A decision system for matching vibration parameters of an attached vibrator

Abstract

This application relates to the field of building construction technology and discloses a vibration parameter matching decision system for attached vibrators. The system includes: a sensing and execution dual-mode vibrator network, a vibration acoustic fingerprint generation and analysis module, a fingerprint evolution time-series inference module, and a multi-objective dynamic decision engine. The system generates a vibration acoustic fingerprint composed of multi-dimensional physical features from response signals. By constructing a fingerprint evolution time-series sequence and comparing it with a benchmark database, the system infers the concrete material properties online, and then calls a mapping function matched to the properties to evaluate the real-time compaction state. Finally, based on the evaluated compaction, the multi-objective dynamic decision engine uses a model predictive control strategy to generate the optimal dynamic operation instructions for each vibrator node. This invention constructs a closed-loop adaptive control system of detection, evaluation, decision-making, and execution, which can match the optimal vibration parameters in real time, significantly improving the vibration quality and uniformity of concrete, reducing energy consumption, and reducing reliance on human experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building construction, in particular to an attached vibrator vibration parameter matching decision system. BACKGROUND

[0002] Concrete vibration is a key process in concrete structure construction, and its purpose is to remove the trapped bubbles in the concrete by vibration, reduce internal voids, and make the concrete mixture dense and uniform, so as to ensure the strength, durability and impermeability of the hardened concrete and other performance indicators. The attached vibrator can transmit vibration energy to a large area of concrete through the formwork, and is widely used in the construction of components such as walls, dams and tunnel linings.

[0003] However, the current operation process of the attached vibrator is still largely an open-loop control process that relies on preset parameters and manual experience. The operator usually sets the vibration frequency and operation time of the vibrator according to the construction specifications or personal experience, and subjectively judges whether the vibration is complete by observing the macro phenomena such as concrete surface paste and bubble escape during the vibration process. This method lacks objective and real-time feedback on the evolution of the internal density of the concrete.

[0004] Due to the inability to accurately perceive the dynamic changes of the internal state of the concrete, this traditional operation mode is prone to uncontrollable vibration quality. If the vibration is insufficient, there will be a lot of voids left in the concrete, forming defects such as honeycomb and pitted surface, which seriously affect the structural safety; on the contrary, if the vibration is excessive, it will cause the segregation and stratification of the concrete mixture, making the coarse aggregate sink and the cement mortar float, which will also deteriorate the mechanical properties and uniformity of the concrete. In addition, different batches of concrete may have differences in mix proportion, slump, etc., and fixed vibration parameters are difficult to adapt to the fluctuations in material performance. The whole process not only depends heavily on the proficiency and responsibility of the operator, but also often leads to unnecessary energy consumption and prolongs the construction time, making it difficult to meet the strict requirements of modern and lean construction for quality control and efficiency improvement. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an attached vibrator vibration parameter matching decision system, which solves the problem that the vibration quality of the attached vibrator during the vibration process is heavily dependent on manual experience and difficult to ensure uniformity and consistency due to the inability to perceive the internal density of the concrete in real time and adaptively adjust the vibration parameters.

[0006] To achieve the above purpose, the present application is implemented by the following technical solutions:

[0007] The first aspect of the present application provides an attached vibrator vibration parameter matching decision system, comprising:

[0008] a perception-execution dual-mode vibrator network configured to control multiple vibrator nodes within the network to switch between a vibration mode and a detection mode, and in the detection mode, to emit a detection signal by a designated at least one vibrator node and to receive a response signal after penetrating a concrete medium by the remaining at least one vibrator node;

[0009] a vibration-acoustic fingerprint generation and analysis module connected to the perception-execution dual-mode vibrator network and configured to extract multiple preset physical features from the response signal to generate a vibration-acoustic fingerprint representing an internal state of the concrete medium;

[0010] a fingerprint evolution time sequence inference module connected to the vibration-acoustic fingerprint generation and analysis module and configured to construct a fingerprint evolution time sequence composed of the vibration-acoustic fingerprints at multiple time points, and to infer material properties of the concrete medium and to evaluate a compactness state thereof based on the fingerprint evolution time sequence;

[0011] a multi-objective dynamic decision engine connected to the fingerprint evolution time sequence inference module and configured to generate a dynamic operation instruction aiming to optimize a balance between concrete vibration quality, energy consumption and efficiency based on the inferred material properties and the evaluated compactness state, and to issue the dynamic operation instruction to the perception-execution dual-mode vibrator network.

[0012] Preferably, the perception-execution dual-mode vibrator network adopts a time-division multiplexing mechanism to divide a working time into alternating vibration time slots and detection time slots. In the vibration time slots, all vibrator nodes perform vibration operations according to the dynamic operation instruction; in the detection time slots, a grid scanning of the concrete medium is achieved by rotating the vibrator nodes emitting the detection signal.

[0013] Preferably, the vibration-acoustic fingerprint generation and analysis module is specifically configured to extract at least two physical features selected from the following group: time of flight, energy attenuation, spectral centroid, phase dispersion, kurtosis, to generate the vibration-acoustic fingerprint.

[0014] In one specific embodiment, the fingerprint evolution time sequence inference module includes a reference database having multiple reference entries, one reference entry corresponding to one material category and including a reference fingerprint evolution time sequence corresponding to the material category and a mapping function generated based on experimental data of the material category. The fingerprint evolution time sequence inference module is specifically configured to infer the material properties of the concrete medium online by calculating distances between the current fingerprint evolution time sequence and each reference fingerprint evolution time sequence in the reference database.

[0015] In one embodiment, the fingerprint evolution time series inference module is configured to compare the computed minimum distance with a preset abnormality threshold, and generate an alarm signal of material performance abnormality when the minimum distance is greater than the abnormality threshold.

[0016] In one embodiment, the fingerprint evolution time series inference module is further configured to determine a material category according to the minimum distance between the fingerprint evolution time series and a reference fingerprint evolution time series trajectory, and call the mapping function corresponding to the determined material category in the reference database to map the current vibration acoustic fingerprint to the compactness state, so as to realize calibration of compactness evaluation.

[0017] Preferably, to achieve the balance between the quality, energy consumption and efficiency of the concrete vibration, the multi-objective dynamic decision engine is configured to solve a comprehensive cost function to generate the dynamic operation instruction. The comprehensive cost function is defined as the weighted sum of a quality cost term , an energy consumption cost term and a time cost term :

[0018] ;

[0019] wherein , , are preset weight coefficients corresponding to the quality, energy consumption and time cost terms respectively.

[0020] The quality cost term is represented as the norm difference between the current evaluated compactness state vector and the target compactness state vector :

[0021] ;

[0022] The energy consumption cost term is represented as the total power consumption of all vibrator nodes within the control time interval :

[0023] ;

[0024] wherein is the total number of vibrator nodes, is the instantaneous power of the th vibrator node at time . The time cost term is a constant for minimizing the number of operation steps in the optimization solution. ​

[0025] Preferably, the multi-objective dynamic decision engine adopts a model predictive control strategy, at each decision time instant, predicts the system evolution within a finite time window in the future based on the current compactness state and the material properties, and solves the comprehensive cost function on the fly to generate the dynamic operation instructions.

[0026] Preferably, the dynamic operation instructions include individualized operation parameters for each of the vibrator nodes, the operation parameters including at least one of working mode, vibration frequency, vibration amplitude and duration.

[0027] The second aspect of the present application provides a method for matching vibration parameters of an attached vibrator, comprising the following steps:

[0028] By means of the perception-execution dual-mode vibrator network, multiple vibrator nodes within the network are controlled to switch between a vibration mode and a detection mode, and in the detection mode, a detection signal is transmitted and a response signal after penetrating the concrete medium is received;

[0029] From the response signal, multiple preset physical features are extracted to generate a vibration-acoustic fingerprint representing the internal state of the concrete medium;

[0030] A fingerprint evolution time sequence composed of the vibration-acoustic fingerprints at multiple time points is constructed, and based on the fingerprint evolution time sequence, the material properties of the concrete medium are inferred on-line, and its compactness state is evaluated;

[0031] Based on the inferred material properties and the evaluated compactness state, dynamic operation instructions for each of the vibrator nodes aiming to optimize the balance between concrete vibration quality, energy consumption and efficiency are generated, and the perception-execution dual-mode vibrator network is controlled to execute.

[0032] The present application provides an attached vibrator vibration parameter matching decision system. It has the following advantages:

[0033] 1、The fingerprint evolution time sequence composed of the vibration-acoustic fingerprints at multiple time points is constructed and analyzed by the fingerprint evolution time sequence inference module, and compared with the reference fingerprint evolution time sequence trajectories of different material categories in the reference database, so that the actual material properties of the current concrete medium can be inferred on-line. Based on this inference result, the system can call the specific mapping function corresponding to the material category to evaluate the compactness, realize the dynamic calibration of the evaluation model, ensure that the control decision is always based on the true physical characteristics of the vibrated material, and significantly enhance the adaptability to the on-site material fluctuations.

[0034] 2、The multi-target dynamic decision engine is arranged, the engine generates dynamic operation instructions by solving a comprehensive cost function containing quality cost item, energy consumption cost item and time cost item, the mechanism enables the system to balance energy consumption and operation time actively and quantitatively while pursuing the final vibrating quality standard, avoids energy waste and time extension caused by excessive vibrating, or quality hidden danger caused by pursuing speed, and optimizes resource utilization of the whole operation process.

[0035] 3、The dual-mode vibrator network is perceived and executed, the vibrator itself has the function of a detection sensor, characteristic data representing the internal state of concrete can be continuously and non-invasively acquired by combining the vibration acoustic fingerprint generation and analysis module, the mechanism changes the internal density evolution process which cannot be seen into a real-time monitored data sequence, provides direct and continuous input for subsequent inference and decision, constitutes a complete closed loop from internal state perception to external action execution, and thus changes the traditional open loop and experience-dependent vibrating operation into a data-driven precise control process. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 It is a structure block diagram of the attached vibrator vibrating parameter matching decision system of the application;

[0037] Figure 2 It is a working timing and scanning mechanism process schematic diagram of the perception-execution dual-mode vibrator network of the application;

[0038] Figure 3 It is a processing flow schematic diagram of generating vibration acoustic fingerprints from response signals of the application;

[0039] Figure 4 It is a working principle block diagram of the fingerprint evolution timing inference module of the application;

[0040] Figure 5 It is a working flow schematic diagram of the multi-target dynamic decision engine adopting a model predictive control strategy of the application;

[0041] Figure 6 It is a flow chart of the attached vibrator vibrating parameter matching decision method of the application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the application specification. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0043] Reference to the drawings Figure 1 , Figure 1 is a structural block diagram of an attached vibrator vibration parameter matching decision system according to an embodiment of the present application. The attached vibrator vibration parameter matching decision system provided by the present application can include a construction site perception-execution dual-mode vibrator network, a vibration acoustic fingerprint generation and analysis module, a fingerprint evolution time sequence inference module, and a multi-objective dynamic decision engine.

[0044] The construction site perception-execution dual-mode vibrator network is configured to perform detection and vibration work in the physical world. The network is configured to control multiple vibrator nodes in the network to switch between a detection mode and a vibration mode. In the detection mode, the construction site network controls at least one vibrator node to emit a detection signal and controls at least one remaining vibrator node to receive a response signal after penetrating the concrete medium, and the response signal is output to the vibration acoustic fingerprint generation and analysis module. In the vibration mode, the construction site network receives a dynamic work instruction from the multi-objective dynamic decision engine and drives the vibrator nodes to perform vibration work accordingly.

[0045] The construction site vibration acoustic fingerprint generation and analysis module is connected to the construction site perception-execution dual-mode vibrator network. The function of the module is to extract multiple preset physical features from the received response signal and combine these physical features to generate a vibration acoustic fingerprint that can represent the physical state of the concrete medium at the current time. The vibration acoustic fingerprint is output to the fingerprint evolution time sequence inference module.

[0046] The construction site fingerprint evolution time sequence inference module is connected to the construction site vibration acoustic fingerprint generation and analysis module. The module receives and stores the vibration acoustic fingerprints generated by the construction site vibration acoustic fingerprint generation and analysis module at different time points to construct a fingerprint evolution time sequence composed of multiple time points. Based on the time sequence, the module performs two core functions: one is to infer the material performance of the current concrete medium online, and the other is to evaluate the density state of the concrete medium online. The inferred material performance and the evaluated density state are jointly output as the state of the system and are transmitted to the multi-objective dynamic decision engine.

[0047] The construction site multi-objective dynamic decision engine is connected to the construction site fingerprint evolution time sequence inference module and connected to the construction site perception-execution dual-mode vibrator network. The engine receives the material performance and the density state output by the construction site fingerprint evolution time sequence inference module and generates a dynamic work instruction based on the input, which aims to optimize the balance between the quality of concrete vibration, energy consumption, and efficiency. To achieve this function, the engine is configured to solve a comprehensive cost function , the expression of which is

[0048] ;

[0049] wherein, is the comprehensive cost function; is the quality cost term; is the energy consumption cost term; is the time cost term; are preset weight coefficients respectively corresponding to the quality, energy consumption and time cost terms.

[0050] The construction execution engine minimizes the comprehensive cost function by optimization solving, thereby generating optimal dynamic operation instructions and issuing the instructions to the construction perception-execution dual-mode vibrator network for execution.

[0051] In summary, the system of the embodiment collects signals through the construction perception-execution dual-mode vibrator network, generates state fingerprints via the construction vibration acoustic fingerprint generation and analysis module, completes the inference and evaluation of the internal state of the concrete by the construction fingerprint evolution time sequence inference module, and finally makes optimized decisions according to the evaluation results by the construction multi-objective dynamic decision engine and issues control instructions, forming a data-driven closed-loop control structure.

[0052] Referring to the accompanying Figure 2 , Figure 2 is a schematic diagram of the working time sequence and scanning mechanism of the perception-execution dual-mode vibrator network according to an embodiment of the present application. The construction perception-execution dual-mode vibrator network is an execution mechanism for the system of the present application to interact with the physical world, and its specific structure and working mode are as follows.

[0053] In one embodiment, the construction perception-execution dual-mode vibrator network includes a plurality of vibrator nodes attached to the formwork and a network control unit. Each vibrator node, in addition to containing an eccentric block and a driving motor for generating vibration, is structurally integrated with a high-sensitivity vibration sensor, such as a piezoelectric acceleration sensor. The network control unit is responsible for sending control instructions to each vibrator node and receiving signals collected by the vibration sensor.

[0054] The dual-mode nature of the construction vibrator node is embodied in its two distinct modes of operation: the vibrating mode and the probing mode. In the vibrating mode, the network control unit instructs the vibrator node's drive motor to operate at a pre-set high power, high amplitude parameter, generating intense mechanical vibrations, the purpose of which is to liquefy the concrete, expel internal air bubbles, and achieve compaction. In the probing mode, the vibrator node's operation is quite different. The network control unit instructs the vibrator selected as the transmitting node to generate a pre-set, wideband probing signal waveform, such as a Chirp signal, at an extremely low power. The energy of the probing signal is controlled at a level that does not significantly affect the state of compaction of the concrete. At the same time, the network control unit instructs the remaining vibrators selected as receiving nodes to stop all active vibrations and only activate their internal vibration sensors for collecting the probing signal transmitted through the concrete medium.

[0055] The reason for choosing the Chirp signal is that this signal has good autocorrelation properties, with its energy spread in the time domain, so that a higher signal-to-noise ratio can be obtained while keeping a relatively low peak power. At the same time, its wideband nature allows the response obtained by probing with this signal to contain richer information about the absorption and scattering of different frequency components by the medium, which is beneficial for subsequent feature extraction of the vibrational acoustic fingerprint.

[0056] To coordinate the operation of the two modes, the construction network control unit adopts a time-division multiplexing working mechanism. Referring to the accompanying Figure 2 , the network control unit divides the total working time into periodically alternating vibrating time slots and probing time slots . The duration of a vibrating time slot,

[0057] for example, a few seconds, during which all vibrator nodes in the network are in the vibrating mode, performing the vibrating job according to the instructions issued by the multi-objective dynamic decision engine. Following each vibrating time slot is a probing time slot, which has a duration of, for example, a few hundred milliseconds. Before entering the probing time slot, there is a short silent period to allow the macroscopic disturbance caused by the vibrating to decay.

[0058] In the construction probing time slot, the network control unit performs a grid-based scan of the concrete medium. This scan is achieved through a strategy of rotating the transmitting node. Specifically, in a probing time slot, the network control unit first designates the first vibrator node as the transmitting node and all the remaining vibrator nodes as receiving nodes. After the transmission is complete, the second vibrator node is immediately designated as the transmitting node and the remaining nodes as receiving nodes. This process is repeated in turn until all vibrator nodes in the network have completed a transmission of the probing signal as the transmitting node.

[0059] By the above-mentioned scanning strategy of round-robin transmission, the system is able to obtain a set of response signals covering multiple different propagation paths in the monitoring area in each probing time slot. For example, from the node transmitting, the signals received by the node characterize the acoustic properties of the concrete medium on the path mixing the response signals of all paths obtained in a probing time slot, a raw data set reflecting the spatial distribution of the internal state of the concrete is formed. The data set is then transmitted to the vibration acoustic fingerprint generation and analysis module for subsequent processing.

[0060] Referring to the accompanying Figure 3 , Figure 3 is a schematic diagram of a processing flow for generating a vibration acoustic fingerprint from response signals according to an embodiment of the present application. The building construction vibration acoustic fingerprint generation and analysis module is responsible for converting the raw, unstructured response signals collected by the perception-execution bimodal vibrator network into a structured feature vector that quantitatively characterizes the internal state of the concrete, i.e., the vibration acoustic fingerprint.

[0061] In an embodiment, the module receives a set of response signals collected by the perception-execution bimodal vibrator network in a probing time slot. Before feature extraction, the module first performs a pre-processing operation on each response signal, which includes band-pass filtering with a passband range matching the frequency spectrum of the transmitted probing signal (e.g., a linear frequency modulated signal) to filter out environmental and equipment noise, and signal windowing to intercept a specific time period containing valid signals.

[0062] After pre-processing, the module extracts a set of pre-defined physical features from each response signal. These physical features specifically include:

[0063] Time of Flight (ToF): This feature represents the propagation time of the probing signal from the transmitting node to the receiving node through the concrete medium. It is calculated by cross-correlating the received response signal with the known, reference original transmitted signal. The time delay corresponding to the peak of the cross-correlation function is determined as the time of flight. This value is directly related to the propagation speed of the acoustic wave in the medium, which reflects the elastic modulus and density of the medium.

[0064] Energy Attenuation: This feature quantifies the degree of signal energy loss during propagation. It is calculated by first computing the total energy of the received signal within a pre-defined time window, usually by integrating the square of the signal amplitude. Then, the energy value is compared with the known total energy of the transmitted signal, and the ratio or decibel (dB) difference is calculated. The magnitude of energy attenuation is related to the internal damping of the medium, as well as the scattering effects caused by pores and aggregates.

[0065] Spectral Centroid: This feature describes the distribution of the signal spectral energy. It is calculated by first performing a Fourier transform on the time-domain response signal to obtain its frequency spectrum. Then, the weighted average frequency of the spectrum is calculated, where the weight of each frequency point is its corresponding spectral amplitude. The mathematical expression is:

[0066] ;

[0067] where, is the final calculation result, usually representing the weighted average or some comprehensive index, is the spectral amplitude at frequency point , and is the frequency value of frequency point . Since high-frequency components attenuate more quickly in media such as concrete, changes in the spectral centroid can reflect the selective absorption characteristics of the medium for different frequency components.

[0068] Phase Dispersion: This feature is used to measure the consistency of the phase changes of the signal at different frequencies. When the signal passes through a porous or non-uniform medium, the propagation path and speed of different frequency components will produce slight differences, causing the phase relationship to become disordered. This module quantifies the degree of dispersion by analyzing the instantaneous frequency of the signal or statistically analyzing the phase spectrum of the spectrum (such as calculating its variance or standard deviation).

[0069] Kurtosis: This feature is a statistical moment of the signal amplitude distribution, used to measure the sharpness of the distribution or the thickness of the tail. In the initial stage of vibration, due to the presence of a large amount of random scattering and multipath propagation, the amplitude distribution of the received signal tends to be Gaussian. As the degree of compaction increases, the signal propagation path tends to be stable, and the signal waveform becomes more deterministic, with a corresponding change in the kurtosis value.

[0070] After extracting the numerical values of the above-mentioned physical features, the vibration acoustic fingerprint generation and analysis module combines these numerical values into a multi-dimensional feature vector, which is defined as the vibration acoustic fingerprint at the current time and under the current propagation path . For example:

[0071] ;

[0072] wherein, represents Time of Flight (ToF) feature, used to measure the signal propagation time, such as ranging applications of ultrasonic, light wave, etc. sensors, represents Attenuation, used to quantify the intensity reduction of the signal due to medium loss during propagation, represents Centroid, used to describe the center position of data distribution, represents Dispersion, used to quantify the dispersion degree of data, represents Kurtosis, used to measure the kurtosis or tail thickness of the probability distribution.

[0073] The module collects the vibratory acoustic fingerprints of all the paths generated in a detection time slot, forms a fingerprint matrix, and outputs it to the fingerprint evolution time inference module.

[0074] Referring to the drawings Figure 4 , Figure 4 is a working principle diagram of the fingerprint evolution time inference module according to an embodiment of the present application. The building construction fingerprint evolution time inference module is the core computing unit of the system of the present application, which receives the fingerprint data output by the vibratory acoustic fingerprint generation and analysis module, and outputs the inference results of the concrete material performance and the evaluation value of the compactness state.

[0075] In an embodiment, the module contains a pre-constructed reference database. The reference database is generated by experiment in the offline stage. The specific generation process is as follows: prepare multiple concrete standard specimens with determined material properties (for example, different grades, different mix proportions, different slump); during the whole process of vibrating these specimens, at fixed time intervals, use the detection mechanism of the present application to collect the response signals and generate the vibratory acoustic fingerprints, thereby obtaining a complete fingerprint evolution time trajectory of each type of standard specimen from the initial state to the completely compacted state, and this trajectory is the reference fingerprint evolution time trajectory. At the same time, at multiple key time points in the vibrating process, the real compactness values of the specimens at these points are obtained through independent physical measurement means (for example, gamma ray density meter or core sampling). Based on the fingerprint data collected at multiple time points and the corresponding real compactness values, through regression analysis (for example, polynomial regression or training a neural network model), a specific mapping function is constructed for each type of material, which can map the instantaneous vibratory acoustic fingerprint to the quantitative compactness value. Finally, the building construction reference database stores multiple reference entries, and each entry contains the identification of a material category, its corresponding reference fingerprint evolution time trajectory, and the associated mapping function.

[0076] In one specific embodiment, the construction mapping function is a Multi-Layer Perceptron (MLP) neural network. The input layer of the network has the same number of nodes as the dimension of the construction vibration acoustic fingerprint, receiving a fingerprint vector as input. The network contains at least one hidden layer, with e.g. Rectified Linear Unit (ReLU) as activation function. Its output layer has one node, outputting a scalar value, which is the predicted quantized density value. The network is supervised trained by backpropagation algorithm, using the offline collected fingerprint real density data, until the mean square error loss function between its predicted output and the real density value converges below a pre-set threshold.

[0077] When the system is online running, this module first performs the online inference of material properties. It maintains a current fingerprint evolution time series in memory, which is composed of the vibration acoustic fingerprints generated by the system in all detection time slots since its start. When a new fingerprint is received, it is appended to the end of the time series. Subsequently, the module computes the similarity between the current fingerprint evolution time series and each reference fingerprint evolution time series stored in the reference database, using Dynamic Time Warping (DTW) algorithm. DTW algorithm can find the optimal nonlinear alignment path between two unequal length time series, and compute the minimum cumulative distance. Its computational formula is:

[0078] ;

[0079] where, is the distance metric of dynamic time warping, is the current fingerprint evolution time series, is the th reference trajectory, and are a fingerprint vector in the two sequences, is the Euclidean distance between two fingerprint vectors, this symbol means for all possible matching paths (or matching ways) this summation expression computes the sum of squared distance (or similarity) metrics between all corresponding points in the matching path , represents an alignment path, this is a distance function, representing the metric between the th point in time series and the th point in time series .

[0080] This module determines the material category corresponding to the baseline fingerprint evolution timeline trajectory with the minimum DTW distance to the current fingerprint evolution timeline as the material property of the concrete currently being vibrated.

[0081] After determining the material category, the module then performs a calibration assessment of the compaction state. Specifically, the module retrieves the specific mapping function corresponding to the newly determined material category from the benchmark database. Then, the latest vibratory acoustic fingerprint vector, output by the vibratory acoustic fingerprint generation and parsing module, is used as input and substituted into the mapping function for calculation. The output of this function is the quantitative compaction state assessment value of the concrete medium at the current moment. This method of first identifying the material and then selecting the corresponding assessment model achieves dynamic calibration of the compaction assessment.

[0082] In addition, this module is also equipped with an abnormal state alarm function. After calculating the DTW distance between the current time series and all reference trajectories, this module will obtain the minimum distance value. With a preset abnormal threshold Compare. If This indicates that the current vibration response behavior of the concrete differs significantly from the behavior of all known standard materials in the database. At this point, the module generates an alarm signal indicating abnormal material properties to alert the operator.

[0083] Pre-set abnormal thresholds in building construction The method for determining the distance is as follows: During the offline construction of the benchmark database, the DTW distance between any two different benchmark fingerprint evolution time-series trajectories in the database is calculated to form an inter-class distance set. The statistical characteristics of this set are then taken, such as the mean plus three standard deviations. Alternatively, the 95th percentile value of the set can be used as the threshold for construction anomalies. This method ensures that the threshold is an objectively set based on the existing data distribution, rather than a subjectively specified one.

[0084] See attached document Figure 5 , Figure 5 This is a schematic diagram illustrating the workflow of a multi-objective dynamic decision engine employing a model predictive control strategy according to an embodiment of the present invention. The multi-objective dynamic decision engine for building construction is the central control and optimization unit of the system of the present invention. It receives the evaluation results of the internal state of the concrete and outputs optimal control commands for the vibrator network.

[0085] In one embodiment, the engine employs a Model Predictive Control (MPC) strategy to generate dynamic operation instructions. The core of this strategy is that, at each control cycle (e.g. at the end of each probing time slot), instead of only computing the optimal instruction at the current time, the engine predicts the system behavior in a finite time window (i.e. prediction horizon) in the future based on a system model, and then performs rolling optimization based on the prediction.

[0086] The specific implementation process of this MPC strategy includes the following steps:

[0087] Firstly, the engine obtains the current density state vector and the material performance parameters inferred by the fingerprint evolution time series inference module. The engine contains a state prediction model inside, which can predict the state evolution sequence of the system in the future time steps based on the current density state , material performance parameters, and a set of assumed future control input sequences 1). Here, the control input is a vector containing the operation parameters (such as frequency, amplitude) of all vibrator nodes at time . This state prediction model can be established by system identification on experimental data homologous to the benchmark database in the offline stage.

[0088] In one specific embodiment, the construction state prediction model is a Long Short-Term Memory (LSTM) network. The input of this LSTM network is the combination of the current density state and the current applied control input . The network learns the time evolution law of the density state under different control inputs through its internal loop structure and gating mechanism. Its output is the predicted value of the density state at the next time . By reusing the predicted output as the input of the next time step, this model can realize the autoregressive prediction of the state evolution sequence in the future time steps.

[0089] Rolling optimization: after obtaining the state prediction, the engine solves an open-loop optimization problem in the prediction horizon , whose goal is to find an optimal control sequence:

[0090] ;

[0091] wherein, It is a set representing the optimized or selected sequence of control signals. To control the time domain, so that the comprehensive cost function in the future prediction time domain is minimize, This is a list of elements within a set, representing the sequence from time [time]. Start to control signals , Indicates time The optimal control signal or control input Indicates time (in The optimal control signal, is a non-negative integer representing the length of the signal sequence or the size of the control time window. In this case, the aforementioned comprehensive cost function is expressed in the prediction time domain as... The cumulative form of the above:

[0092] ;

[0093] in, For at any time Predicted time The density state vector; The target compaction state vector; and In the control step Energy and time costs, These are the preset weighting coefficients corresponding to the quality, energy consumption, and time cost items, respectively.

[0094] During the solution process, various constraints must be satisfied, such as the vibration frequency and amplitude of each vibrator node must be within their physically permissible upper and lower limits. This optimization problem can be solved using numerical optimization algorithms such as Sequential Quadratic Programming (SQP).

[0095] Command execution and feedback correction: After calculating the optimal control sequence Subsequently, a key feature of the MPC strategy is that the engine only executes the first control command in the sequence. As the final dynamic operation instruction, it is issued to the sensing-execution dual-mode vibrator network for execution. The remainder of the sequence... Then it is discarded. At the next decision-making moment... The system will measure the new actual density state. The engine uses this new measured state as the initial condition and repeats the state prediction and rolling optimization process described above. This feedback correction mechanism based on the latest measured state enables the control process to effectively cope with model mismatch and external disturbances.

[0096] Finally, the dynamic operation instruction generated by the engine is a set of instructions containing personalized operation parameters for each vibrator node in the network. The construction operation parameters specifically include the working mode (vibration or standby) of the node in the next vibration time slot, the vibration frequency, the vibration amplitude, and the duration of the operation. In this way, the engine achieves fine, forward-looking, and adaptive control of the entire vibration process.

[0097] Referring to the accompanying Figure 6 , Figure 6 is a flowchart of an attachment vibrator vibration parameter matching decision method according to an embodiment of the present application. The overall workflow of the present application will be described in detail below in combination with the foregoing system structure.

[0098] Step S1: System initialization.

[0099] Before the start of the vibration operation, the system is powered on and initialized. This process includes loading the pre-constructed reference database into memory, which contains the reference fingerprint evolution time sequence trajectories and mapping functions corresponding to various material categories. At the same time, the target density state vector and the weight coefficients in the comprehensive cost function are set.

[0100] Step S2: Perform a detection scan.

[0101] The system enters the first working cycle. The construction sensing and execution dual-mode vibrator network first enters the detection time slot. After a short period of silence, the network performs a complete gridding scan, that is, by rotating to designate different vibrator nodes as transmitting nodes and the remaining nodes as receiving nodes, a set of response signals covering multiple propagation paths in the monitoring area is collected.

[0102] Step S3: Generate vibration acoustic fingerprints. The construction sensing and execution dual-mode vibrator network transmits the set of collected original response signals to the vibration acoustic fingerprint generation and analysis module. The module pre-processes each signal and extracts a set of physical features such as time of flight, energy attenuation, and spectral centroid from it, and combines these features into a multi-dimensional feature vector, which is the vibration acoustic fingerprint corresponding to each propagation path at the current time.

[0103] Step S4: Infer material properties and evaluate density.

[0104] The construction vibration acoustic fingerprint generation and analysis module outputs the newly generated fingerprint matrix to the fingerprint evolution time series inference module. This module first appends the new fingerprint data to the end of the current fingerprint evolution time series sequence being constructed. Then, it calculates the distance between this sequence and each benchmark trajectory in the benchmark database using the Dynamic Time Warping (DTW) algorithm, determining the current concrete material properties based on the material category corresponding to the minimum distance. Next, it calls a specific mapping function associated with that material category, inputting the latest fingerprint vector to calculate the quantized density state vector at the current moment. .

[0105] Step S5: Perform multi-objective optimization decision-making.

[0106] The construction fingerprint evolution time sequence inference module infers the inferred material performance parameters and the evaluated density state vector. This information is then transmitted to the multi-objective dynamic decision engine. This engine uses... Given the current state, and utilizing its internal state prediction model and model predictive control (MPC) strategy, the optimal control sequence for the next control time domain is calculated by solving the problem of minimizing the comprehensive cost function over a finite prediction time domain. .

[0107] Step S6: Issue and execute dynamic job instructions.

[0108] The multi-objective dynamic decision engine for building construction calculates the optimal control sequence. In the middle, only the first control instruction is extracted. This is then used as a dynamic work instruction for the next vibration time slot and sent to the construction sensing and execution dual-mode vibrator network. The network then enters the vibration time slot and executes the vibration operation precisely according to the working mode, frequency, amplitude, and other parameters set for each vibrator node in the instruction.

[0109] Step S7: Determine if the task is completed and loop.

[0110] After one vibration interval, the system evaluates the current density state vector. With the target compactness state vector Compare. If All components are greater than or equal to If the corresponding component is detected, the vibration operation is considered complete, and the system stops operating. Otherwise, the system will return to step S2, enter the next detection time slot, and begin a new round of detection, evaluation, decision-making, and execution closed-loop control cycle until the operation is completed.

Claims

1. An attached vibrator vibration parameter matching decision system, characterized in that, The method comprises the following steps: a perception-execution dual-mode vibrator network is configured to control multiple vibrator nodes in the network to switch between a vibration mode and a detection mode, and in the detection mode, a detection signal is emitted by at least one designated vibrator node, and a response signal after penetrating a concrete medium is received by at least one remaining vibrator node; a vibration acoustic fingerprint generation and analysis module is connected to the perception-execution dual-mode vibrator network and is configured to extract multiple preset physical features from the response signal to generate a vibration acoustic fingerprint representing the internal state of the concrete medium; a fingerprint evolution time sequence inference module is connected to the vibration acoustic fingerprint generation and analysis module and is configured to construct a fingerprint evolution time sequence composed of the vibration acoustic fingerprints at multiple time points, and to infer the material performance of the concrete medium and to evaluate the compactness state of the concrete medium based on the fingerprint evolution time sequence; the fingerprint evolution time sequence inference module further comprises a reference database, the reference database stores multiple reference entries, and one reference entry corresponds to one material category and comprises a reference fingerprint evolution time sequence corresponding to the material category and a mapping function generated based on experimental data of the material category; the fingerprint evolution time sequence inference module is specifically configured to infer the material performance of the concrete medium by calculating the distance between the current fingerprint evolution time sequence and each reference fingerprint evolution time sequence in the reference database; the fingerprint evolution time sequence inference module is further configured to: determine a material category according to the minimum distance between the fingerprint evolution time sequence and the reference fingerprint evolution time sequence; and call the mapping function corresponding to the determined material category in the reference database to map the current vibration acoustic fingerprint to the compactness state for calibration of the compactness evaluation; a multi-target dynamic decision engine is connected to the fingerprint evolution time sequence inference module and is configured to generate dynamic operation instructions aiming to optimize the balance between concrete vibration quality, energy consumption and efficiency based on the inferred material performance and the evaluated compactness state, and to issue the dynamic operation instructions to the perception-execution dual-mode vibrator network.

2. The attached vibrator vibration parameter matching decision system according to claim 1, characterized in that, the perception-execution dual-mode vibrator network is configured to: divide the working time into alternating vibration time slots and detection time slots using a time division multiplexing mechanism; in the vibration time slots, all vibrator nodes perform vibration operations according to the dynamic operation instructions; in the detection time slots, the vibrator nodes that emit the detection signal are rotated to perform grid scanning on the concrete medium.

3. The attached vibrator vibration parameter matching decision system according to claim 1, characterized in that, the vibration acoustic fingerprint generation and analysis module is specifically configured to extract at least two physical features selected from the following group to generate the vibration acoustic fingerprint: time of flight, energy attenuation, spectral centroid, phase dispersion and kurtosis.

4. The attached vibrator vibration parameter matching decision system according to claim 1, characterized in that, the fingerprint evolution time sequence inference module is further configured to compare the calculated minimum distance with a preset abnormal threshold, and when the minimum distance is greater than the abnormal threshold, generate an alarm signal for material performance abnormality.

5. The attached vibrator vibration parameter matching decision system according to claim 1, characterized in that, The multi-objective dynamic decision engine is configured to solve a comprehensive cost function to generate the dynamic operation instructions, the comprehensive cost function including a weighted sum of a quality cost term, an energy cost term and a time cost term.

6. The attached vibrator vibration parameter matching decision system according to claim 5, characterized in that, The multi-objective dynamic decision engine adopts a model predictive control strategy, at each decision-making time, based on the current density state and the material performance, predicts the system evolution in a finite time window in the future, and solves the comprehensive cost function to generate the dynamic operation instructions.

7. The attached vibrator vibration parameter matching decision system according to claim 1, characterized in that, The dynamic operation instructions include individualized operation parameters for each of the vibrator nodes, the operation parameters including at least one of working mode, vibration frequency, vibration amplitude and duration.

8. An attached vibrator vibration parameter matching decision method, characterized in that, The method comprises the following steps: The perception-execution dual-mode vibrator network is used to control multiple vibrator nodes in the network to switch between the vibration mode and the detection mode, and in the detection mode, to emit detection signals and receive response signals after penetrating the concrete medium; A plurality of preset physical features are extracted from the response signals to generate a vibration-acoustic fingerprint representing the internal state of the concrete medium; A fingerprint evolution time sequence composed of the vibration-acoustic fingerprints at multiple time points is constructed, and the following sub-steps are performed to infer the material performance of the concrete medium and evaluate its density state: First, by calculating the distance between the current fingerprint evolution time sequence and a plurality of reference fingerprint evolution trajectory tracks stored in a reference database, a reference fingerprint evolution trajectory track with the smallest distance to the current fingerprint evolution time sequence is determined, and the material category corresponding to the smallest distance trajectory track is determined as the material performance of the concrete medium; Then, a mapping function corresponding to the determined material category in the reference database is called, and the current latest vibration-acoustic fingerprint is input into the mapping function to evaluate the density state of the concrete medium; Based on the inferred material performance and the evaluated density state, dynamic operation instructions for each of the vibrator nodes are generated, aiming to optimize the balance between concrete vibration quality, energy consumption and efficiency, and the perception-execution dual-mode vibrator network is controlled to execute.

Images