Arch rib displacement prediction method and system based on differential interference radar

Abstract

The invention provides an arch rib displacement prediction method and system based on differential interference radars, and the method comprises the steps: 1, taking the vertex of an arch bridge as a calibration point, respectively selecting a plurality of groups of symmetric measurement points at the two sides of the arch bridge, installing the differential interference radars on the symmetric measurement points through differential interference radar targets, performing real-time monitoring on a transverse displacement difference value between the symmetrical measuring points on the two sides of the arch bridge through a differential interference radar to obtain original time history data; step 2, according to the original time history data, calculating time history data of displacement components of the arch rib in the transverse bridge direction, the longitudinal bridge direction and the vertical direction; step 3, performing noise reduction on the time history data by adopting an improved wavelet threshold denoising method; and 4, artificial intelligence is adopted to predict the time history data after noise reduction, and predicted arch rib displacement is obtained. According to the method, the arch rib deformation can be efficiently monitored, and the trend of the arch rib deformation can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the fields of wavelet analysis, machine learning and overall improvement technology, and in particular to a method and system for predicting arch rib displacement based on differential interferometry radar. Background Art

[0002] Differential interferometry radar is an intelligent deformation monitoring system with continuous multi-point measurements that has emerged in recent years. It combines a number of advanced remote sensing technologies such as linear frequency modulation continuous wave signal pulse compression and interferometry, and its monitoring accuracy can reach sub-millimeter level. It has the advantages of long range, high measurement accuracy, and can work continuously all day and all weather, unattended, etc. Compared with traditional deformation monitoring methods, the advantages of differential interferometry radar technology are: the range can reach several kilometers, operators do not need to enter the detection area, and multiple observation points can be monitored continuously in real time and all weather, with high monitoring accuracy.

[0003] The main advantages of the overall lifting construction method are: (1) The arch rib structure is assembled on the ground or on a low-position support, avoiding the dangers brought by high-altitude operations, greatly reducing the difficulty of assembly and welding, and ensuring the quality of arch rib structure welding and the linear shape of the arch rib; (2) The construction period is short, the dangers brought by high-altitude operations are reduced, the construction quality is easy to ensure, and the structural integrity is good. However, due to the large deformation of the steel arch ribs, high synchronization requirements, sensitive joint adjustment, and difficult to control construction stability during the overall lifting process, it is difficult to meet the high-precision, intelligent and multi-angle monitoring requirements using traditional monitoring methods; traditional status assessment methods have problems such as low efficiency and low accuracy. The overall lifting process collects test data from many aspects, but has not yet been effectively applied to the status assessment of arch bridges. The overall data utilization rate is low and the level of intelligence of arch bridge status assessment is not high; and with the development of artificial intelligence and big data technologies, bridge structure status assessment technologies based on machine learning and other methods are rapidly emerging. These methods can automatically extract features and rules, and convert a large amount of bridge inspection data into quantitative structural status assessment results, thereby improving the accuracy and efficiency of the assessment. Therefore, a displacement test method for the large-stage overall lifting process of the arch rib based on differential interferometry radar was proposed to provide guidance for the application of overall lifting technology in arch bridge construction under special environments. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a displacement prediction method and system for the overall lifting process of a large-stage steel tube concrete arch bridge based on differential interferometry radar, the purpose of which is to achieve efficient monitoring of arch rib deformation and accurately predict the trend of arch rib deformation.

[0005] In order to achieve the above object, the specific scheme of the present invention is as follows:

[0006] The arch rib displacement prediction method based on differential interferometry radar includes the following steps:

[0007] Step 1: Taking the apex of the arch bridge as the calibration point, select multiple groups of symmetrical measuring points on both sides of the arch bridge, install differential interferometry radars on the symmetrical measuring points through differential interferometry radar targets, and use differential interferometry radars to monitor the lateral displacement difference between the symmetrical measuring points on both sides of the arch bridge in real time to obtain original time history data;

[0008] Step 2, obtaining the time history data of the displacement components of the arch rib in the transverse direction, along the bridge direction, and vertical direction components according to the original time history data described in step 1;

[0009] Step 3, using the improved wavelet threshold denoising method to reduce the noise of the time course data;

[0010] Step 4: Use artificial intelligence prediction based on the denoised time-history data to obtain the predicted arch rib displacement.

[0011] Furthermore, the differential interferometer radar target described in step S1 includes a target body, an L-shaped metal connecting piece, a suction cup mechanism and a connector, wherein the target body is formed by connecting a metal plate and an L-shaped metal plate to form a semi-enclosed structure, a plurality of L-shaped metal connecting pieces are respectively installed at the connection between the metal plate and the L-shaped metal plate and at the bending of the L-shaped metal plate, a plurality of suction cup mechanisms are respectively installed on the outside of the target body, the connector includes a connecting panel, a connecting arm and an arc block, one side of the connecting panel is connected to the target body through the suction cup mechanism, and the two ends of the other side are respectively connected to one end of the connecting arm, and the other ends of the two connecting arms are respectively connected to the arc blocks, and the suction cup mechanism is composed of a suction cup with a nut and a suction cup with a threaded rod.

[0012] Furthermore, the formula for the lateral displacement difference in step 1 is as follows:

[0013] C=X i -X j

[0014] Where C is the difference of the symmetrical measuring points; i is 1, 2, 3; X i is the transverse displacement component of measuring points 1, 2, and 3, and the values ​​of j are 1#, 2#, and 3#; X j It is the transverse displacement component of measuring points 1#, 2#, and 3#.

[0015] Furthermore, the original time history data described in step 1 is data after pulse compression of linear frequency modulation continuous wave signal, differential interferometry target displacement change, removal of static clutter and atmospheric correction;

[0016] The linear frequency modulated continuous wave signal is expressed as:

[0017]

[0018] Assume that the transmitted signal is S T(t), ignoring the influence of signal amplitude after a modulation period of time τ, the received signal is obtained as

[0019] S R (t) = S T (t-τ) (2)

[0020] The echo signal and the transmitted signal are mixed according to the mixing formula. The mixing formula is as follows:

[0021] S IF (t) = S T (t) / S T (t-τ) (3)

[0022] Pulse compression can directly perform Fourier transform on the mixing signal, the formula is as follows:

[0023]

[0024] So, |S IF (f)|=|sin c(π(ατ-f)(T r -τ))|,|S IF The peak value of (f)| appears at That is, the target point location;

[0025] In the formula, S IF (t) is the mixing signal; S IF (f) is the mixing signal after Fourier transformation; T t is the period; τ is the echo delay, τ=2R / c, c is the electromagnetic wave velocity, λ is the wavelength, d is the distance between the target and the radar; j represents the resolution unit with i as the center and L as the length range; f c is the carrier frequency; t is the time within a modulation cycle; T r is the time within one modulation cycle;

[0026] The formula for measuring the target displacement change by differential interferometry is as follows:

[0027]

[0028] In the formula, The phase of the echo of the first transmitted signal received by the radar; is the echo phase of the second transmitted signal received; λ is the wavelength of the transmitted signal; is the target displacement change;

[0029] The static clutter removal is performed by fitting the center of the discrete point trajectory, and the calculation formula is as follows:

[0030] a=∑x i L i / (2L)

[0031] b=∑y i L i / (2L)

[0032]

[0033] Where L is the sum of the arc lengths between two adjacent points of all discrete points; L i For any point (x i ,y i ) and the sum of the arc lengths between two adjacent points; For any point (x i ,y i ) is the weight coefficient of the point; (a, b) represents the center of the circle;

[0034] The expression for calculating the echo phase of the target point by atmospheric correction is as follows:

[0035]

[0036] In the formula, is the echo phase of the target point; r n is the distance from the launch point; T is the atmospheric dry temperature, P d is the dry air pressure, e is the water vapor pressure; H is the humidity.

[0037] Furthermore, the displacement component formulas of the arch rib in the transverse direction, along the bridge direction, and vertical direction of step 2 are respectively as follows:

[0038] d x =d*cosθ*cosβ

[0039] d y =d*cosθ*cosα

[0040] d z =d*sinθ

[0041] Where, d x is the displacement component of the arch rib in the transverse direction; d y is the displacement component of the arch rib along the bridge direction; d z is the displacement component of the arch rib in the vertical direction; d is the distance between the target and the radar.

[0042] Furthermore, the threshold function of the improved wavelet threshold denoising method for denoising the time course data in step 3 is as follows:

[0043]

[0044] Where W J,K is the wavelet coefficient before processing; is the processed wavelet coefficient; λ is the set threshold; Sgn(·) is the sign function; J is the maximum decomposition scale, j = 1, 2, ..., J; K is the length of the wavelet coefficient, k = 1, 2, ..., n j , n j =N / 2 J-i+1 , N is the length of the signal; α and β are adjustment factors used to adjust The speed of change.

[0045] Furthermore, the method of using artificial intelligence to predict the denoised time course data described in step 4 comprises the following steps:

[0046] Step 41, preprocessing the time-course data after improved wavelet threshold denoising, and dividing the time-course data into a training set and a test set to ensure the separation of model training and evaluation;

[0047] Step 42, select a support vector machine model, determine a kernel function, use the training set to train the support vector machine regression model, and use the trained model to predict data for future time periods;

[0048] Step 43, using the test set to evaluate the performance of the model, comparing the difference between the model prediction results and the true values, adjusting the hyperparameters or feature selection of the model based on the evaluation results, and visualizing the prediction results and the true values ​​to obtain the predicted arch rib displacement.

[0049] A real-time prediction system for arch rib linear shape for implementing the displacement prediction method, comprising:

[0050] The original time history data obtaining module is used to obtain the original time history data of the steel tube concrete arch rib according to the differential interferometer radar;

[0051] A data processing module is used to eliminate errors based on the original time history data and obtain time history data of components in different directions of the arch bridge;

[0052] A denoising module is used to obtain denoised time history data based on the processed time history data of different directional components;

[0053] The artificial intelligence prediction module is used to predict the displacement of the arch rib based on the time history data of different directional components after noise reduction;

[0054] The early warning module is used to determine whether to issue an early warning based on the displacement prediction value of the arch rib combined with the deformation rate, acceleration, and accumulated deformation at the current moment.

[0055] Furthermore, the original time course data obtaining module includes:

[0056] A target recognition unit is used to identify the position of the target on the steel tube concrete arch rib;

[0057] A differential interference processing unit is used to perform differential interference processing on radar data at two adjacent moments to obtain a phase difference between the two adjacent moments;

[0058] The data processing module comprises:

[0059] Error elimination unit, used to remove mechanical errors and atmospheric errors caused by static clutter during radar monitoring;

[0060] A coordinate conversion unit is used to obtain the time history data of the arch rib in the transverse direction, along the bridge direction and vertical direction by performing spatial coordinate transformation on the time history data;

[0061] The denoising module comprises:

[0062] The data preprocessing unit is used to preprocess the time history data of the components in each direction of the arch rib, remove abnormal values ​​and fill in missing values;

[0063] The wavelet denoising unit is used to decompose the signal into low-frequency and high-frequency parts of different scales, perform threshold processing on the high-frequency part to remove noise, and then reconstruct the processed low-frequency signal and the denoised high-frequency signal;

[0064] The prediction module comprises:

[0065] The model unit selects appropriate kernel functions and parameters for the denoised time-course data to construct a prediction model;

[0066] A training unit, which is used to train the prediction model using the training set;

[0067] Evaluation unit, used to evaluate model performance on the test set, including accuracy, recall, and F1 value;

[0068] The prediction unit is used to use the trained model to make predictions on new data based on the evaluation results and explain the results.

[0069] Advantages of the present invention

[0070] (1) The present invention provides a method and system for real-time monitoring of deformation of steel tube concrete arch bridges based on differential interferometer radar data, which processes the acquired radar data in real time, converts the original time-series data into time-series data in different directions, performs noise reduction on the data, trains and predicts the noise-reduced data using a support vector machine, thereby improving the monitoring accuracy of the deformation of steel tube concrete arch bridges, and proposes a method for predicting the deformation of steel tube concrete arch bridges.

[0071] (2) The present invention uses differential interferometry radar to monitor the displacement of the overall lifting process of the reinforced concrete arch bridge, realizing all-day, all-weather long-distance visual intelligent monitoring.

[0072] (3) The present invention can not only capture the changes in the arch rib structure during the lifting process by accurately grasping the lateral displacement difference between the symmetrical measuring points, but also improve the real-time perception of the overall lifting state of the arch bridge, providing key data for subsequent structural analysis.

[0073] (4) The present invention adopts an improved wavelet threshold denoising method, which achieves the advantages of removing noise and retaining signals by adjusting the size of wavelet coefficients, which not only reduces the noise of the data but also improves the authenticity of the reconstructed signal.

[0074] (5) The present invention uses machine learning to predict the displacement of reinforced concrete arch bridges. The machine learning model can automatically learn and extract useful features without human intervention and can adapt to a variety of data types and features. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a flow chart of the displacement prediction method of the present invention.

[0076] Figure 2 To achieve Figure 1 Working principle diagram of the displacement prediction system of the displacement prediction method.

[0077] Figure 3 for Figure 1 Flow chart of data processing for overall lifting displacement prediction of a medium arch bridge.

[0078] Figure 4 for Figure 1 Schematic diagram of the structure of the differential interferometer radar target.

[0079] Figure 5 for Figure 4 Schematic diagram of the connection with the connector.

[0080] Figure 6 for Figure 1 Schematic diagram of the arrangement of symmetrical measuring points on the arch bridge.

[0081] Figure 7 for Figure 1 Schematic diagram of coordinate transformation based on target and radar.

[0082] Figure 8 for Figure 1 Comparison chart before and after denoising based on improved wavelet threshold.

[0083] In the figure:

[0084] 1. Metal plate; 2. L-shaped metal plate; 3. L-shaped metal connecting piece; 4. Suction cup mechanism; 401. Suction cup with nut; 402. Suction cup with threaded rod; 5. Connecting panel; 6. Connecting arm; 7. Arc block. DETAILED DESCRIPTION

[0085] The present invention is further explained and illustrated below in conjunction with the accompanying drawings and specific embodiments. It should be noted that this specific embodiment does not limit the scope of rights of the present invention.

[0086] like Figures 1 to 8 As shown, the arch rib displacement prediction method based on differential interferometry radar provided in this specific embodiment includes the following steps:

[0087] Step 1: Taking the apex of the arch bridge as the calibration point, select multiple groups of symmetrical measuring points on both sides of the arch bridge, install differential interferometry radars on the symmetrical measuring points through differential interferometry radar targets, and use differential interferometry radars to monitor the lateral displacement difference between the symmetrical measuring points on both sides of the arch bridge in real time to obtain original time history data;

[0088] The differential interferometer radar target includes a target body, an L-shaped metal connecting piece, a suction cup mechanism and a connector. The target body is formed by connecting a metal plate and an L-shaped metal plate by continuous welding to form a semi-enclosed structure. A plurality of L-shaped metal connecting pieces are respectively installed at the connection between the metal plate and the L-shaped metal plate and at the bending of the L-shaped metal plate as additional reinforcement. A plurality of suction cup mechanisms are respectively installed on the outside of the target body, and are used for the connection between the target and the component. In view of the case where the cross-section of the component is circular, a connector is added to make it easier for the target to be tightly connected with the component to ensure the firmness of the connection. The connector includes a connecting panel, a connecting arm and an arc block. One side of the connecting panel is connected to the target body through a suction cup mechanism, and the two ends of the other side are respectively connected to one end of the connecting arm, and the other ends of the two connecting arms are respectively connected to the arc block. The suction cup mechanism is composed of a suction cup with a nut and a suction cup with a threaded rod. The suction cup with a nut is composed of a suction cup with a diameter of 10 mm and a nut with a diameter of 4 mm and a length of 30 mm, and the suction cup with a threaded rod is composed of a suction cup with a diameter of 10 mm and a threaded rod with a diameter of 4 mm and a length of 30 mm.

[0089] In this embodiment, three groups of symmetrical measuring points are selected on both sides of the arch bridge, and the apex of the arch bridge is selected as the calibration point. The markings of the symmetrical measuring points and the calibration points are as follows: Figure 5 As shown, they are calibration point, measuring point 1, measuring point 2, measuring point 3, measuring point 1#, measuring point 2#, and measuring point 3#. The above-mentioned differential interferometer radar target is installed, and the displacement change of the arch bridge structure is obtained by monitoring the lateral displacement difference between the symmetrical measuring points.

[0090] By monitoring the calibration points, the overall displacement value of the arch bridge can be obtained. The displacement of the measuring point minus the displacement of the calibration point is calculated to obtain the accurate lateral displacement difference.

[0091] The lateral displacement difference is C=X i -X j , where C is the difference of the symmetrical measuring points, X i(i is 1, 2, 3) are the transverse displacement components of measuring points 1, 2, and 3, respectively, X j (j is 1#, 2#, 3#) is the transverse displacement component of measuring points 1#, 2#, 3#.

[0092] Firstly, the differential interferometer radar target is searched in the radar monitoring system, and the original time course data is output after the pulse compression of the linear frequency modulated continuous wave signal, differential interferometer measurement, static clutter removal and atmospheric correction are performed inside the differential interferometer radar.

[0093] The pulse compression technology of the linear frequency modulated continuous wave signal enables the radar to lock the target. The calculation process includes: assuming that the transmitted signal can be expressed as Where f c is the carrier frequency, T r is the frequency sweep period (PRI), t is the time in one modulation period, and α is the sweep frequency. Ignoring the influence of signal amplitude, the received signal is S R (t) = S T (t-τ), where τ is the echo delay, τ = 2R / c, c is the electromagnetic wave velocity, λ is the wavelength, and d is the distance between the target and the radar. The echo signal is mixed with the transmitted signal to obtain the intermediate frequency echo signal after frequency modulation:

[0094]

[0095] Pulse compression can be done directly by taking a Fourier transform of the IF signal:

[0096]

[0097] So, |S IF (f)|=|sinc(π(ατ-f)(T r -τ))|,|S IF The peak value of (f)| appears at

[0098] In the above formula, S T (t) is the transmitted signal; S R (t) is the received signal; S IF (t) is the mixing signal; S IF (f) is the mixing signal after Fourier transformation; T t is the period; τ is the echo delay, τ=2R / c, c is the electromagnetic wave velocity, λ is the wavelength, d is the distance between the target and the radar; j represents the resolution unit with i as the center and L as the length range; f c is the carrier frequency; t is the time within a modulation cycle; T r is the time within one modulation cycle;

[0099] The phase of the echo of the first transmitted signal received by the radar is The echo phase of the second transmitted signal received is The phase difference between the two echo signals is The precise displacement change of the target can be determined based on the corresponding relationship between the phase and the wavelength. The calculation formula for the displacement change of the target measured by differential interferometry is as follows:

[0100]

[0101] In the formula, The phase of the echo of the first transmitted signal received by the radar; is the echo phase of the second transmitted signal received; λ is the wavelength of the transmitted signal; is the target displacement change;

[0102] The static clutter is caused by a constant vector deviation of the observed signal vector due to the radar identifying multiple points. Due to the existence of static clutter, the complex signal obtained after data processing forms a series of discrete point trajectories in polar coordinates that deviate from the origin of the coordinate system. Therefore, the method to remove static clutter is to fit the center of the discrete point trajectory, that is, the center of the fitting circular curve deviates from the constant vector of the origin of the coordinate system. The calculation process for removing static clutter is as follows:

[0103] Assume that the sum of the arc lengths between two adjacent points of all discrete points is L, and any point (x i ,y i ) and the sum of the arc lengths between the two adjacent points is L i , the arbitrary point (x i ,y i ) has a weight coefficient of The center of the circle can be expressed as (a, b), and we get a = ∑x i L i / (2L), b=∑y i L i / (2L), radius

[0104] To this end, the static clutter component is obtained, and then the discrete points can be translated as a whole toward the coordinate origin.

[0105] The atmospheric correction refers to the phase deviation caused by the change of parameters such as temperature, humidity and air pressure in the atmosphere. The elimination process includes:

[0106] According to the radar transmission frequency at time t, c The electromagnetic wave can be calculated from the emission point r n The echo phase of the target point at Where n is the refractive index. In a small area, the observation time is short, and the expression of the target point echo phase can be simplified to Therefore, when studying a stable target, different meteorological conditions lead to differences in the atmospheric refractive index, that is, there is a phase difference between t1 and t2.

[0107] When it comes to atmospheric refraction, the refractive index is often used, called the N unit. Where T is the atmospheric dry temperature, P d is the dry air pressure, and e is the water vapor pressure. The relationship between the atmospheric refractive index n(t) and the refractive index N is as follows: n(t) = N × 10 -6 +1. Total atmospheric pressure P = P d +e. Use Tetens empirical formula to calculate saturated water vapor pressure e s , The water vapor pressure is then calculated from the saturated water vapor pressure and relative humidity In summary, the echo phase expression of the target point can be obtained:

[0108]

[0109] In the formula, is the echo phase of the target point; r n is the distance from the launch point; T is the atmospheric dry temperature, P d is the dry air pressure, e is the water vapor pressure; H is the humidity.

[0110] Step 2, obtaining the time history data of the displacement components of the arch rib in the transverse direction, along the bridge direction, and vertical direction components according to the original time history data described in step 1;

[0111] The original displacement-time data is converted into the displacement-time data of the arch rib in the transverse direction, along the bridge direction and vertical direction according to the coordinate transformation. The specific method is as follows: Since the radar first transmits the signal to the target and then receives the reflected signal, the displacement data of the overall lifting of the arch rib is displayed in the radar system as the displacement of the arch rib on the straight line between the arch rib and the radar. Figure 6 As shown, assuming that the object to be measured is a measuring point on a steel tube concrete arch bridge, the x-axis direction is the horizontal direction of the arch bridge, the y-axis direction is the longitudinal direction of the arch bridge, and the z-axis direction is the vertical direction of the arch bridge. Take point d(x,y,z) for analysis, and the coordinate transformation formula is as follows:

[0112] The formula for the transverse displacement component of an arch bridge is: x =d*cosθ*cosβ;

[0113] The formula for the displacement component of an arch bridge along the bridge direction is: y =d*cosθ*cosα;

[0114] The formula for the vertical displacement component of an arch bridge is: z =d*sinθ.

[0115] Where, d x is the displacement component of the arch rib in the transverse direction; d y is the displacement component of the arch rib along the bridge direction; d z is the displacement component of the arch rib in the vertical direction; d is the distance between the target and the radar.

[0116] Step 3: Use the improved wavelet threshold denoising method to reduce the noise of the time course data; Figure 4 As shown in FIG. 1 , the process of wavelet threshold denoising includes wavelet decomposition, threshold processing and wavelet reconstruction. The wavelet decomposition is to determine a reasonable decomposition level according to the radar data, and use discrete wavelet transform to decompose the data into low-frequency signals and high-frequency signals. The low-frequency signal is usually the actual monitoring data; the high-frequency signal is usually the noise data.

[0117] This embodiment uses an improved threshold function to adjust the size of the wavelet coefficients to remove noise and retain signal characteristics.

[0118] According to the process of engineering structure condition monitoring, the actual monitoring data is usually a low-frequency signal or a stable signal, while the noise monitoring data is usually a high-frequency signal. The wavelet threshold denoising method is based on this feature to denoise the original monitoring data.

[0119] The monitoring data can be regarded as a time series consisting of real information and noise information, that is, f(t)=s(t)+x(t), where: f(t) is the actual measured data; s(t) is the real data; x(t) is the noise data.

[0120] The main steps of wavelet threshold denoising are as follows: after selecting the appropriate wavelet basis function and decomposition layer number, f(t) is decomposed into several multi-scale layers. When performing wavelet decomposition on a signal, the larger the number of decomposition layers, the more obvious the different characteristics of noise and signal performance, and the more conducive to the separation of the two. However, the larger the number of decomposition layers, the greater the distortion of the reconstructed signal will be, which will have a poor impact on the effect of signal denoising to a certain extent.

[0121] The frequency range of wavelet decomposition is related to the sampling frequency. If N layers of decomposition are performed, the range of each frequency band is: Where: F s is the sampling frequency, f max is the maximum frequency of the signal. s =2f max It is obtained according to the Nyquist sampling theorem.

[0122] The wavelet coefficients are restricted according to the selected threshold calculation method and threshold processing function. The high-frequency coefficients and low-frequency coefficients after threshold processing are reconstructed to obtain the denoised monitoring sequence. The calculation formula of the threshold λ is: Where: M is the signal length; σ is the standard deviation of the noise; the estimation formula of σ is Where: median is the intermediate value function; ω j,k is the wavelet transform coefficient before denoising.

[0123] This embodiment uses an improved threshold function to adjust the size of the wavelet coefficients to remove noise and retain signal characteristics. The improved threshold function is as follows:

[0124]

[0125] Where W J,K is the wavelet coefficient before processing, is the processed wavelet coefficient, λ is the set threshold, Sgn(·) is the sign function, J is the maximum decomposition scale, j=1,2,...,J,K is the length of the wavelet coefficient, k=1,2,...,n j , n j =N / 2 J-i+1 , N is the length of the signal. α and β are adjustment factors used to adjust By adjusting the two adjustment factors α and β, the threshold function can be controlled more conveniently, while eliminating the discontinuity of the threshold function, making the function approach the desired position more quickly, ensuring the continuity of the threshold function, avoiding the occurrence of pseudo-Gibbs phenomenon, and reducing the deviation caused by signal compression due to wavelet decomposition and reconstruction, thereby improving the approximation between the reconstructed signal and the real signal.

[0126] like Figure 8 As shown in the figure, the comparison diagram before and after denoising based on the improved wavelet threshold function of the present invention shows that the denoised data not only retains the low-frequency information, but also removes the high-frequency noise. The results show that this method can remove noise and protect the energy characteristics of the signal itself when processing differential interferometer radar data, and has a good effect.

[0127] Step 4: The method of using artificial intelligence to predict the time-history data after noise reduction in step 4 is used to obtain the predicted arch rib displacement. The artificial intelligence prediction uses a support vector machine regression (hereinafter referred to as SVR regression model) algorithm to predict the time-history data after noise reduction. The specific steps are as follows:

[0128] Step 41, pre-process the time-history data after improved wavelet threshold denoising, and divide the time-history data into a training set and a test set to ensure the separation of model training and evaluation. For displacement-time data, it may be necessary to extract features from it, such as rate of change, acceleration, etc., for model training.

[0129] Step 42, select a support vector machine model, determine a kernel function, use the training set to train the SVR regression model, and use the trained model to predict data for future time periods;

[0130] Step 43, use the test set to evaluate the performance of the model, compare the differences between the model prediction results and the true values, and adjust the model's hyperparameters or feature selection based on the evaluation results to improve the performance and generalization ability of the model. Visualize the prediction results and the true values ​​to obtain the predicted arch rib displacement, so as to more intuitively understand the effect of the model prediction.

[0131] like Figure 2 As shown, this embodiment also provides a real-time prediction system for arch rib linear shape for implementing the displacement prediction method, including:

[0132] The original time history data obtaining module is used to obtain the original time history data of the steel tube concrete arch rib according to the differential interferometer radar;

[0133] The original time history data obtaining module includes:

[0134] A target recognition unit is used to identify the position of the target on the steel tube concrete arch rib;

[0135] The differential interference processing unit is used to perform differential interference processing on radar data at two adjacent moments to obtain the phase difference between the two adjacent moments.

[0136] A data processing module is used to eliminate errors based on the original time history data and obtain time history data of components in different directions of the arch bridge;

[0137] The data processing module comprises:

[0138] Error elimination unit, used to remove mechanical errors and atmospheric errors caused by static clutter during radar monitoring;

[0139] The coordinate conversion unit is used to obtain the time-history data of the arch rib in the transverse direction, along the bridge direction and vertical direction by performing spatial coordinate transformation on the time-history data.

[0140] A denoising module is used to obtain denoised time history data based on the processed time history data of different directional components;

[0141] The denoising module comprises:

[0142] The data preprocessing unit is used to preprocess the time history data of the components in each direction of the arch rib, remove abnormal values ​​and fill in missing values;

[0143] The wavelet denoising unit is used to decompose the signal into low-frequency and high-frequency parts of different scales, perform threshold processing on the high-frequency part to remove noise, and then reconstruct the processed low-frequency signal and the denoised high-frequency signal.

[0144] The artificial intelligence prediction module is used to predict the displacement of the arch rib based on the time history data of different directional components after noise reduction;

[0145] The artificial intelligence prediction module includes:

[0146] The model unit selects appropriate kernel functions and parameters for the denoised time-course data to construct a prediction model;

[0147] A training unit, which is used to train the prediction model using the training set;

[0148] Evaluation unit, used to evaluate model performance on the test set, including accuracy, recall, and F1 value;

[0149] The prediction unit is used to use the trained model to make predictions on new data based on the evaluation results and explain the results.

[0150] The early warning module is used to determine whether to issue an early warning based on the displacement prediction value of the arch rib combined with the deformation rate, acceleration, and accumulated deformation at the current moment.

[0151] The time course data described in steps 1 to 4 above are stored in the form of two-dimensional data consisting of displacement and time, and a deformation value is stored at each time position.

Claims

1. Arch rib displacement prediction method based on differential interferometry radar, It is characterized in that The steps include: Step 1: Taking the apex of the arch bridge as the calibration point, select multiple groups of symmetrical measuring points on both sides of the arch bridge, install differential interferometry radars on the symmetrical measuring points through differential interferometry radar targets, and use differential interferometry radars to monitor the lateral displacement difference between the symmetrical measuring points on both sides of the arch bridge in real time to obtain original time history data; Step 2, obtaining the time history data of the displacement components of the arch rib in the transverse direction, along the bridge direction, and vertical direction components according to the original time history data described in step 1; Step 3, using the improved wavelet threshold denoising method to reduce the noise of the time course data; Step 4: Use artificial intelligence to predict the time-history data after noise reduction to obtain the predicted arch rib displacement.

2. The method according to claim 1, It is characterized in that The differential interferometer radar target described in step S1 includes a target body, an L-shaped metal connecting piece, a suction cup mechanism and a connector. The target body is formed by connecting a metal plate and an L-shaped metal plate to form a semi-enclosed structure. A plurality of L-shaped metal connecting pieces are respectively installed at the connection between the metal plate and the L-shaped metal plate and at the bending of the L-shaped metal plate. A plurality of suction cup mechanisms are respectively installed on the outside of the target body. The connector includes a connecting panel, a connecting arm and an arc block. One side of the connecting panel is connected to the target body through a suction cup mechanism, and the two ends of the other side are respectively connected to one end of the connecting arm, and the other ends of the two connecting arms are respectively connected to the arc blocks. The suction cup mechanism is composed of a suction cup with a nut and a suction cup with a threaded rod.

3. The method according to claim 1, It is characterized in that The formula for the lateral displacement difference in step 1 is as follows: C=X i -X j Where C is the difference of the symmetrical measuring points; i is 1, 2, 3; X i is the transverse displacement component of measuring points 1, 2, and 3, and the values ​​of j are 1#, 2#, and 3#; X j It is the transverse displacement component of measuring points 1#, 2#, and 3#.

4. The method according to claim 1, It is characterized in that The original time history data described in step 1 is the data processed by pulse compression of linear frequency modulation continuous wave signal, differential interferometry measurement of target displacement change, removal of static clutter and atmospheric correction to calculate the echo phase of the target point; The pulse compression formula of the linear frequency modulated continuous wave signal is as follows: In the formula, S IF (t) is the mixing signal; S IF (f) is the mixing signal after Fourier transformation; T t is the period; τ is the echo delay, τ=2R / c, c is the electromagnetic wave velocity, λ is the wavelength, d is the distance between the target and the radar; j represents the resolution unit with i as the center and L as the length range; f c is the carrier frequency; t is the time within a modulation cycle; T r is the time within one modulation cycle; The formula for measuring the target displacement change by differential interferometry is as follows: In the formula, The phase of the echo of the first transmitted signal received by the radar; is the echo phase of the second transmitted signal received; λ is the wavelength of the transmitted signal; is the target displacement change; The static clutter removal is performed by fitting the center of the discrete point trajectory, and the calculation formula is as follows: a=∑x i L i / (2L) b=∑y i L i / (2L) Where L is the sum of the arc lengths between two adjacent points of all discrete points; L i For any point (x i ,y i ) and the sum of the arc lengths between two adjacent points; For any point (x i ,y i ) is the weight coefficient of the point; (a, b) represents the center of the circle; The expression for calculating the echo phase of the target point by atmospheric correction is as follows: In the formula, is the echo phase of the target point; r n is the distance from the launch point; T is the atmospheric dry temperature, P d is the dry air pressure, e is the water vapor pressure; H is the humidity.

5. The method according to claim 1, It is characterized in that The displacement component formulas of the arch rib in the transverse direction, along the bridge direction, and vertical direction of step 2 are as follows: d x =d*cosθ*cosβ d y =d*cosθ*cosα d z =d*sinθ Where, d x is the displacement component of the arch rib in the transverse direction; d y is the displacement component of the arch rib along the bridge direction; d z is the displacement component of the arch rib in the vertical direction; d is the distance between the target and the radar.

6. The method according to claim 1, It is characterized in that The threshold function of the improved wavelet threshold denoising method for denoising the time course data in step 3 is as follows: Where W J,K is the wavelet coefficient before processing; is the processed wavelet coefficient; λ is the set threshold; Sgn(·) is the sign function; J is the maximum decomposition scale, j = 1, 2, ..., J; K is the length of the wavelet coefficient, k = 1, 2, ..., n j , n j =N / 2 J-i+1 , N is the length of the signal; α and β are adjustment factors used to adjust The speed of change.

7. The method according to claim 1, It is characterized in that The method of using artificial intelligence to predict the time-course data after noise reduction described in step 4 comprises the following steps: Step 41, preprocessing the time-course data after improved wavelet threshold denoising, and dividing the time-course data into a training set and a test set; Step 42, select a support vector machine model, determine a kernel function, use the training set to train the support vector machine regression model, and use the trained model to predict data for future time periods; Step 43, using the test set to evaluate the performance of the model, comparing the difference between the model prediction results and the true values, adjusting the hyperparameters or feature selection of the model based on the evaluation results, and visualizing the prediction results and the true values ​​to obtain the predicted arch rib displacement.

8. A real-time prediction system for arch rib linear shape for implementing the methods described in claims 1 to 7, It is characterized in that include: The original time history data obtaining module is used to obtain the original time history data of the steel tube concrete arch rib according to the differential interferometer radar; A data processing module is used to eliminate errors based on the original time history data and obtain time history data of components in different directions of the arch bridge; A denoising module is used to obtain denoised time history data based on the processed time history data of different directional components; The artificial intelligence prediction module is used to predict the displacement of the arch rib based on the time history data of different directional components after noise reduction; The early warning module is used to determine whether to issue an early warning based on the displacement prediction value of the arch rib combined with the deformation rate, acceleration, and accumulated deformation at the current moment.

9. The system according to claim 6, It is characterized in that The original time history data obtaining module includes: A target recognition unit is used to identify the position of the target on the steel tube concrete arch rib; A differential interference processing unit is used to perform differential interference processing on radar data at two adjacent moments to obtain a phase difference between the two adjacent moments; The data processing module comprises: Error elimination unit, used to remove mechanical errors and atmospheric errors caused by static clutter during radar monitoring; A coordinate conversion unit is used to obtain the time history data of the arch rib in the transverse direction, along the bridge direction and vertical direction by performing spatial coordinate transformation on the time history data; The denoising module comprises: The data preprocessing unit is used to preprocess the time history data of the components in each direction of the arch rib, remove abnormal values ​​and fill in missing values; The wavelet denoising unit is used to decompose the signal into low-frequency and high-frequency parts of different scales, perform threshold processing on the high-frequency part to remove noise, and then reconstruct the processed low-frequency signal and the denoised high-frequency signal; The artificial intelligence prediction module includes: The model unit selects appropriate kernel functions and parameters for the denoised time-course data to construct a prediction model; A training unit, which is used to train the prediction model using the training set; Evaluation unit, used to evaluate model performance on the test set, including accuracy, recall, and F1 value; The prediction unit is used to use the trained model to make predictions on new data based on the evaluation results and explain the results.

Images