sub-matrix items and nominal damping coefficient sub-matrix items, which specifically characterize the load dimension of the automotive component under test, in the vibration table-under-test vehicle component coupling system based on physical degree of freedom decoupling in the generalized controlled object matrix, with the corresponding stiffness reduction coefficients, to generate a dynamic model matrix with updated parameters. During this process, the mechanical stiffness matrix elements of the vibration table body remain nominally locked, thereby avoiding deformation of the basic characteristics of the closed-loop control model. Before issuing the next control command, the H∞ robust controller re-executes the solution calculation process in step S3 based on the updated dynamic model to solve for a new optimal control quantity that satisfies the H∞ norm optimization objective. By reducing the set value of the stiffness parameter in advance at the algorithm model end, the system performs feedforward pre-compensation for the response deviation that will be caused by the softening of the tested component, suppressing the waveform tracking steady-state error caused by parameter drift.

[0055] Furthermore, the method also includes the following auxiliary process: Throughout the closed-loop vibration test, the operating parameters of the vibration table (including but not limited to driving voltage, driving current, peak feedback acceleration, and peak feedback displacement) and the strain response signals of the automotive components under test are simultaneously monitored in real time. The comparison module performs a high-frequency periodic comparison between the real-time monitored parameter values ​​and the system-configured safety threshold vector. When any of the load parameters, displacement parameters, or strain parameters exceeds the preset safety threshold limit, the system hardware interrupt mechanism immediately triggers the emergency stop unit, sending a disconnect signal to the contactor or solid-state relay to cut off the power supply to the vibration table and the drive module. After the test cycle reaches the preset duration or the test command is completed, the system gradually reduces the control quantity according to the descending ramp curve until it stops. Subsequently, the controller extracts the measured signal data stored throughout the test process and performs a Fourier transform to calculate the measured transfer function. The measured transfer function is then compared with the frequency domain characteristics of the nominal model in the initialized generalized controlled object. Based on the theory of cumulative damage from vibration fatigue, the rainflow counting algorithm is used to statistically analyze the amplitude of alternating load and the number of cycles during the test period. Combined with the SN fatigue curve of the material, the cumulative fatigue damage and predicted fatigue life of the automotive parts under test are calculated, and a standard test report containing time history, frequency response and life prediction data is output.

[0056] Corresponding to the above method embodiments, this application also provides an automotive component vibration testing system based on H∞ robust control, used to implement the above steps S1 to S4 and all subordinate method steps.

[0057] like Figure 6 As shown, specifically, the system includes the following hardware and its connection structure: The host computer monitoring module is used to issue test commands and configure reference waveform parameters. This module is an industrial computer with a human-machine interface, responsible for integrating the test parameters input by the operator and distributing initialization and running commands to various subsystems.

[0058] The reference waveform setting unit, which exists as a software process or an independent microcontroller, is connected to the host computer monitoring module via an internal communication bus. This unit is configured with various waveform generation algorithms to output preset discrete sequences of reference waveforms in digital form, such as sinusoidal frequency sweeps, random vibrations, or impact vibrations, based on parameter configurations.

[0059] The H∞ robust control module includes a high-performance microprocessor or digital signal processing device. Its reference input port is communicatively connected to the data output port of the reference waveform setting unit to receive a reference waveform signal. The processor's memory and logic units integrate a construction unit for analyzing the generalized controlled object, a linear matrix inequality solving unit for performing matrix operations, and a performance weighting function design unit for configuring frequency domain filtering characteristics.

[0060] The vibration execution and drive module provides mechanical excitation energy. This module includes a drive control module and a vibration table. The low-voltage input terminal of the drive control module is connected to the control output terminal of the H∞ robust control module via a digital-to-analog converter interface, used to receive and amplify the voltage control commands output by the controller; the drive control module internally contains a power electronic inverter amplifier circuit. The electromagnetic drive winding of the vibration table is electrically connected to the high-voltage output terminal of the drive control module via a power cable. The vibration table surface is equipped with tooling fixtures and fasteners for fixing the automotive parts under test.

[0061] A closed-loop feedback and signal processing module is used to acquire and clean multi-source signals. This module includes a signal acquisition module (containing an accelerometer, displacement sensor, and strain sensor) arranged on the vibration table surface and the surface of the automotive parts under test, and a filtering and preprocessing module electrically connected in series with it. The analog signal output terminals of each sensor are connected to the multi-channel analog input terminals of the filtering and preprocessing module via shielded cables. The filtering and preprocessing module integrates a programmable gain amplifier, an anti-aliasing low-pass filter, an analog-to-digital converter, and a digital signal processing chip that performs digital low-pass filtering and baseline correction. The digital bus output terminal of the filtering and preprocessing module is directly connected to the feedback input port of the H∞ robust control module, thereby forming a complete and high-precision closed-loop control loop from physical actuators, sensor network, signal cleaning circuit to controller logic operation.

[0062] Through the coordinated operation of the above-mentioned components and logical steps, the system disclosed in this application embodiment can complete the task of high-precision and robust closed-loop vibration testing of automotive parts purely through objective physical modeling, matrix solving, multi-source temporal correlation identification, and energy integral-based thermal feedforward compensation algorithm.

[0063] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A vibration testing method for automotive components based on H∞ robust control, characterized in that, Includes the following steps: S1. Fix the automotive parts to be tested on a vibration table, and arrange sensors on the vibration table and the automotive parts to be tested respectively, so as to synchronously collect the actual state signals reflecting the operating status of the equipment and the strain signals reflecting the response of the parts. S2. Set a reference waveform that matches the actual working conditions of the automotive parts under test, and construct a generalized controlled object for the physical properties of the vibration table; wherein, the generalized controlled object specifically includes: a vibration table dynamic model based on the correction parameters of the initial acquired data, and an external disturbance channel, a sensor noise channel, and a control input weighting channel for constraint drive module overload that are mapped in parallel with the vibration table dynamic model; S3. Design a multi-objective performance weighting function that works in conjunction with the generalized controlled object, specifically including: a low-frequency tracking weighting function with low-pass filtering characteristics to reduce low-frequency tracking lag; a high-frequency noise suppression weighting function with high-pass filtering characteristics to suppress waveform distortion caused by high-frequency interference; and a control amplitude limiting weighting function; and transform the multi-objective performance weighting function into an H∞ norm optimization objective, and obtain the optimal control strategy by solving linear matrix inequalities to configure an H∞ robust controller; S4. During the closed-loop test, the real-time acquired actual state signal and strain signal are simultaneously subjected to a digital low-pass filter with a preset cutoff frequency to filter out external high-frequency interference, and baseline correction is performed to remove signal drift. Subsequently, the denoised and corrected accurate feedback signal is directly input into the H∞ robust controller, so that it forms a closed-loop collaborative suppression mechanism with the high-frequency noise suppression weighting function. Based on the deviation between the reference waveform and the accurate feedback signal and the optimal control strategy, the H∞ robust controller dynamically outputs control quantities to the drive module to control the vibration table to achieve high-precision waveform tracking.

2. The vibration testing method for automotive components based on H∞ robust control according to claim 1, characterized in that, In step S3, the step of obtaining the optimal control strategy by solving linear matrix inequalities specifically includes: Under the boundary conditions of uncertainty, external disturbance and control constraints of the vibration table dynamic model, the minimization problem that satisfies the H∞ norm optimization objective is solved in real time, and the optimal control quantity is extracted and sent to the H∞ robust controller to achieve dynamic adaptive adjustment in response to load changes within the control cycle.

3. The vibration testing method for automotive components based on H∞ robust control according to claim 1, characterized in that, Step S4 further includes a sensor response discrimination step based on multi-source temporal correlation: Within a set sliding time window, the displacement time sequence in the actual state signal and the strain time sequence in the strain signal are extracted synchronously. The cross-correlation function between the displacement time series and the strain time series is calculated, and the delay time corresponding to the maximum cross-correlation coefficient is extracted to calculate the dynamic phase difference characteristic value at the current excitation frequency.

4. The vibration testing method for automotive components based on H∞ robust control according to claim 3, characterized in that, The sensor response discrimination step also includes: The discrete dynamic phase difference characteristic values ​​are continuously recorded, and their phase difference change rate with respect to time is calculated. When the phase difference change rate exceeds the preset non-continuous step boundary threshold within the set continuous sampling period, and the corresponding cross-correlation coefficient decreases by more than a set proportion, it is determined that a non-dynamic constraint abrupt change has occurred and it is confirmed that the corresponding sensor adhesive interface has delamination state, thereby generating a first state signal. When generating the first state signal, the H∞ robust controller performs extreme value penalty amplification on the performance weighting factor of the information feedback channel corresponding to the abnormal sensor in the generalized controlled object matrix, so as to force the optimization solver to minimize the feedback gain of the failed channel, and re-solve the linear matrix inequality based on the updated generalized controlled object matrix to output a new optimal control quantity.

5. The vibration testing method for automotive components based on H∞ robust control according to claim 1, characterized in that, Step S4 further includes a feedforward adjustment step for the drift of dynamic parameters of the automotive component under test: During the closed-loop vibration test, the instantaneous displacement data stream within each complete vibration cycle is extracted, and the corresponding instantaneous force data stream is extracted based on the acceleration of the vibration table and the set equivalent mass. The discrete numerical integration algorithm is used to calculate the hysteresis energy loss in a single cycle. The calculation logic is to obtain the sum of the products of the instantaneous force data stream and the corresponding displacement increment at each sampling point in a single cycle.

6. The vibration testing method for automotive components based on H∞ robust control according to claim 5, characterized in that, The feedforward adjustment step further includes: The hysteresis energy loss over multiple vibration cycles is accumulated and calculated to generate a cumulative heat accumulation index. When the cumulative heat accumulation index reaches or exceeds a preset first critical threshold, it is determined that the dynamic stiffness of the automotive component under test has decayed, and a second state signal is generated.

7. The vibration testing method for automotive components based on H∞ robust control according to claim 6, characterized in that, The feedforward adjustment step further includes: In response to the second state signal, the corresponding stiffness reduction coefficient is obtained according to the preset energy-stiffness reduction mapping table; In the dynamic model of the vibration table-tested automotive component coupled system in the generalized controlled object, the nominal stiffness coefficient submatrix and damping coefficient submatrix corresponding to the body dimension of the tested automotive component are multiplied by the corresponding stiffness reduction coefficients to generate the updated dynamic model. The H∞ robust controller re-solves for the optimal control quantity that satisfies the H∞ norm optimization objective based on the updated dynamic model, so as to achieve feedforward pre-compensation of the control quantity by reducing the stiffness parameter setpoint in advance.

8. The vibration testing method for automotive components based on H∞ robust control according to claim 1, characterized in that, In step S1, the actual state signal is collected by an acceleration sensor and a displacement sensor fixed at the center of the vibration table surface, and the strain signal is collected by a strain sensor attached to the key test position of the automotive component under test. In step S4, before performing the digital low-pass filtering, the step of amplifying the weak original acquisition signal by a signal amplification unit is included, and the cutoff frequency of the digital low-pass filtering is configured to be higher than the highest frequency of the reference waveform to completely filter out inherent sensor noise.

9. The vibration testing method for automotive components based on H∞ robust control according to claim 1, characterized in that, The method also includes the following auxiliary processes: Throughout the closed-loop vibration test, the operating parameters of the vibration table and the response signals of the automotive parts under test are monitored synchronously in real time. When the load, displacement, or strain exceeds the preset safety threshold, the emergency stop unit is triggered and the power supply to the vibration table and drive module is cut off. After the test, the measured transfer function is compared with the nominal model in the generalized controlled object. Based on the vibration fatigue cumulative damage theory, the fatigue life of the tested automotive parts is calculated and a standard test report is generated.

10. A vibration testing system for automotive components based on H∞ robust control, characterized in that, The system for implementing the method as described in any one of claims 1 to 9 comprises: The host computer monitoring module is used to issue test commands and configure reference waveform parameters; the reference waveform setting unit is connected to the host computer monitoring module and is used to output preset sinusoidal sweep frequency, random vibration or impact vibration reference waveforms. The H∞ robust control module has its reference input terminal connected to the output terminal of the reference waveform setting unit, and its internal processor integrates a linear matrix inequality solving unit and a performance weighting function design unit. The vibration execution and drive module includes a drive control module whose input terminal is connected to the output terminal of the H∞ robust control module, and a vibration table electrically connected to the output terminal of the drive control module for fixing the automotive parts to be tested. The closed-loop feedback and signal processing module includes a signal acquisition module arranged on the vibration table and the surface of the automotive parts under test, and a filtering preprocessing module connected in series thereafter; the output end of the signal acquisition module is connected to the input end of the filtering preprocessing module, and the output end of the filtering preprocessing module is connected to the feedback input end of the H∞ robust control module, so as to form a closed-loop control loop that transmits accurate feedback signals to the controller.

Images