A method and apparatus for detecting key dimensions of large precast components

Abstract

This invention provides a method and apparatus for detecting key dimensions of large precast components, relating to the technical field of dimension detection methods. The method includes simultaneously acquiring the original time-domain signals of vibration at various measuring points; uniformly dividing the power spectral density curve of each measuring point into multiple sub-bands; selecting the sub-band with the lowest energy as a noise reference value and setting a signal-to-noise ratio threshold; constructing resonance peak extraction rules; constructing the actual resonance frequency feature vector of the measuring points of the precast component; constructing a resonance frequency prediction model using the finite element method; determining the vibration type through a finite element solver; constructing a predicted resonance frequency vector; calculating the frequency difference to obtain the final dimension of the precast component. By acquiring the original time-domain signals of each measuring point and combining them with material parameters, wavelet denoising and fast Fourier transform are used to improve the accuracy of feature extraction; and by comparing the actual resonance frequency with the predicted resonance frequency, the final dimension is corrected.

Description

Technical Field

[0001] This invention relates to the field of dimensional inspection methods, specifically to a method and apparatus for inspecting key dimensions of large precast components. Background Technology

[0002] In the production of precast components, the accuracy of critical dimensions is crucial to the safety and performance of the structure. However, traditional dimensional inspection methods often rely on manual measurement, which is susceptible to human error and fails to fully utilize the relationship between material properties and vibration characteristics, resulting in insufficient inspection accuracy. This makes it difficult to effectively identify and correct dimensional errors in components during construction, potentially leading to safety hazards and a decline in structural performance.

[0003] The existing technology, disclosed in CN119554967A, uses Building Information Modeling (BIM) technology to construct a BIM model, and utilizes UAV aerial surveying technology and 3D laser scanners to construct point cloud models and overall centimeter-level and millimeter-level realistic 3D models of prefabricated components for inspection and testing of key dimensions of prefabricated components. BIM technology is used to pre-process and resolve internal details and collision issues of hidden components in prefabricated components, visualizes complex segmental structures in advance, and imports automatically calculated quantities into factory processing to control costs and improve the standardization of prefabricated components. 3D laser scanning technology and UAV aerial surveying technology are used for complementary modeling. However, this method is costly, requiring significant time and manpower for inspecting the key dimensions of prefabricated components. It does not comprehensively consider the physical properties of materials and the geometric parameters of components, increasing the uncertainty of manual intervention in the inspection process. Therefore, there is an urgent need for an efficient and reliable method for key dimension inspection.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for detecting key dimensions of large precast components, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for detecting key dimensions of large precast components, comprising the following steps:

[0008] S1: Set multiple measuring points on the surface of the precast component to be tested, and synchronously acquire the original time-domain signal of the vibration of each measuring point to obtain the material parameters of the precast component, including elastic modulus, density and Poisson's ratio;

[0009] S2: Normalize the original time-domain signal of each measurement point, use wavelet denoising to denoise the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising.

[0010] S3: Divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, construct the resonance peak extraction rule, extract the peak of the resonance frequency of each measurement point, and construct the actual resonance frequency feature vector of the measurement point of the prefabricated component based on the peak of the resonance frequency of each measurement point.

[0011] S4: Define the dimension vector for each measuring point. The dimension vector includes the length, width, and thickness of the precast component. Construct a material parameter vector based on the material parameters. Concatenate the dimension vector and the material parameter vector. Use the finite element method to construct a resonant frequency prediction model. Use the concatenated composite vector as the input of the model and the corresponding predicted resonant frequency as the output of the model. Based on the material parameters, determine the vibration type using a finite element solver. The vibration type includes the bending, torsion, and expansion / contraction states of the precast component. Construct a predicted resonant frequency vector based on the predicted resonant frequency and the vibration type.

[0012] S5: Based on the actual resonant frequency eigenvector and the predicted resonant frequency vector, calculate the frequency difference, map the frequency difference to the size error through logarithmic transformation, and sum the actual size and the size error to obtain the final size of the precast component.

[0013] Furthermore, wavelet denoising is used to denoise the original time-domain signals at each measurement point, specifically including the following steps:

[0014] The normalized original time-domain signal at each measurement point is decomposed into multi-scale wavelet coefficients to obtain multi-scale detail coefficients and approximation coefficients. Soft thresholding is then applied to these multi-scale detail coefficients to obtain the thresholded detail coefficients. The denoised signal is reconstructed using these thresholded detail coefficients and approximation coefficients. Independent denoising of the original time-domain signal at bending, torsion, and stretching points is performed using inverse wavelet transform to obtain the denoised signal. Time-domain signal at each measurement point The power spectral density curves of the original time-domain signals at each measurement point after noise reduction are calculated using Fast Fourier Transform.

[0015] Furthermore, the power spectral density curves of the original time-domain signals at each measurement point after noise reduction are calculated using Fast Fourier Transform, specifically including the following steps:

[0016] Obtain the noise-reduced first The time-domain signal at each measurement point is continuous. With sampling frequency Discretize to obtain a length of The discrete sequence is obtained; a fast Fourier transform is performed on the discrete signal to calculate the square of the spectral amplitude; the positive frequency part is retained, and the energy is doubled to generate the corresponding frequency value.

[0017] Furthermore, based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency at each measuring point. Based on the peak of the resonant frequency at each measuring point, the actual resonant frequency feature vector of the measuring points of the precast component is constructed. The specific steps are as follows:

[0018] Calculate the total energy of each subband:

[0019] ;

[0020] Indicates the first The first measuring point The energy carried by a person; Indicates the first One-sided power spectral density at each measurement point; Indicates the length of the discrete sequence; Indicates the discrete sampling frequency; Indicates the first A set of frequency indices for each sub-band; Indicates the number of sub-band indices;

[0021] Choose the one with the lowest energy. Each sub-band serves as a noise reference:

[0022] ;

[0023] in, The set of indices representing the lowest energy subband;

[0024] For each sub-band, calculate its signal-to-noise ratio:

[0025] ;

[0026] Indicates the first The first measuring point Signal-to-noise ratio of each band;

[0027] Set global threshold Valid resonance peaks are screened, resonance peak extraction rules are constructed, and resonance frequency peaks are extracted:

[0028] ;

[0029] in, Indicates the first The first measuring point The frequency points corresponding to the resonance peaks detected within each sub-band; Indicates peak width; Indicates the preset minimum peak width; Indicates the first Each measuring point is at Signal power at the location; Indicates the power ratio threshold;

[0030] Based on the structural dynamics model, the theoretical frequency range of the original time-domain signal at each measuring point is defined as follows:

[0031] ;

[0032] in, Indicates the frequency at the theoretical bending point; Indicates the frequency at the theoretical point of torsion; Indicates the frequency at the theoretical scaling point;

[0033] Extracted resonance peaks Matching with the theoretical interval, filtering effective frequencies, for each measurement point Construct a three-dimensional feature vector:

[0034] ;

[0035] in, Indicates the first Frequency values ​​at the bends of each measuring point; Indicates the first Frequency values ​​at the torsion points of each measuring point; Indicates the first Frequency values ​​at the expansion joints of each measuring point; Indicates the first The actual resonant frequency eigenvector of each measurement point.

[0036] Furthermore, a material parameter vector is constructed based on the material parameters, and the size vector and the material parameter vector are concatenated. The specific steps are as follows:

[0037] Each measuring point corresponds to a finite local region on the component surface, with a size of [missing information]. Cube range:

[0038] Using a laser scanner to acquire the area around the measurement point The point cloud data of the region is projected along the main axis, and the 95% confidence space of the projection interval is taken. The length is then calculated as the length of the measurement point.

[0039] Five parallel measurement lines are set up in the horizontal range. The actual width of each line is measured by a contact displacement sensor. The median value of the width range that appears most frequently is selected as the width of the measurement point through histogram analysis.

[0040] Within a cube centered on the measuring point, four additional measuring points symmetrically distributed are selected, and a total of five thickness measurements are performed, one at the center point and four at the symmetrical points. The average thickness of the measuring points is calculated using Gaussian weighting, and the average thickness is taken as the final thickness of the measuring point.

[0041] Define the dimension vector:

[0042] ;

[0043] in, Indicates the first The dimension vector of each measuring point; Indicates the first The length of the material at each measuring point; Indicates the first The width of the material at each measuring point; Indicates the first The thickness of the material at each measuring point;

[0044] Construct a material parameter vector based on the material's elastic modulus, density, and Poisson's ratio:

[0045] ;

[0046] in, Represents a vector of material parameters; Indicates the elastic modulus; Indicates density; Indicates Poisson's ratio;

[0047] By concatenating the dimension vector and the material parameter vector, we obtain the combined vector:

[0048] ;

[0049] in, This represents the combined vector after concatenation.

[0050] Furthermore, a resonant frequency prediction model is constructed using the finite element method. The concatenated composite vector is used as the model input, and the corresponding predicted resonant frequency is used as the model output. The specific steps are as follows:

[0051] The geometric model of the prefabricated component is automatically generated using computer programming. Material parameters are assigned to the generated geometric model, the geometric model is meshed, and eigenvalue equations are constructed.

[0052] ;

[0053] in,

[0054] ;

[0055] ;

[0056] Indicates the first Cross-sectional area of ​​each measuring point; Indicates the first Stiffness matrix of each measuring point; Indicates the first Mass matrix of each measurement point; Indicates the first The natural angular frequencies of the first mode; Indicates the first The mode shape vector of the first mode;

[0057] Solving the resonance equation yields the resonance frequency:

[0058] ;

[0059] Indicates the first One resonant frequency; Indicates the resonant frequency index;

[0060] The model output is the front One resonant frequency vector:

[0061] ;

[0062] Indicates the first The resonance vector of each measuring point.

[0063] Furthermore, based on material parameters, the vibration type is determined using a finite element solver. The vibration type includes bending, torsion, and expansion / contraction states of the precast component. Based on the predicted resonance frequency and vibration type, a predicted resonance frequency vector is constructed, specifically including the following steps:

[0064] The displacement direction weight ratio is defined as a modal discrimination index, and the vibration type is determined by the finite element solver:

[0065] ;

[0066] ;

[0067] ;

[0068] Indicates the axial displacement component; Indicates the lateral displacement component; Represents the angular displacement component; This indicates the proportion of axial displacement energy to total energy; This indicates the proportion of lateral displacement energy to total energy; This indicates the proportion of angular displacement energy to total energy; Denotes the Euclidean norm;

[0069] Design discrimination rules:

[0070] ;

[0071] Indicates the type of vibration;

[0072] Based on the predicted resonance frequency and vibration type, a predicted resonance frequency vector is constructed for each measuring point:

[0073] ;

[0074] Indicates the first The predicted resonant frequency vector matrix for each measurement point; Indicates the first The bending resonance frequency at each measuring point; Indicates the first Torsional resonance frequency at each measuring point; Indicates the first The stretching resonant frequency at each measuring point.

[0075] Further, the frequency difference is calculated, and the frequency difference is mapped to the dimensional error through logarithmic transformation. The actual size is then summed with the dimensional error to obtain the final size of the precast component. This process specifically includes the following steps:

[0076] Obtain the difference between the actual resonance frequency feature vector and the predicted resonance frequency vector corresponding to the vibration type at each measuring point:

[0077] ;

[0078] in,

[0079] ;

[0080] in, The mapping correlation coefficient is obtained by linearly fitting the size error and frequency error; Indicates the first The first measuring point Frequency difference between vibration types; Indicates the first The first measuring point The actual resonant frequency of each vibration type; Show the first The first measuring point Predicted resonant frequencies for each vibration type; Indicates the first The first measuring point Size error for each vibration type; This represents a correction constant, ensuring that the input to the logarithmic function is positive.

[0081] Adding the dimensional error to the original measured dimension yields the corrected dimension, which is:

[0082] ;

[0083] in, Indicates the corrected number The dimensions of each measuring point; Indicates the first The original dimensions of each measuring point.

[0084] The present invention also provides a device for detecting critical dimensions of large precast components, the device being used to perform the above-described detection method, comprising:

[0085] The data acquisition module is used to set multiple measuring points on the surface of the precast component to be tested, synchronously acquire the original time-domain signal of the vibration of each measuring point, and acquire the material parameters of the precast component, including elastic modulus, density and Poisson's ratio;

[0086] The power spectrum calculation module is used to normalize the original time-domain signal of each measurement point, use wavelet denoising to reduce the noise of the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising.

[0087] The actual resonant frequency module is used to uniformly divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency of each measurement point. Based on the peak of the resonant frequency of each measurement point, the actual resonant frequency feature vector of the measurement point of the prefabricated component is constructed.

[0088] The resonant frequency prediction module defines the dimension vector for each measuring point, which includes the length, width, and thickness of the precast component. A material parameter vector is constructed based on material parameters. The dimension vector and material parameter vector are concatenated, and a resonant frequency prediction model is built using the finite element method. The concatenated composite vector is used as the model input, and the corresponding predicted resonant frequency is used as the model output. Based on the material parameters, the vibration type is determined using a finite element solver. The vibration type includes bending, torsion, and expansion / contraction states of the precast component. Based on the predicted resonant frequency and the vibration type, a predicted resonant frequency vector is constructed.

[0089] The dimensional error module calculates the frequency difference based on the actual resonant frequency eigenvector and the predicted resonant frequency vector. It then maps the frequency difference logarithmically to the dimensional error, and sums the actual size and the dimensional error to obtain the final size of the precast component. Compared with the prior art, the advantages of this invention are:

[0090] Wavelet denoising and Fast Fourier Transform (FFT) are applied to the original time-domain signal at the measurement point to extract accurate resonant frequency features. Compared with existing technologies, this improves the efficiency and accuracy of detection. Wavelet denoising effectively removes random noise from the signal, improving the signal-to-noise ratio, while FFT rapidly generates the power spectral density, providing a reliable foundation for subsequent frequency extraction and a solid data basis for subsequent finite element analysis and dimensional correction. The overall scheme achieves high-precision resonant frequency detection and analysis, thereby promoting the dimensional optimization and quality control of precast components. Through systematic signal processing and model prediction, the impact of human error is significantly reduced, improving detection efficiency. Attached Figure Description

[0091] Figure 1 This is a schematic diagram of the overall method flow of the present invention.

[0092] Figure 2 This is a schematic diagram of the overall system flow of the present invention. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0094] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0095] Example:

[0096] Please see Figure 1 The present invention provides a technical solution:

[0097] A method for detecting key dimensions of large precast components, comprising the following steps:

[0098] S1: Set multiple measuring points on the surface of the precast component to be tested, and synchronously acquire the original time-domain signal of the vibration of each measuring point to obtain the material parameters of the precast component, including elastic modulus, density and Poisson's ratio;

[0099] S2: Normalize the original time-domain signal of the measurement point, use wavelet denoising to denoise the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising.

[0100] The wavelet denoising process is used to reduce noise in the original time-domain signals of each measurement point, specifically including the following steps:

[0101] The normalized original time-domain signal of the measurement points is subjected to multi-scale wavelet decomposition to obtain multi-scale detail coefficients and approximation coefficients:

[0102] ;

[0103] in,

[0104] ;

[0105] ;

[0106] Indicates the first The first measuring point Layer detail factor; Indicates the first The first measuring point Layer approximation coefficient; Describe the wavelet basis functions; Represents the scaling function; Indicates the first Layer Detail coefficients; Indicates the level index; Indicates the index of detail coefficients; Indicates the first Layer Detail coefficients; Indicates the maximum number of decomposition levels; Indicates the first The normalized original time-domain signal of each measurement point;

[0107] Soft thresholding is performed on the multi-scale detail coefficients to obtain the thresholded detail coefficients:

[0108] ;

[0109] in,

[0110] ;

[0111] Indicates the first The first measuring point Standard deviation of layer noise; The length of the signal; Indicates the number of th values ​​after threshold processing The first measuring point Layer Detail coefficients; Indicates the first The first measuring point Adaptive thresholding for layer wavelet decomposition;

[0112] Reconstruct the denoised signal using the thresholded coefficients:

[0113] ;

[0114] in,

[0115] ;

[0116] Represents the thresholded first... The first measuring point Layer detail factor;

[0117] Independent denoising is performed on the original time-domain signals at bending, torsion, and expansion points:

[0118] ;

[0119] in, Indicates the first The thresholded first measurement point Layer Detail coefficients; Indicates the first The first measuring point Layer approximation coefficient; Indicates the inverse wavelet transform; Indicates the first after noise reduction The time-domain signal of each measurement point.

[0120] The calculation of the power spectral density of the original time-domain signal at each measurement point after noise reduction using Fast Fourier Transform specifically includes the following steps:

[0121] Obtain the noise-reduced first The time-domain signal at each measurement point is continuous. With sampling frequency Discretize to obtain a length of Discrete sequences:

[0122] ;

[0123] in,

[0124] ;

[0125] in, Indicates the index of a discrete signal sequence; express Discrete signals at each measurement point;

[0126] Perform a Fast Fourier Transform on a discrete signal:

[0127] ;

[0128] in,

[0129] ;

[0130] Indicates the first Discrete signals after Fourier transform of each measurement point; Indicates the index of scattered frequency components;

[0131] Calculate the square of the spectral amplitude:

[0132] ;

[0133] in, Represents the square of the spectral amplitude of a discrete signal;

[0134] Preserve the positive frequency portion and double the energy:

[0135] ;

[0136] Indicates the first One-sided power spectral density at each measurement point;

[0137] Generate the corresponding frequency values:

[0138] ;

[0139] Indicates the length of the discrete sequence; Indicates the first The actual frequency value corresponding to each frequency point.

[0140] S3: Divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, construct the resonance peak extraction rule, extract the peak of the resonance frequency of each measurement point, and construct the actual resonance frequency feature vector of the measurement point of the prefabricated component based on the peak of the resonance frequency of each measurement point.

[0141] The steps are as follows: Based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency at each measuring point. Based on the peak of the resonant frequency at each measuring point, an actual resonant frequency feature vector of the measuring points of the precast component is constructed.

[0142] Calculate the total energy of each subband:

[0143] ;

[0144] Indicates the first The first measuring point The energy carried by a person; Indicates the first One-sided power spectral density at each measurement point; Indicates the length of the discrete sequence; Indicates the discrete sampling frequency; Indicates the first A set of frequency indices for each sub-band; Indicates the number of sub-band indices;

[0145] Choose the one with the lowest energy. Each sub-band serves as a noise reference:

[0146] ;

[0147] in, The set of indices representing the lowest energy subband;

[0148] For each sub-band, calculate its signal-to-noise ratio:

[0149] ;

[0150] Indicates the first The first measuring point Signal-to-noise ratio of each band;

[0151] Set global threshold Valid resonance peaks are screened, resonance peak extraction rules are constructed, and resonance frequency peaks are extracted:

[0152] ;

[0153] in, Indicates the first The first measuring point The frequency points corresponding to the resonance peaks detected within each sub-band; Indicates peak width; Indicates the preset minimum peak width; Indicates the first Each measuring point is at Signal power at the location; Indicates the power ratio threshold;

[0154] Based on the structural dynamics model, the theoretical frequency range of the original time-domain signal at each measuring point is defined as follows:

[0155] ;

[0156] in, Indicates the frequency at the theoretical bending point; Indicates the frequency at the theoretical point of torsion; Indicates the frequency at the theoretical scaling point;

[0157] Extracted resonance peaks Matching with the theoretical interval, filtering effective frequencies, for each measurement point Construct a three-dimensional feature vector:

[0158] ;

[0159] in, Indicates the first Frequency values ​​at the bends of each measuring point; Indicates the first Frequency values ​​at the torsion points of each measuring point; Indicates the first Frequency values ​​at the expansion joints of each measuring point; Indicates the first The actual resonant frequency eigenvector of each measurement point.

[0160] In the above process, the energy of each frequency band is first calculated through sub-band energy analysis. Select the lowest energy subband to estimate the noise floor. And calculate the subband signal-to-noise ratio. Subsequently, based on the signal-to-noise ratio threshold Power ratio threshold and minimum peak width Triple constraints extract effective resonance peaks Finally, these resonance peaks were matched with the bending, torsion, and stretching frequency ranges predicted by structural dynamics theory to select characteristic frequencies that conform to physical laws. , , Ultimately, this is used to construct a three-dimensional feature vector that characterizes the dynamic properties of the structure. .

[0161] S4: Define a dimension vector, which includes the length, width, and thickness of the precast component. Construct a material parameter vector based on the material parameters. Concatenate the dimension vector and the material parameter vector. Use the finite element method to construct a resonant frequency prediction model. Use the concatenated composite vector as the input of the model and the corresponding predicted resonant frequency as the output of the model. Determine the vibration type using a finite element solver based on the material parameters. The vibration type includes the bending, torsion, and expansion / contraction states of the precast component. Construct a predicted resonant frequency vector based on the predicted resonant frequency and the vibration type.

[0162] The specific steps for constructing a material parameter vector based on material parameters and concatenating the size vector and the material parameter vector are as follows:

[0163] Each measuring point corresponds to a finite local region on the component surface, with a size of [missing information]. Cube range:

[0164] Using a laser scanner to acquire the area around the measurement point The point cloud data of the region is projected along the main axis, and the 95% confidence space of the projection interval is taken. The length is then calculated as the length of the measurement point.

[0165] The principal axis direction here refers to a limited local area on the surface of the component. The main geometric extension direction within the cube range usually corresponds to the maximum dimension axis of the component in space. For example, for a beam-shaped structure in the shape of an I-beam, its main axis direction is usually the length direction of the beam, that is, the longest dimension axis. For a plate-shaped structure such as a metal plate, the main axis is the normal or long side direction of the plate surface. After laser scanning the point cloud data, the point cloud is projected onto the coordinate axis of this direction, and the length dimension is determined by the projection distribution.

[0166] The computational length refers to the length of the point cloud data projected along the principal axis, resulting in a set of one-dimensional coordinate values. The 95% confidence space of the projected coordinates is then calculated, and the span of this interval, which is the difference between the maximum and minimum values, represents the measurement point. The length; reflects the effective structural dimensions of the local area of ​​the measuring point in the principal axis direction;

[0167] For example, after the local point cloud of an I-beam is projected onto the principal axis, the coordinate range is... cm, outlier removal After cm, the 95% confidence space is cm, then cm. Five parallel measuring lines are set up in the horizontal range. The actual width of each line is measured using a contact displacement sensor. The median value of the width interval that occurs most frequently is selected as the width of the measuring point through histogram analysis.

[0168] The lateral range here refers to the plane perpendicular to the principal axis direction, i.e. Within the measurement range, five measurement lines parallel to the main axis are set in this plane. The actual width value of each line is measured by a contact displacement sensor; the final width... Take the median of the highest frequency interval in the histogram of all measurements;

[0169] For example, the principal axis direction, along the beam length, is the X-axis; the transverse plane perpendicular to the principal axis direction is the YZ plane; and five measurement lines parallel to the principal axis direction are set at the following locations: cm, the Z-direction dimension of each line is measured using a displacement sensor, and the measured value is [value missing]. cm, of which For outliers, perform histogram analysis, where, If it appears 4 times, take the median value. cm as Within a cube centered on the measuring point, four additional measuring points symmetrically distributed were selected, and a total of five thickness measurements were performed: one at the center point and four at the symmetrical points. The average thickness of the measuring points was calculated using Gaussian weighting, and the average thickness was taken as the final thickness of the measuring point.

[0170] Define the dimension vector:

[0171] ;

[0172] in, Indicates the first The dimension vector of each measuring point; Indicates the first The length of the material at each measuring point; Indicates the first The width of the material at each measuring point; Indicates the first The thickness of the material at each measuring point;

[0173] Construct a material parameter vector based on the material's elastic modulus, density, and Poisson's ratio:

[0174] ;

[0175] in, Represents a vector of material parameters; Indicates the elastic modulus; Indicates density; Indicates Poisson's ratio;

[0176] By concatenating the dimension vector and the material parameter vector, we obtain the combined vector:

[0177] ;

[0178] in, This represents the combined vector after concatenation.

[0179] The method of constructing a resonance frequency prediction model using the finite element method involves using the concatenated composite vector as the model input and the corresponding predicted resonance frequency as the model output. The specific steps are as follows:

[0180] The geometric model of the prefabricated component is automatically generated using computer programming. Material parameters are then assigned to the generated geometric model, and the geometric model is meshed.

[0181] The Python programming language is used in conjunction with a geometric modeling library to generate a parametric geometric model. The generated parametric geometric model is saved in a standardized format, and the mesh generation library gmsh is used to discretize the geometric model to generate a mesh.

[0182] Construct eigenvalue equations:

[0183] ;

[0184] in,

[0185] ;

[0186] ;

[0187] Indicates the first Cross-sectional area of ​​each measuring point; Indicates the first Stiffness matrix of each measuring point; Indicates the first Mass matrix of each measurement point; Indicates the first The natural angular frequencies of the first mode; Indicates the first The mode shape vector of the first mode;

[0188] Solving the resonance equation yields the resonance frequency:

[0189] ;

[0190] Indicates the first One resonant frequency; Indicates the resonant frequency index;

[0191] The model output is the front One resonant frequency vector:

[0192] ;

[0193] Indicates the first The resonance vector of each measuring point.

[0194] The process of determining the vibration type based on material parameters using a finite element solver, where the vibration type includes bending, torsion, and expansion / contraction states of the precast component, and constructing a predicted resonance frequency vector based on the predicted resonance frequency and vibration type, specifically includes the following steps:

[0195] The displacement direction weight ratio is defined as a modal discrimination index, and the vibration type is determined by the finite element solver:

[0196] ;

[0197] ;

[0198] ;

[0199] Indicates the axial displacement component; Indicates the lateral displacement component; Represents the angular displacement component; This indicates the proportion of axial displacement energy to total energy; This indicates the proportion of lateral displacement energy to total energy; This indicates the proportion of angular displacement energy to total energy; Denotes the Euclidean norm;

[0200] Design discrimination rules:

[0201] ;

[0202] Indicates the type of vibration;

[0203] Based on the predicted resonance frequency and vibration type, a predicted resonance frequency vector is constructed for each measuring point:

[0204] ;

[0205] Indicates the first The predicted resonant frequency vector matrix for each measurement point; Indicates the first The bending resonance frequency at each measuring point; Indicates the first Torsional resonance frequency at each measuring point; Indicates the first The stretching resonant frequency at each measuring point.

[0206] In the above process, because the vibration type discrimination rule ensures that each measuring point corresponds to only a single vibration type through a mutual exclusion threshold, the predicted vector... The three-dimensional structure is essentially a type-frequency standardized container, where each of its three elements is bound to a specific type condition, such as... Activation only occurs during scaling, with only one position storing the effective frequency; the remaining positions are theoretically empty, facilitating comparison with measured vectors. Alignment, and also avoids the contradiction of frequency coexistence through conditional constraints.

[0207] S5: Based on the actual resonant frequency eigenvector and the predicted resonant frequency vector, calculate the frequency difference, map the frequency difference to the size error through logarithmic transformation, and sum the actual size and the size error to obtain the final size of the precast component.

[0208] The frequency difference is calculated, and then mapped to the dimensional error using a logarithm. The actual size is summed with the dimensional error to obtain the final size of the precast component. This process includes the following steps:

[0209] Obtain the difference between the actual resonance frequency feature vector and the predicted resonance frequency vector corresponding to the vibration type at each measuring point:

[0210] ;

[0211] in,

[0212] ;

[0213] in, The mapping correlation coefficient is obtained by linearly fitting the size error and frequency error; Indicates the first The first measuring point Frequency difference between vibration types; Indicates the first The first measuring point The actual resonant frequency of each vibration type; Show the first The first measuring point Predicted resonant frequencies for each vibration type; Indicates the first The first measuring point Size error for each vibration type; This represents a correction constant, ensuring that the input to the logarithmic function is positive.

[0214] when When the actual resonant frequency is higher than the predicted frequency, the size correction is positive; when When the actual frequency is lower than the predicted frequency, the size correction is negative; dependent variable Specifically, it reflects the precast components at a specific point. The dimensional error correction amount at that location is obtained by adjusting the frequency difference. The processing quantifies the difference between the actual and predicted dimensions of the component.

[0215] Adding the dimensional error to the original measured dimension yields the corrected dimension, which is:

[0216] ;

[0217] in, Indicates the corrected number The dimensions of each measuring point; Indicates the first The original dimensions of each measuring point.

[0218] The present invention also provides a critical dimension detection device for large precast components, the detection device being used to perform the above-described detection method, comprising:

[0219] The data acquisition module is used to set multiple measuring points on the surface of the precast component to be tested, synchronously acquire the original time-domain signal of the vibration of each measuring point, and acquire the material parameters of the precast component, including elastic modulus, density and Poisson's ratio;

[0220] The power spectrum calculation module is used to normalize the original time-domain signal of each measurement point, use wavelet denoising to reduce the noise of the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising.

[0221] The actual resonant frequency module is used to uniformly divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency of each measurement point. Based on the peak of the resonant frequency of each measurement point, the actual resonant frequency feature vector of the measurement point of the prefabricated component is constructed.

[0222] The resonant frequency prediction module defines the dimension vector for each measuring point, which includes the length, width, and thickness of the precast component. A material parameter vector is constructed based on material parameters. The dimension vector and material parameter vector are concatenated, and a resonant frequency prediction model is built using the finite element method. The concatenated composite vector is used as the model input, and the corresponding predicted resonant frequency is used as the model output. Based on the material parameters, the vibration type is determined using a finite element solver. The vibration type includes bending, torsion, and expansion / contraction states of the precast component. Based on the predicted resonant frequency and the vibration type, a predicted resonant frequency vector is constructed.

[0223] The dimensional error module calculates the frequency difference based on the actual resonant frequency eigenvector and the predicted resonant frequency vector. This frequency difference is then logarithmically mapped to the dimensional error. The actual dimension and the dimensional error are summed to obtain the final dimension of the precast component. All the above formulas are dimensionless numerical calculations derived from software simulations using a large amount of collected data, resulting in formulas that most closely reflect real-world conditions. The preset parameters in these formulas can be set by those skilled in the art based on specific circumstances.

[0224] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0225] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0226] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting key dimensions of large precast components, characterized by the following steps: include: S1: Set multiple measuring points on the surface of the precast component to be tested, and synchronously acquire the original time-domain signal of the vibration of each measuring point to obtain the material parameters of the precast component, including elastic modulus, density and Poisson's ratio; S2: Normalize the original time-domain signal of each measurement point, use wavelet denoising to denoise the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising. S3: Divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, construct the resonance peak extraction rule, extract the peak of the resonance frequency of each measurement point, and construct the actual resonance frequency feature vector of the measurement point of the prefabricated component based on the peak of the resonance frequency of each measurement point. S4: Define the dimension vector for each measuring point. The dimension vector includes the length, width, and thickness of the precast component. Construct a material parameter vector based on the material parameters. Concatenate the dimension vector and the material parameter vector. Use the finite element method to construct a resonant frequency prediction model. Use the concatenated composite vector as the input of the model and the corresponding predicted resonant frequency as the output of the model. Based on the material parameters, determine the vibration type using a finite element solver. The vibration type includes the bending, torsion, and expansion / contraction states of the precast component. Construct a predicted resonant frequency vector based on the predicted resonant frequency and the vibration type. S5: Based on the actual resonant frequency eigenvector and the predicted resonant frequency vector, calculate the frequency difference, map the frequency difference to the size error through logarithmic transformation, and sum the actual size and the size error to obtain the final size of the precast component.

2. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The wavelet denoising process is used to reduce noise in the original time-domain signals of each measurement point, specifically including the following steps: The normalized original time-domain signal at each measurement point is decomposed into multi-scale wavelet coefficients to obtain multi-scale detail coefficients and approximation coefficients. Soft thresholding is then applied to these multi-scale detail coefficients to obtain the thresholded detail coefficients. The denoised signal is reconstructed using these thresholded detail coefficients and approximation coefficients. Independent denoising of the original time-domain signal at bending, torsion, and stretching points is performed using inverse wavelet transform to obtain the denoised signal. Time-domain signal at each measurement point The power spectral density curves of the original time-domain signals at each measurement point after noise reduction are calculated using Fast Fourier Transform.

3. The method for detecting key dimensions of large precast components according to claim 2, characterized in that, The calculation of the power spectral density curves of the original time-domain signals at each measurement point after noise reduction using Fast Fourier Transform specifically includes the following steps: Obtain the noise-reduced first The time-domain signal at each measurement point is continuous. With sampling frequency Discretize to obtain a length of The discrete sequence is obtained; a fast Fourier transform is performed on the discrete signal to calculate the square of the spectral amplitude; the positive frequency part is retained, and the energy is doubled to generate the corresponding frequency value.

4. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The steps are as follows: Based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency at each measuring point. Based on the peak of the resonant frequency at each measuring point, an actual resonant frequency feature vector of the measuring points of the precast component is constructed. Calculate the total energy of each subband: ； Indicates the first The first measuring point The energy carried by a person; Indicates the first One-sided power spectral density at each measurement point; Indicates the length of the discrete sequence; Indicates the discrete sampling frequency; Indicates the first A set of frequency indices for each sub-band; Indicates the number of sub-band indices; Choose the one with the lowest energy. Each sub-band serves as a noise reference: ； in, The set of indices representing the lowest energy subband; For each sub-band, calculate its signal-to-noise ratio: ； Indicates the first The first measuring point Signal-to-noise ratio of each band; Set global threshold Valid resonance peaks are screened, resonance peak extraction rules are constructed, and resonance frequency peaks are extracted: ； in, Indicates the first The first measuring point The frequency points corresponding to the resonance peaks detected within each sub-band; Indicates peak width; Indicates the preset minimum peak width; Indicates the first Each measuring point is at Signal power at the location; Indicates the power ratio threshold; Based on the structural dynamics model, the theoretical frequency range of the original time-domain signal at each measuring point is defined as follows: ； in, Indicates the frequency at the theoretical bending point; Indicates the frequency at the theoretical point of torsion; Indicates the frequency at the theoretical scaling point; Extracted resonance peaks Matching with the theoretical interval, filtering effective frequencies, for each measurement point Construct a three-dimensional feature vector: ； in, Indicates the first Frequency values ​​at the bends of each measuring point; Indicates the first Frequency values ​​at the torsion points of each measuring point; Indicates the first Frequency values ​​at the expansion joints of each measuring point; Indicates the first The actual resonant frequency eigenvector of each measurement point.

5. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The specific steps for constructing a material parameter vector based on material parameters and concatenating the size vector and the material parameter vector are as follows: Each measuring point corresponds to a finite local region on the component surface, with a size of [missing information]. Cube range: Using a laser scanner to acquire the area around the measurement point The point cloud data of the region is projected along the main axis, and the 95% confidence space of the projection interval is taken. The length is then calculated as the length of the measurement point. Five parallel measurement lines are set up in the horizontal range. The actual width of each line is measured by a contact displacement sensor. The median value of the width range that appears most frequently is selected as the width of the measurement point through histogram analysis. Within a cube centered on the measuring point, four additional measuring points symmetrically distributed are selected, and a total of five thickness measurements are performed, one at the center point and four at the symmetrical points. The average thickness of the measuring points is calculated using Gaussian weighting, and the average thickness is taken as the final thickness of the measuring point. Define the dimension vector: ； in, Indicates the first The dimension vector of each measuring point; Indicates the first The length of the material at each measuring point; Indicates the first The width of the material at each measuring point; Indicates the first The thickness of the material at each measuring point; Construct a material parameter vector based on the material's elastic modulus, density, and Poisson's ratio: ； in, Represents a vector of material parameters; Indicates the elastic modulus; Indicates density; Indicates Poisson's ratio; By concatenating the dimension vector and the material parameter vector, we obtain the combined vector: ； in, This represents the combined vector after concatenation.

6. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The method of constructing a resonance frequency prediction model using the finite element method involves using the concatenated composite vector as the model input and the corresponding predicted resonance frequency as the model output. The specific steps are as follows: Computer software is used to generate geometric models of prefabricated components. Material parameters are assigned to the generated geometric models, which are then meshed, and eigenvalue equations are constructed. ； in, ； ； Indicates the first Cross-sectional area of ​​each measuring point; Indicates the first Stiffness matrix of each measuring point; Indicates the first Mass matrix of each measurement point; Indicates the first The natural angular frequencies of the first mode; Indicates the first The mode shape vector of the first mode; Solving the resonance equation yields the resonance frequency: ； Indicates the first One resonant frequency; Indicates the resonant frequency index; The model output is the front One resonant frequency vector: ； Indicates the first The resonance vector of each measuring point.

7. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The vibration type is determined using a finite element method (FEM) based on material parameters. The vibration type includes bending, torsion, and expansion / contraction states of the precast component. A predicted resonance frequency vector is constructed based on the predicted resonance frequency and vibration type. Includes the following steps: The displacement direction weight ratio is defined as a modal discrimination index, and the vibration type is determined by the finite element solver: ； ； ； Indicates the axial displacement component; Indicates the lateral displacement component; Represents the angular displacement component; This indicates the proportion of axial displacement energy to total energy; This indicates the proportion of lateral displacement energy to total energy; This indicates the proportion of angular displacement energy to total energy; Denotes the Euclidean norm; Design discrimination rules: ； Indicates the type of vibration; Based on the predicted resonance frequency and vibration type, a predicted resonance frequency vector is constructed for each measuring point: ； Indicates the first The predicted resonant frequency vector matrix for each measurement point; Indicates the first The bending resonance frequency at each measuring point; Indicates the first Torsional resonance frequency at each measuring point; Indicates the first The stretching resonant frequency of each measuring point.

8. The method for detecting key dimensions of large precast components according to claim 1, characterized in that, The calculation of the frequency difference involves mapping the frequency difference to the dimensional error using a logarithm, and then summing the actual size and the dimensional error to obtain the final size of the prefabricated component. This process specifically includes the following steps: Obtain the difference between the actual resonance frequency feature vector and the predicted resonance frequency vector corresponding to the vibration type at each measuring point: ； in, ； in, The mapping correlation coefficient is obtained by linearly fitting the size error and frequency error; Indicates the first The first measuring point Frequency difference between vibration types; Indicates the first The first measuring point The actual resonant frequency of each vibration type; Show the first The first measuring point Predicted resonant frequencies for each vibration type; Indicates the first The first measuring point Size error for each vibration type; This represents a correction constant, ensuring that the input to the logarithmic function is positive. Adding the dimensional error to the original measured dimension yields the corrected dimension, which is: ； in, Indicates the corrected number The dimensions of each measuring point; Indicates the first The original dimensions of each measuring point.

9. A device for detecting key dimensions of large precast components, characterized in that: The detection device is used to perform the detection method according to any one of claims 1-8, including: The data acquisition module is used to set multiple measuring points on the surface of the precast component to be tested, synchronously acquire the original time-domain signal of the vibration of each measuring point, and acquire the material parameters of the precast component, including elastic modulus, density and Poisson's ratio; The power spectrum calculation module is used to normalize the original time-domain signal of each measurement point, use wavelet denoising to reduce the noise of the original time-domain signal of each measurement point, and use fast Fourier transform to calculate the power spectral density curve of the original time-domain signal of each measurement point after denoising. The actual resonant frequency module is used to uniformly divide the power spectral density curve of each measurement point into multiple sub-bands, select the sub-band with the lowest energy as the noise reference value and set the signal-to-noise ratio threshold. Based on the signal-to-noise ratio threshold, a resonant peak extraction rule is constructed to extract the peak of the resonant frequency of each measurement point. Based on the peak of the resonant frequency of each measurement point, the actual resonant frequency feature vector of the measurement point of the prefabricated component is constructed. The resonant frequency prediction module defines the dimension vector for each measuring point, which includes the length, width, and thickness of the precast component. A material parameter vector is constructed based on material parameters. The dimension vector and material parameter vector are concatenated, and a resonant frequency prediction model is built using the finite element method. The concatenated composite vector is used as the model input, and the corresponding predicted resonant frequency is used as the model output. Based on the material parameters, the vibration type is determined using a finite element solver. The vibration type includes bending, torsion, and expansion / contraction states of the precast component. Based on the predicted resonant frequency and the vibration type, a predicted resonant frequency vector is constructed. The size error module is used to calculate the frequency difference based on the actual resonant frequency feature vector and the predicted resonant frequency vector, map the frequency difference to the size error through logarithmic transformation, and sum the actual size and the size error to obtain the final size of the prefabricated component.

Images