A sensor-based automatic adjustment method and system for precast component vibrators

Abstract

This invention relates to the field of vibration testing technology, and discloses a sensor-based automatic adjustment method and system for precast component vibrators. The method includes real-time acquisition of acoustic signals and vibration time-domain data, and extraction of characteristic parameters, including: skewness-kurtosis joint entropy, peak value of normalized time-delay cross-correlation function, and box-count fractal dimension. The characteristic parameters are combined into a real-time characteristic state vector, and a target characteristic state vector is preset. The deviation between the real-time characteristic state vector and the target characteristic state vector is calculated, and the frequency and amplitude of the vibrator are adjusted according to the deviation. When the time change rate of the real-time characteristic state vector is lower than a first threshold, and the norm of the deviation is greater than a second threshold, a non-periodic perturbation sequence is applied to the vibrator. When the norm of the deviation remains lower than the second threshold for a preset time, the vibrator is stopped. This invention, based on a closed-loop control strategy of multi-dimensional information fusion and a special perturbation mechanism, avoids quality defects caused by under-vibration or over-vibration.

Description

Technical Field

[0001] This invention relates to the field of vibration testing technology, specifically to a sensor-based automatic adjustment method and system for precast component vibrators. Background Technology

[0002] In the production of precast concrete components, compaction is a crucial step in ensuring component quality. High-frequency vibration reduces the friction and viscosity between particles within the concrete mixture, allowing aggregates to rearrange under gravity and expelling trapped air bubbles. This process aims to improve the density and homogeneity of the concrete, thereby ensuring the final strength, durability, and surface quality of the component. However, current vibration operations in precast component plants largely rely on the experience of the operators. Workers typically judge the completion of vibration subjectively by observing macroscopic phenomena such as surface slurry return and whether air bubbles have stopped emerging, combined with listening to changes in the sound of the vibrating equipment. This judgment standard is not only vague but also varies from person to person, easily leading to under-vibration or over-vibration. Under-vibration can cause defects such as honeycomb and voids inside the component, severely affecting its mechanical properties and service life; while over-vibration may cause concrete segregation (i.e., coarse aggregates sinking and cement paste floating), which also damages the overall performance of the component.

[0003] Several other vibration control methods have been proposed in existing technologies. A relatively basic one is timed control, which applies vibration for a fixed duration to all components. This "one-size-fits-all" approach cannot adapt to performance fluctuations in different batches of concrete mixes, changes in environmental temperature and humidity, or differences in mold geometry, thus making it difficult to guarantee optimal vibration results. Other improved technologies attempt to achieve feedback control by monitoring a single physical quantity (such as the input current of the vibrator motor or the vibration signal from a single accelerometer on the mold). However, concrete compaction is a complex physical process, and information collected from a single point cannot fully reflect the overall characteristics of concrete as it transforms from a discrete aggregate of particles into a fluid plastic continuous medium, easily leading to under-vibration or over-vibration, and consequently, poor production quality of precast concrete components. Therefore, there is a need to design an automatic adjustment method for vibrators that can characterize the overall compaction state of concrete in multiple dimensions and in real time, and achieve intelligent closed-loop control. Summary of the Invention

[0004] This invention provides a sensor-based automatic adjustment method and system for precast component vibrators to solve the problem in the prior art that the vibration is easily under-vibrated or over-vibrated, resulting in poor production quality of precast concrete components.

[0005] In a first aspect, the sensor-based automatic adjustment method for precast component vibrators of the present invention includes the following steps:

[0006] Acoustic signals from the mold surface of the precast component and vibration time-domain data distributed at multiple spatial nodes on the mold are acquired in real time.

[0007] Based on acoustic signals and vibration time-domain data, feature parameters representing the compaction state of concrete are extracted. These feature parameters include: the skewness-kurtosis joint entropy calculated based on vibration time-domain data within a sliding time window; the peak value of the normalized time-delay cross-correlation function between the acoustic signal and vibration time-domain data of at least one of the spatial nodes; and the box-counting fractal dimension of the spatiotemporal state matrix constructed using vibration time-domain data from multiple spatial nodes.

[0008] The feature parameters are combined into a real-time feature state vector, and a target feature state vector representing the ideal compaction state is preset; the deviation between the real-time feature state vector and the target feature state vector is calculated, and the frequency and amplitude of the vibrator are adjusted in a closed loop according to the deviation;

[0009] When the time change rate of the real-time feature state vector is lower than the first threshold and the norm of the deviation is greater than the second threshold, an aperiodic perturbation sequence is generated and applied to the vibrator; when the norm of the deviation is continuously less than the second threshold within a preset time, it is determined that the vibration is completed and the vibrator is stopped.

[0010] Preferably, the real-time acquisition of acoustic signals from the mold surface of the precast component and vibration time-domain data distributed at multiple spatial nodes on the mold includes:

[0011] A capacitive microphone is installed on the surface of the geometric center of the mold to collect acoustic signals; piezoelectric accelerometers are installed at the four corners and the geometric center of the mold to collect vibration time-domain data.

[0012] Preferably, the skewness-kurtosis joint entropy calculated based on vibration time-domain data within a sliding time window includes:

[0013] Set the sliding time window width to 1 second and the step size to 0.1 seconds; within each time window, calculate the skewness value S and kurtosis value K of the vibration time-domain data of the spatial nodes at the geometric center of the mold; divide the skewness-kurtosis plane into... For a given grid, calculate the joint probability of sample points falling within each grid (i, j). Calculate the skewness-kurtosis joint entropy using the following formula. :

[0014] .

[0015] Preferably, the peak value of the normalized time-delay cross-correlation function between the acoustic signal and the vibration time-domain data of at least one of the spatial nodes includes:

[0016] acoustic signals Downsampling is performed to make the sampling frequency consistent with the sampling frequency of the vibration time-domain data, resulting in the downsampled acoustic signal. Normalized time-delay cross-correlation is then performed between the vibration time-domain data from the center of the mold and the downsampled acoustic signal to obtain the cross-correlation function value. The cross-correlation function value is calculated using the following formula:

[0017] ;

[0018] in, The time delay is At that time, the cross-correlation function value between the vibration time-domain data and the downsampled acoustic signal, For the downsampled acoustic signal in The value at time, The mean of the downsampled acoustic signal. The mean of the vibration time-domain data. For vibration time domain data in The value at time, The standard deviation of the downsampled acoustic signal. The standard deviation of the vibration time-domain data;

[0019] Set the maximum delay value to Its unit is milliseconds, obtained by traversing... turn up The maximum value, and The maximum value is taken as the peak value of the normalized time-delay cross-correlation function.

[0020] Preferably, the box-count fractal dimension of the spatiotemporal state matrix constructed using vibration time-domain data from multiple spatial nodes includes:

[0021] Using M data points collected within a 1-second time window as the time dimension and 5 spatial nodes distributed at the four corners and geometric center of the mold as the spatial dimension, a... A 3D spatiotemporal state matrix; using box counting, calculate the minimum number of boxes required to cover the trajectory of the matrix. With the side length of the box The relationship between box counts and fractal dimension Calculated using the following formula:

[0022] ;

[0023] Fitting by linear regression and The absolute value of the slope of the relationship line is obtained. .

[0024] Preferably, the step of adjusting the frequency and amplitude of the vibrator according to the deviation closed-loop includes:

[0025] A multivariable proportional-integral-derivative (MIDR) controller is used for adjustment; the deviation vector between the real-time characteristic state vector and the target characteristic state vector is used as the input of the MIDR controller, and the output is the control increment of frequency and amplitude.

[0026] Preferably, the frequency adjustment range is set to 50-150Hz, and the amplitude adjustment range is set to 0.5-2.0mm.

[0027] Preferably, generating and applying a non-periodic perturbation sequence to the vibrator includes:

[0028] Generating chaotic sequences using the Logistic mapping function The Logistic mapping function is:

[0029] ;

[0030] The non-periodic perturbation sequence is obtained by normalizing the chaotic sequence.

[0031] An aperiodic disturbance sequence is superimposed onto the frequency control signal output by a multivariable proportional-integral-derivative controller; the duration of the aperiodic disturbance sequence is 3 seconds, and the superimposed amplitude is 10% of the amplitude of the main control signal.

[0032] Preferably, the sampling frequency of the capacitive microphone is 44.1 kHz; and the sampling frequency of the piezoelectric accelerometer is 2 kHz.

[0033] Secondly, the sensor-based automatic adjustment system for precast vibrators of the present invention includes a memory and a processor. The memory stores computer instructions, and when the processor executes the computer instructions, it implements the aforementioned sensor-based automatic adjustment method for precast vibrators.

[0034] The beneficial effects of this invention are as follows: By fusing vibration time-domain data and surface acoustic signals from multiple spatial nodes on the precast mold, and extracting multi-dimensional feature parameters such as skewness-kurtosis joint entropy, cross-modal time-delay cross-correlation peak value, and spatiotemporal fractal dimension, a feature vector representing the overall compaction state of concrete is constructed. By monitoring the rate of change of the feature vector and the target deviation, potential compaction stagnation phenomena in local areas are identified and addressed. When the vibration process stalls, a non-periodic perturbation sequence is applied to disrupt the local equilibrium of the vibration process, guiding the concrete state towards the ideal compaction point. Based on a closed-loop control strategy using multi-dimensional information fusion and a special perturbation mechanism, the accuracy of judgment is improved, avoiding quality defects caused by under-vibration or over-vibration, thereby ensuring the high uniformity and stability of precast product quality. Attached Figure Description

[0035] Figure 1 This is a schematic flowchart of a sensor-based automatic adjustment method for precast component vibrators provided in an embodiment of the present invention. Detailed Implementation

[0036] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0037] like Figure 1 As shown, an embodiment of the sensor-based automatic adjustment method for precast component vibrators provided by the present invention specifically includes the following steps:

[0038] S1, real-time acquisition of acoustic signals from the mold surface of the precast component and vibration time-domain data distributed at multiple spatial nodes on the mold.

[0039] Specifically, five triaxial accelerometers are attached or magnetically fixed at key locations on the outer wall of the precast mold, such as the four corners and the center point, to collect vibration signals. Simultaneously, a high-fidelity microphone is placed approximately 0.5 meters above the mold to collect acoustic signals emitted by the concrete surface during vibration. Finally, a multi-channel synchronous data acquisition card is used to synchronously sample and convert the analog signals from all sensors at a sampling frequency of 20kHz, thereby obtaining a digitized vibration and acoustic time-domain data stream.

[0040] S2, based on acoustic signals and vibration time-domain data, extract feature parameters representing the compaction state of concrete. The feature parameters include: skewness-kurtosis joint entropy calculated based on vibration time-domain data within a sliding time window, peak value of the normalized time-delay cross-correlation function between the acoustic signal and vibration time-domain data of at least one of the spatial nodes, and box-count fractal dimension of the spatiotemporal state matrix constructed using vibration time-domain data from multiple spatial nodes.

[0041] Specifically, a sliding time window of 1 second is set, sliding once every 0.1 seconds. For the vibration data of the center point accelerometer within the time window, the skewness and kurtosis values ​​are calculated, and the skewness-kurtosis pairs of all sample points are treated as a two-dimensional random variable. The joint entropy of skewness-kurtosis is calculated by constructing a two-dimensional probability distribution histogram. Simultaneously, for the acoustic signal and the center point vibration signal within the same time window, the normalized time-delay cross-correlation function between them is calculated, and the maximum peak value of this function is extracted. The vibration data of the five accelerometers within the same time window are arranged into a matrix, where rows represent time sampling points and columns represent spatial nodes. Then, the box counting algorithm is applied to this matrix to calculate the box-count fractal dimension.

[0042] S3, combine the feature parameters into a real-time feature state vector, and preset a target feature state vector representing the ideal compaction state; calculate the deviation between the real-time feature state vector and the target feature state vector, and adjust the frequency and amplitude of the vibrator in a closed loop according to the deviation.

[0043] Specifically, the calculated skewness-kurtosis joint entropy, the peak value of the normalized time-delay cross-correlation function, and the box-count fractal dimension are combined into a three-dimensional real-time feature state vector. The target feature state vector is obtained through a pre-calibration experiment: a standard specimen is vibrated, and the moment when the specimen reaches its optimal compaction state is confirmed through manual experience and subsequent non-destructive testing of the component. The feature state vector corresponding to this moment is then recorded and stored in the control system as the target value.

[0044] The Euclidean distance between the real-time feature state vector and the target feature state vector is calculated to obtain a scalar deviation value. This deviation value is then input as an error signal into a PID controller. The PID controller performs calculations based on preset proportional, integral, and derivative parameters, outputting two control signals. These signals are used to adjust the frequency converter controlling the vibrator motor and the servo system controlling the eccentric block angle, respectively, thereby achieving continuous closed-loop regulation of the vibrator's vibration frequency and amplitude.

[0045] S4, when the time change rate of the real-time feature state vector is lower than the first threshold and the norm of the deviation is greater than the second threshold, an aperiodic perturbation sequence is generated and applied to the vibrator; when the norm of the deviation is continuously less than the second threshold within a preset time, it is determined that the vibration is completed and the vibrator is stopped.

[0046] Specifically, the norm of the difference between the current feature state vector and the feature state vector at the previous moment, i.e., the rate of change over time, is calculated in real time. If this rate of change is lower than a preset first threshold for a continuous period of time, for example, 0.01, while the norm of the current deviation is still greater than a preset second threshold, for example, 0.2, then it is determined that the vibration process has entered a local stagnation. At this time, a short-time vibration sequence composed of multiple combinations of different frequencies and amplitudes is generated, for example, vibrating at 60Hz and 80% amplitude for 0.5 seconds, and then vibrating at 85Hz and 95% amplitude for 0.3 seconds. This disturbance sequence instruction is sent to the vibrator controller for execution, thereby breaking the mechanical equilibrium inside the concrete.

[0047] The system continuously monitors the deviation range. When this range first falls below the second threshold of 0.2, a timer is started. If the deviation range remains below the second threshold of 0.2 for the next 5 seconds, the concrete is considered to have reached the ideal compaction state, and the vibration process is complete. At this point, the control system sends a stop command to the vibrator's frequency converter and servo system, ending the vibration operation.

[0048] In an optional embodiment, the real-time acquisition of acoustic signals from the mold surface of the precast component and vibration time-domain data distributed at multiple spatial nodes on the mold includes:

[0049] A capacitive microphone is installed on the surface of the geometric center of the mold to collect acoustic signals; piezoelectric accelerometers are installed at the four corners and the geometric center of the mold to collect vibration time-domain data.

[0050] The sampling frequency of the capacitive microphone is 44.1 kHz; the sampling frequency of the piezoelectric accelerometer is 2 kHz.

[0051] Specifically, a capacitive microphone is mounted on the geometric center surface of the mold to detect the airborne acoustic signals emitted by the concrete during vibration. To ensure signal fidelity and detail, the microphone's sampling frequency is set to a relatively high 44.1 kHz. Simultaneously, to comprehensively monitor the structural vibration state of the mold, five piezoelectric accelerometers are arranged on the mold surface. One sensor, like the microphone, is located at the geometric center of the mold, while the other four are mounted at the four corners. This distributed layout enables the detection of the spatial distribution characteristics of vibration. These piezoelectric accelerometers are used to acquire vibration time-domain data. The value of i ranges from 1 to 5, representing five different measurement point locations, and their sampling frequency is uniformly set to 2kHz.

[0052] In an optional embodiment, the skewness-kurtosis joint entropy calculated based on vibration time-domain data within a sliding time window includes:

[0053] Set the sliding time window width to 1 second and the step size to 0.1 seconds; within each time window, calculate the skewness value S and kurtosis value K of the vibration time-domain data of the spatial nodes at the geometric center of the mold; divide the skewness-kurtosis plane into... For a given grid, calculate the joint probability of sample points falling within each grid (i, j). Calculate the skewness-kurtosis joint entropy using the following formula. :

[0054] .

[0055] Specifically, a sliding time window with a width of 1 second is used to extract data segments. This time window slides forward in 0.1-second increments, thus achieving near real-time analysis of the vibration signal. Within each 1-second time window, only the vibration time-domain data collected by the sensor located at the geometric center of the mold is selected as the analysis object. Within each time window's data segment, two key statistical indicators are calculated: skewness S and kurtosis K. Skewness S represents the asymmetry of the data distribution, while kurtosis K represents the sharpness of the data distribution. These two values ​​constitute a two-dimensional coordinate point (S, K). As the time window slides, a series of such coordinate points are obtained.

[0056] To represent the complexity of the joint distribution of these two indices, the two-dimensional plane consisting of all S and K values ​​is divided into a rectangular grid of NS rows and NK columns. The joint probability is estimated by counting the number of coordinate points falling within each grid (i, j). Finally, using the formula for calculating information entropy, the skewness-kurtosis joint entropy H(S, K) is obtained.

[0057] In an optional embodiment, the peak value of the normalized time-delay cross-correlation function between the acoustic signal and the vibration time-domain data of at least one of the spatial nodes includes:

[0058] acoustic signals Downsampling is performed to make the sampling frequency consistent with the sampling frequency of the vibration time-domain data, resulting in the downsampled acoustic signal. Normalized time-delay cross-correlation is then performed between the vibration time-domain data from the center of the mold and the downsampled acoustic signal to obtain the cross-correlation function value. The cross-correlation function value is calculated using the following formula:

[0059] ;

[0060] in, The time delay is At that time, the cross-correlation function value between the vibration time-domain data and the downsampled acoustic signal, For the downsampled acoustic signal in The value at time, The mean of the downsampled acoustic signal. The mean of the vibration time-domain data. For vibration time domain data in The value at time, The standard deviation of the downsampled acoustic signal. The standard deviation of the vibration time-domain data;

[0061] Set the maximum delay value to Its unit is milliseconds, obtained by traversing... turn up The maximum value, and The maximum value is taken as the peak value of the normalized time-delay cross-correlation function.

[0062] Specifically, given the original acoustic signal The sampling frequency is 44.1 kHz, which is much higher than the 2 kHz of the vibration signal. Therefore, it is necessary to perform a downsampling operation on the acoustic signal to reduce the sampling frequency to 2 kHz, thereby obtaining a downsampled acoustic signal that is time-aligned with the vibration data. The vibration time-domain data v(t) from the center of the mold was selected as the reference signal and compared with the downsampled acoustic signal. Perform normalized time-delay cross-correlation calculation to generate a cross-correlation function. This function displays the similarity between two signals at different time delays τ, normalizing the value to a range of -1 to 1. The maximum value of the time delay is set. For 50ms, in The time at which the two signals best match is searched within the time delay range. The maximum value is found by calculating and comparing the cross-correlation function values ​​corresponding to all τ within this time delay range. This maximum value is taken as the peak value of the normalized time-delay cross-correlation function, which reflects the strongest linear correlation between the acoustic signal and the structural vibration.

[0063] In an optional embodiment, the box-counting fractal dimension of the spatiotemporal state matrix constructed using vibration time-domain data from multiple spatial nodes includes:

[0064] Using M data points collected within a 1-second time window as the time dimension and 5 spatial nodes distributed at the four corners and geometric center of the mold as the spatial dimension, a... A 3D spatiotemporal state matrix; using box counting, calculate the minimum number of boxes required to cover the trajectory of the matrix. With the side length of the box The relationship between box counts and fractal dimension Calculated using the following formula:

[0065] ;

[0066] Fitting by linear regression and The absolute value of the slope of the relationship line is obtained. .

[0067] Specifically, within a 1-second time window, each sensor collects M vibration data points. The data sets from the sensors located at the four corners and the center are merged within one second to form an M-row, 5-column spatiotemporal state matrix. Each row of this matrix represents the vibration state of all spatial points at a given moment, while each column represents the vibration sequence of a single spatial point over time. In the high-dimensional space formed by this M-row, 5-column matrix, each row can be considered a state point, and all points together constitute a state trajectory. A box-counting method is used, employing boxes with side length ε to cover the entire trajectory, and the minimum number of boxes required to completely cover the trajectory is calculated. By continuously reducing the side length of the box And by recounting, a series of results can be obtained about and The corresponding data. Theoretically, the box count fractal dimension. Is when When approaching zero, and The negative limit of the ratio. This invention is achieved by plotting in a double logarithmic coordinate system. right The graphical implementation. These data points typically exhibit a linear relationship, fitted to a straight line via linear regression, with the absolute value of the slope of that line serving as the box-count fractal dimension. The estimated value.

[0068] In an optional embodiment, adjusting the frequency and amplitude of the vibrator according to the deviation closed-loop includes:

[0069] A multivariable proportional-integral-derivative (MIDR) controller is used for adjustment. The deviation vector between the real-time characteristic state vector and the target characteristic state vector is used as the input of the MIDR controller, and the output is the control increment of frequency and amplitude. The frequency adjustment range is set to 50-150Hz, and the amplitude adjustment range is set to 0.5-2.0mm.

[0070] Specifically, the controller is designed to simultaneously manage the two key operating parameters of the vibrator: frequency and amplitude, to achieve closed-loop control of the compaction process. The controller's input is a real-time calculated deviation vector, obtained by subtracting the current real-time characteristic state vector extracted from sensor data from a preset target characteristic state vector representing the ideal compaction state. This deviation vector reflects the difference between the current state and the target state in multiple dimensions. The PID controller calculates two independent output signals based on the proportional, integral, and derivative terms of this deviation vector: the control increment for frequency and the control increment for amplitude. These control increments are sent to the vibrator's drive system to adjust the actual operating parameters. To ensure equipment safety and process stability, the adjustment process is limited to preset ranges: frequency adjustment is constrained between 50Hz and 150Hz, while the vibration amplitude adjustment range is set between 0.5mm and 2.0mm. The controller continuously executes this adjustment cycle until the deviation vector decreases to an acceptable range.

[0071] In an optional embodiment, generating and applying a non-periodic perturbation sequence to the vibrator includes:

[0072] Generating chaotic sequences using the Logistic mapping function The Logistic mapping function is:

[0073] ;

[0074] The non-periodic perturbation sequence is obtained by normalizing the chaotic sequence.

[0075] The aperiodic disturbance sequence is superimposed on the frequency control signal output by the multivariable proportional-integral-derivative controller; the duration of each applied aperiodic disturbance sequence is 3 seconds, and the superimposed amplitude is 10% of the amplitude of the main control signal.

[0076] The intensity of the non-periodic disturbance sequence, i.e. the superimposed amplitude, is set to 10% of the amplitude of the main frequency control signal. Brief and small-amplitude non-periodic disturbances can help the vibration system break out of its existing local optimum, promote the rearrangement of particles inside the concrete, and thus achieve a more uniform compaction effect.

[0077] The implementation principle of the sensor-based automatic adjustment method for precast concrete vibrators in this invention is as follows: This invention integrates the vibration time-domain data and surface acoustic signals from multiple spatial nodes on the precast concrete mold, and extracts multi-dimensional feature parameters such as skewness-kurtosis joint entropy, cross-modal time-delay cross-correlation peak value, and spatiotemporal fractal dimension to construct a feature vector that can represent the overall compaction state of concrete in real time. By monitoring the rate of change of this feature vector and its deviation from the target state, the "compaction stagnation" phenomenon that may occur in local areas can be identified and handled. When the vibration process is detected to be stalled, a non-periodic perturbation sequence is actively applied to break the local equilibrium of the vibration process and guide the concrete state to converge towards the ideal compaction point. This closed-loop control strategy based on multi-dimensional information fusion and a special perturbation mechanism significantly improves the accuracy of judgment, effectively avoids quality defects caused by under-vibration or over-vibration, and thus ensures the high uniformity and stability of precast concrete product quality.

[0078] An embodiment of the sensor-based automatic adjustment system for precast component vibrators provided by the present invention includes a memory and a processor. The memory stores computer instructions, and when the processor executes the computer instructions, it implements the sensor-based automatic adjustment method for precast component vibrators in the above embodiment.

[0079] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A sensor-based method of automatic adjustment of a preform vibrator, characterized in that, The method comprises the following steps: collecting the acoustic signals of the mold surface of the prefabricated part and the vibration time domain data of multiple spatial nodes distributed on the mold in real time; Based on the acoustic signals and the vibration time domain data, the characteristic parameters representing the compactness of the concrete are extracted, including: the skewness-kurtosis joint entropy calculated based on the vibration time domain data within a sliding time window, the normalized time-lag cross-correlation function peak value between the acoustic signals and the vibration time domain data of at least one spatial node, and the box-counting fractal dimension of the space-time state matrix constructed using the vibration time domain data of multiple spatial nodes; The characteristic parameters are combined into a real-time characteristic state vector, and a target characteristic state vector representing an ideal compactness state is preset; the deviation between the real-time characteristic state vector and the target characteristic state vector is calculated, and the frequency and amplitude of the vibrator are adjusted in a closed loop according to the deviation; When the time variation rate of the real-time characteristic state vector is lower than a first threshold value, and the norm of the deviation is greater than a second threshold value, a non-periodic disturbance sequence is generated and applied to the vibrator; when the norm of the deviation is continuously less than the second threshold value within a preset time, it is determined that the vibration is completed and the vibrator is stopped.

2. The method of claim 1, wherein, The real-time collection of the acoustic signals of the mold surface of the prefabricated part and the vibration time domain data of multiple spatial nodes distributed on the mold comprises: A condenser microphone is installed on the geometric center surface of the mold to collect acoustic signals; piezoelectric acceleration sensors are installed on the four corners and the geometric center of the mold to collect vibration time domain data.

3. The method of claim 1, wherein, The skewness-kurtosis joint entropy calculated based on the vibration time domain data within a sliding time window comprises: The width of the sliding time window is set to 1 second, and the step is 0.1 second; in each time window, the skewness value S and the kurtosis value K of the vibration time domain data of the spatial node of the geometric center position of the mold are calculated; the skewness-kurtosis plane is divided into a grid, and the joint probability of the sample points falling in each grid (i, j) is counted ; the skewness-kurtosis joint entropy is calculated according to the following formula : Skewness-kurtosis joint entropy = -∑P(i,j)log(P(i,j)) : 。 4. The method of claim 1, wherein, The normalized time-lag cross-correlation function peak value between the acoustic signals and the vibration time domain data of at least one spatial node comprises: Acoustic signal The acoustic signal is down-sampled to make the sampling frequency consistent with the sampling frequency of the vibration time-domain data, to obtain a down-sampled acoustic signal; vibration time-domain data at a center position of the mold is selected to perform normalized time-lag cross-correlation calculation with the down-sampled acoustic signal, to obtain a cross-correlation function value; the cross-correlation function value is calculated by the following formula: ； in, The time delay is At that time, the cross-correlation function value between the vibration time-domain data and the downsampled acoustic signal, For the downsampled acoustic signal in The value at time, The mean of the downsampled acoustic signal. The mean of the vibration time-domain data. For vibration time domain data in The value at time, The standard deviation of the downsampled acoustic signal. The standard deviation of the vibration time-domain data; The maximum value of the time lag is set to in milliseconds by iterating over finding the maximum value of and setting the maximum value of as the normalized time-lagged cross-correlation function peak.

5. The method of claim 2, wherein the method further comprises: The box-counting fractal dimension of the space-time state matrix constructed using the vibration time domain data of multiple spatial nodes comprises: A 5-dimensional spatio-temporal state matrix was constructed with M data points collected in 1 second time window as time dimension, and 5 spatial nodes distributed at the four corners and the geometric center of the mold as spatial dimension The minimum number of boxes required to cover the matrix trajectory was calculated by box-counting method The relationship between the box length and the box-counting fractal dimension was calculated by the following formula: ； The linear regression fit The slope of the linear relationship is found to be .​ 6. The method of claim 1, wherein, The closed-loop adjustment of the frequency and amplitude of the vibrator according to the deviation comprises: A multivariable proportional-integral-derivative controller is used for adjustment; the deviation vector between the real-time characteristic state vector and the target characteristic state vector is taken as the input of the multivariable proportional-integral-derivative controller, and the output is the control increment of the frequency and amplitude.

7. The method of claim 6, wherein the method further comprises: The adjustment range of the frequency is set to 50-150Hz, and the adjustment range of the amplitude is set to 0.5-2.0mm.

8. The method of claim 1, wherein, The generation and application of a non-periodic disturbance sequence to the vibrator comprises: Logistic mapping function is used to generate chaotic sequence Logistic mapping function is: ； The non-periodic disturbance sequence is obtained by normalizing the chaotic sequence; The non-periodic disturbance sequence is superimposed on the frequency control signal output by the multivariable proportional-integral-derivative controller; the duration of the non-periodic disturbance sequence is 3 seconds, and the superimposed amplitude is 10% of the amplitude of the main control signal.

9. The method of claim 2, wherein the method further comprises: The sampling frequency of the condenser microphone is 44.1kHz; the sampling frequency of the piezoelectric acceleration sensor is 2kHz.

10. A sensor-based automatic adjustment system for a precast unit vibrator, comprising: The method comprises a memory and a processor, and the memory stores computer instructions; when the processor executes the computer instructions, the automatic adjustment method of the sensor-based vibrator for prefabricated parts is realized.

Images