A pile foundation deformation distribution type testing method and system

Abstract

The present application relates to the technical field of pile foundation deformation measurement, in particular to a kind of pile foundation deformation distributed test method and system, the method comprises: obtaining the time series strain value sequence of the sampling point of tower pile foundation, calculate dynamic disturbance index and extract quasi-static signal, determine long-term strain value increment by time domain trend analysis;Through space-time decoupling operation, the correlation of the time evolution behavior of the sampling point and its physically adjacent sampling point is compared, to determine the spatial localization index;The spatial localization index is used as the confidence weight, and the long-term strain value increment is weighted and distributed to determine the adaptive global trend component and the local abnormal component respectively;Finally, spatial integration operation is respectively performed on global trend component and local abnormal component, to obtain two decoupled deformation curves to determine the deformation state.The method improves the accuracy of deformation test.

Description

Technical Field

[0001] This invention relates to the field of pile foundation deformation measurement technology. Specifically, it relates to a distributed testing method and system for pile foundation deformation. Background Technology

[0002] The pile foundations of tall structures such as power transmission towers and communication towers in coastal areas are fundamental to ensuring the safe and stable use of these structures, making the monitoring of their deformation status crucial. Distributed fiber optic sensing technology, particularly BOTDR (Optical Time Domain Reflectometry) based on Brillouin scattering, utilizes sensing fibers pre-embedded inside the pile foundation or bonded to its surface to measure deformation through strain distribution profiles formed by the strain values ​​at various sampling points along the pile's length.

[0003] BOTDR-based deformation testing mainly relies on two core steps: First, by measuring the Brillouin frequency shift and performing temperature compensation, the strain value is calculated based on the strain sensing coefficient. Second, by spatially integrating the strain profile along the length of the pile foundation, the total deformation or displacement is obtained.

[0004] However, the pole and pile foundations in coastal areas are exposed to high sea wind loads, high salt spray, and high humidity for a long time. The pole and pile foundations will not only experience high-frequency strain due to the high-frequency sea wind loads, but also suffer local damage due to the corrosive environment. The rust products produced by corrosion expand in volume and exert compressive pressure on the surrounding concrete, thereby causing local strain in the pole and pile foundations.

[0005] Therefore, the strain value obtained by calculating the strain value based on the strain sensing coefficient is actually a complex composite signal. It mixes the extremely weak corrosion strain with the huge dynamic strain of wind load and the global strain value generated by the actual foundation settlement. The deformation test results are easily dominated and submerged by strong interference signals, which reduces the sensitivity of the test system and leads to inaccurate deformation test results. Summary of the Invention

[0006] To address the problem that the strain value obtained by existing BOTDR-based deformation testing is a complex signal, mixing extremely weak corrosion strain with the massive dynamic strain of wind load and the global strain generated by actual foundation settlement, resulting in strong interference signals dominating and drowning out the real corrosion signal, thus reducing the accuracy of deformation testing, this invention proposes a distributed testing method and system for pile foundation deformation.

[0007] In a first aspect, the present invention provides a distributed testing method for pile foundation deformation, comprising:

[0008] By using fiber optic sensors deployed on the surface of the tower pile foundation, strain values ​​at multiple sampling points along the pile foundation axis at multiple times are obtained to form a time-series strain value sequence for each sampling point.

[0009] For each sampling point, the dynamic disturbance index of the sampling point is calculated based on the time-series strain value sequence of the sampling point. The quasi-static signal of the sampling point is extracted from the time-series strain value sequence using the dynamic disturbance index. Time-domain trend analysis is performed on the quasi-static signal to determine a long-term rate of change. Combined with the time length corresponding to the time-series strain value sequence, the long-term strain value increment of the sampling point is determined.

[0010] Spatiotemporal decoupling is performed on the quasi-static signals of all sampling points. The correlation between the strain value of each sampling point and the strain value of its physical neighboring sampling points over time is compared to determine the spatial localization index of that sampling point.

[0011] Using the spatial localization index as a confidence weight, a weighted allocation operation is performed on the long-term strain value increment to determine the adaptive global trend component and local anomaly component of the sampling point, respectively.

[0012] The deformation curve is determined based on the adaptive global trend component and local anomaly component of all sampling points to achieve distributed deformation testing of tower pile foundations.

[0013] This technical solution first filters out wind load interference by calculating a dynamic disturbance index that incorporates local spatial correlation in the time domain. Then, it uses two parallel analysis paths: one is to determine the mixed long-term strain value increment through time domain trend analysis, and the other is to determine the spatial localization index through spatiotemporal decoupling analysis. Finally, it uses the spatial localization index to weight and distribute the long-term strain value increment, completely separating the adaptive global trend component and the local anomaly component, thus solving the problem of strain parameter contamination and improving the accuracy of deformation testing.

[0014] Preferably, the dynamic disturbance index of the sampling point is determined as follows: The time-series strain value sequence of each sampling point is obtained, and the normalized standard deviation of the time-series strain value sequence is used as the time volatility index of that sampling point; the first time-series strain value sequence of the preceding adjacent sampling point and the second time-series strain value sequence of the following adjacent sampling point are obtained; the first Pearson correlation coefficient and the second Pearson correlation coefficient between the time-series strain value sequence of that sampling point and the first and second time-series strain value sequences are calculated respectively; the average of the first and second Pearson correlation coefficients is determined as the local spatial correlation index of that sampling point; the time volatility index and the local spatial correlation index are nonlinearly fused to obtain the dynamic disturbance index of that sampling point.

[0015] This technical solution achieves an AND gate operation through nonlinear fusion. The final dynamic disturbance index will only be larger when a signal simultaneously satisfies high temporal volatility and high local spatial correlation. This allows the dynamic disturbance index to accurately identify strain characteristics caused by wind loads, while also eliminating sensor noise with the same high volatility but lacking spatial correlation.

[0016] Preferably, the method for extracting the quasi-static signal of the sampling point from the time-series strain value sequence using the dynamic perturbation index is as follows: the product of the normalized root mean square of the time-series strain value sequence of each sampling point and the dynamic perturbation index of the sampling point is negatively correlated and nonlinearly mapped to determine a suppression factor; the suppression factor is then used to multiply the time-series strain value sequence of the sampling point element by element to obtain a new time-series strain value sequence, which is used as the quasi-static signal of the sampling point.

[0017] This technical solution physically constructs an adaptive time-domain filter. It introduces the root mean square as an indicator of the total energy of the signal and multiplies this indicator by the dynamic disturbance index to construct a physical quantity of dynamic disturbance energy. Through a negatively correlated nonlinear mapping, this energy is mapped into a suppression factor, which can accurately suppress signal components identified as strong dynamic interference, while retaining all non-dynamic quasi-static signals.

[0018] Preferably, the method for performing time-domain trend analysis on quasi-static signals to determine a long-term rate of change is as follows: In the quasi-static signal at each sampling point, the gradient direction and gradient magnitude of each strain value are determined based on each strain value and its adjacent strain values ​​to obtain a set of gradient directions and a set of gradient magnitudes; statistical analysis is performed on the set of gradient magnitudes to determine a gradient magnitude threshold; the set of gradient magnitudes is filtered by the gradient magnitude threshold to obtain a subset of gradient magnitudes; the gradient direction subset corresponding to the subset of gradient magnitudes is determined in the set of gradient directions; a weighted average operation is performed on the subset of gradient magnitudes using the gradient direction subset as a weight to obtain a long-term rate of change.

[0019] This technical solution performs statistical analysis on the gradient magnitude set to determine a gradient magnitude threshold and filter it. First, it identifies and removes all contaminated data points that are unlikely to be true trends. It only performs a weighted average operation on the remaining clean gradient subset. The resulting long-term rate of change is a highly robust estimate that is not affected by any instantaneous outliers or sensor jumps, and accurately reflects the true physical change trend.

[0020] Preferably, determining the gradient direction and gradient magnitude of each strain value based on each strain value and its preceding and following strain values ​​includes: for each strain value, calculating the difference between the next adjacent strain value and the previous adjacent strain value, determining the sign of the difference as the gradient direction of the strain value, and determining the absolute value of the difference as the gradient magnitude of the strain value.

[0021] Preferably, the long-term strain value increment of the sampling point is determined as follows: the product of the long-term change rate and the time length corresponding to the time series strain value sequence is determined as the long-term strain value increment of the sampling point.

[0022] Preferably, a spatiotemporal decoupling operation is performed on the quasi-static signal at all sampling points to determine the spatial localization index of that sampling point, including: for the first... The sampling points are used to construct the first sampling point. Local spatiotemporal observation vector of each sampling point :

[0023] ,in, , and The first The quasi-static signal of the previous sampling point of the sampling point, the first sampling point, The quasi-static signal at the sampling point and the... The quasi-static signal of the next sampling point after the sampling point; calculate The local covariance matrix varies over time, with the time length corresponding to the quasi-static signal at that sampling point. Eigenvalue decomposition is performed on this local covariance matrix to obtain three eigenvalues. , and ,and , No. Spatial localization index of each sampling point ,in, To prevent parameters with a denominator of 0.

[0024] This technical solution takes into account the essential difference between the settlement characteristics of the tower pile foundation itself and the local corrosion characteristics. By constructing a local spatiotemporal observation vector and calculating the local covariance matrix generated by the evolution of this vector over time, eigenvalue decomposition is performed. The eigenvalues ​​reflect the difference between the settlement of the tower pile foundation itself and local corrosion. The final spatial localization index can accurately distinguish whether the strain value collected at a sampling point is caused by the settlement of the tower pile foundation itself or by physical corrosion.

[0025] Preferably, a weighted allocation operation is performed on the long-term strain value increment to determine the adaptive global trend component and local anomaly component of the sampling point, including:

[0026] The local anomaly component is calculated by weighting the long-term strain value increment with confidence weights; the adaptive global trend component is calculated by weighting the long-term strain value increment with the value obtained by subtracting the confidence weights from 1.

[0027] Preferably, the deformation curve is determined based on the adaptive global trend component and local anomaly component of all sampling points to achieve distributed deformation testing of the tower pile foundation, including:

[0028] The adaptive global trend components of all sampling points are obtained to form an adaptive global trend component sequence. A first spatial integration operation is performed on the adaptive global trend component sequence to obtain a global deformation curve representing the overall deformation state of the tower and pile foundation. The local anomaly components of all sampling points are obtained to form a local anomaly component sequence. A second spatial integration operation is performed on the local anomaly component sequence to obtain a local deformation curve representing the deformation caused by local damage to the tower. The global deformation curve and the local deformation curve are decoupled and diagnosed to jointly determine the deformation state of the tower and pile foundation.

[0029] Secondly, the present invention also provides a distributed pile foundation deformation testing system, the distributed pile foundation deformation testing system including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of any of the distributed pile foundation deformation testing methods.

[0030] The present invention has the following effects:

[0031] In the time domain, this invention utilizes a dynamic disturbance index that incorporates local spatial correlation to accurately filter out dynamic strain interference caused by wind loads. In the spatial domain, it separates the deformations representing foundation settlement and actual corrosion. This strategy of separating interference layer by layer and decoupling and allocating it using spatiotemporal evolution characteristics solves the problem of strong interference dominating test results and improves the accuracy of deformation testing. Attached Figure Description

[0032] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] Reference Figure 1 A distributed testing method for pile foundation deformation specifically includes the following steps:

[0035] S1: Obtain the time-series strain value sequence for each sampling point along the pile foundation axis.

[0036] Using Brillouin scattering-based optical time-domain reflectometry (BOTDR), sensing optical fibers are laid along the axial direction of the tower pile foundation from the pile top to the pile bottom. Data is continuously collected through a BOTDR demodulator at a preset time sampling rate, once per hour, with a sampling point every 0.5 meters. The strain value of each sampling point along the pile foundation axial direction at each moment is obtained. Up to the current test moment, the strain values ​​of each sampling point over the past month are used to construct a time-series strain value sequence for that sampling point.

[0037] S2: Determine the dynamic disturbance index for each sampling point, and use the dynamic disturbance index to extract the quasi-static signal of that sampling point from the time-series strain value sequence.

[0038] For pole and pile foundations in coastal areas, the strain value is a complex composite signal due to the influence of sea winds and corrosive environment, mainly consisting of four components: wind load interference, sensor noise, global foundation settlement, and local physical corrosion. Among them, wind load and sensor noise manifest as high-frequency interference, while settlement and corrosion manifest as low-frequency quasi-static signals.

[0039] In one embodiment, the dynamic perturbation index for each sampling point is determined based on the following method:

[0040] First, obtain the number Time-series strain value sequence of sampling points And the time-series strain value sequence of the previous adjacent sampling point. and the time-series strain value sequence of the next adjacent sampling point These are denoted as the first time-series strain value sequence and the second time-series strain value sequence, respectively.

[0041] Then, calculate The standard deviation, and obtain all The maximum and minimum values ​​of the standard deviation of the time-series strain value sequence for each sampling point are used to normalize the values ​​using the maximum and minimum values. Normalize the standard deviation to obtain Normalized standard deviation ,Will As an indicator of the time volatility of this sampling point, the following parameters were obtained according to standard Pearson correlation analysis: and Pearson correlation coefficient between ,as well as and Pearson correlation coefficient between Let these be denoted as the first Pearson correlation coefficient and the second Pearson correlation coefficient, respectively. The average of the first and second Pearson correlation coefficients is determined as the local spatial correlation index for that sampling point. .

[0042] Specifically:

[0043] Calculate the first Pearson correlation coefficient The process is as follows:

[0044] Let the first The time-series strain value sequence of each sampling point is as follows: , For the first The first strain value in the time-series strain value sequence of sampling points. For the first The total number of strain values ​​contained in the time-series strain value sequence of each sampling point. For the first The time-series strain value sequence of the sampling point is the first Each strain value.

[0045] Meanwhile, the first strain value sequence is The second strain value sequence is as follows: ;

[0046] calculate and The degree of linear correlation between them, i.e., the first Pearson correlation coefficient :

[0047]

[0048] in, yes , The sequence number of each included strain value, because and If the sequence is one-to-one, then the same index can be used for traversal. yes and The total number of strain values ​​contained in each. for The Each strain value, for The Each strain value, for The average value, for The average value.

[0049] Similarly, using the above formula, to Replace with The second Pearson correlation coefficient can then be calculated, which is used to quantify the synchronicity of two physically adjacent sampling points in terms of their temporal evolution trends.

[0050] Finally, the time volatility indicator Indicators related to local space Perform nonlinear fusion to obtain the first... Dynamic disturbance index of each sampling point Specifically, in a concrete implementation, nonlinear fusion can be achieved using multiplication and the hyperbolic tangent function:

[0051]

[0052] in, For the first Dynamic disturbance index of each sampling point It is the hyperbolic tangent function. For the first Time volatility index of each sampling point For the first Local spatial correlation indicators for each sampling point.

[0053] At the sampling points at the top and bottom of the pile, when obtaining local spatial correlation indicators, the time-series strain value sequence of the sampling point and the time-series strain value sequence of the unique adjacent sampling point of the sampling point are obtained. Then, based on the time-series strain value sequences of the two, a Pearson correlation coefficient is calculated as the local spatial correlation indicator of the sampling point.

[0054] The dynamic disturbance index for each sampling point is calculated in this way to accurately eliminate the influence of wind load interference. In the formula for calculating the dynamic disturbance index, This represents the total fluctuation of all signal components (including high-frequency wind load, high-frequency noise, and low-frequency settlement / corrosion) in a physically represented manner. Wind load interference exhibits high-frequency fluctuation characteristics. The values ​​are very high, and since wind load is a physical excitation, it inevitably causes spatially correlated vibrations in the pile foundation as a continuous body. Therefore, its local spatial correlation index is also very high. The value will be very high, approaching 1, and the dynamic disturbance index will be even larger. Sensor noise also exhibits high-frequency fluctuation characteristics. The noise level is high, but the sensor noise is a random jump due to spatial decoupling, with almost no spatial correlation characteristics. The values ​​will be very low, resulting in a significant suppression of the final dynamic disturbance index. Regarding the physical settlement or corrosion of the tower pile foundation itself, this physical settlement or corrosion is usually caused by the pile foundation as a whole, with local spatial related indicators... The value will be high, but because the process of physical sedimentation or physical corrosion is usually an extremely slow quasi-static change, it has low-frequency characteristics. The dynamic disturbance index will be very small, resulting in a significant suppression of the final dynamic disturbance index. Therefore, the larger the dynamic disturbance index of a sampling point, the more likely the strain value at the sampling point is to include the strain value generated by the interference of wind load. Conversely, the smaller the dynamic disturbance index of a sampling point, the more likely the strain value at the sampling point is to include the strain value generated by sensor noise or physical settlement or physical corrosion of the tower pile foundation. This provides a reliable basis for accurately extracting the quasi-static signal.

[0055] In one embodiment, the method for extracting the quasi-static signal of sampling points from a time-series strain value sequence using a dynamic perturbation index is as follows:

[0056] First, identify a repressor:

[0057] Calculate the first The root mean square (RMS) of the time-series strain values ​​at each sampling point is obtained. The maximum and minimum values ​​among the RMS of the time-series strain values ​​at all sampling points are then normalized using the maximum-minimum normalization method. Perform a normalization operation to obtain the first... The normalized root mean square of each sampling point is used to perform a negatively correlated nonlinear mapping on the product of the dynamic perturbation index and the normalized root mean square of that sampling point to determine a suppression factor. Specifically, this nonlinear mapping can be implemented using the natural exponential function:

[0058]

[0059] in, For the first The inhibition factor at each sampling point It is a natural exponential function. For the first The inhibition factor at each sampling point For the first The normalized root mean square of the time-series strain value sequence of each sampling point.

[0060] In this formula, The larger the value, the more likely it is to be the first. The larger the strain value at each sampling point, the more likely it is to include strain values ​​caused by wind load disturbance; conversely, the smaller the value, the more likely it is to include sensor noise or a true quasi-static signal. The root mean square (RMS) is a standard method in signal processing for measuring the total energy or average power of a time series. It reflects the magnitude of the total energy intensity of the signal; a severe vibration will have a high RMS value. A stable signal will have a low point. , It is a negatively correlated nonlinear mapping, which will The input value in the range is mapped to a... Output value for the interval. Part of it involves calculating a dynamic perturbation energy. Characterizing the total energy of the signal, It is a number in the range of 0 to 1. This is equivalent to the portion of the total energy contributed by the wind load.

[0061] In this formula, the stronger the wind load, the better. The larger, The larger the inhibitory factor, the stronger the inhibition factor. The smaller the value, the more likely the strain value at that sampling point is to include strain values ​​generated by wind load, thus requiring a smaller value. The strain value at this sampling point is suppressed. The weaker the wind load, The smaller, The smaller the value, the stronger the inhibitory factor. The larger the value, the less likely the strain value at that sampling point is to include strain values ​​caused by wind loads, and the more likely it is to include strain values ​​caused by sensor noise, actual settlement of the tower and pile foundations, or physical corrosion. Therefore, a larger value is required. The strain value at this sampling point is retained.

[0062] Then, With the The time-series strain value sequences of each sampling point are multiplied element-wise to obtain a new time-series strain value sequence, which is used as the quasi-static signal of that sampling point. When the dynamic disturbance energy is high, the suppression factor approaches 0, achieving strong suppression; when the dynamic disturbance energy is low, the signal is stable, and the suppression factor approaches 1, preserving the signal to the greatest extent. In this way, an adaptive time-domain filter is constructed, which can accurately suppress signal components identified as strong dynamic interference, while preserving all non-dynamic quasi-static signals.

[0063] S3: By performing time-domain trend analysis on the quasi-static signal at each sampling point, a long-term rate of change is determined, and the long-term strain increment at that sampling point is determined in combination with the time length.

[0064] Since the quasi-static signal at each sampling point may still physically contain two components: one is the real, slow long-term trend, such as the foundation settlement and physical corrosion of the tower pile foundation itself; the other is sensor noise, which needs to be further eliminated to obtain the real strain value increment.

[0065] This step takes into account that sensor noise is usually instantaneous and has a high amplitude, while the settlement or physical corrosion of tower pile foundations is usually a long-term trend and has a lower amplitude.

[0066] Therefore, by analyzing the quasi-static signals at the sampling points and calculating the changes in gradient magnitude and gradient direction, all noise interference that could not be the true trend is physically eliminated. The strain value change process is analyzed based only on the remaining clean and slow data, thus obtaining a long-term strain value increment that is not affected by noise.

[0067] In one embodiment, the method for performing time-domain trend analysis on the quasi-static signal at each sampling point to determine a long-term rate of change is as follows:

[0068] First, in the quasi-static signal at each sampling point, the gradient direction and gradient magnitude of each strain value are determined based on each strain value and its preceding and following strain values, including:

[0069] For each strain value, calculate the difference between the next adjacent strain value and the previous adjacent strain value, as well as the time interval between the next adjacent strain value and the previous adjacent strain value. The sign (positive or negative) of the ratio of the difference and the time interval is used as the gradient direction of the strain value, and the absolute value of the ratio of the difference and the time interval is used as the gradient magnitude of the strain value.

[0070] For example, for the first The quasi-static signal at the nth sampling point strain value The previous adjacent strain value The next adjacent strain value Since the sampling frequency is fixed, it is collected once every hour, that is... ,but The gradient direction is: ,

[0071] The gradient magnitude is , This is a sign function. When the value inside the parentheses is positive, the sign function evaluates to 1; when the value inside the parentheses is negative, the sign function evaluates to -1; and when the value inside the parentheses is 0, the sign function evaluates to 0. It is the absolute value symbol. yes and The time interval between them.

[0072] This is a central difference method. For strain values ​​at the endpoints of a quasi-static signal, forward or backward difference methods can be used for adaptation. The reason for using the central difference method is that it cleverly averages out local noise fluctuations by analyzing the symmetrical adjacent strain values ​​on both sides of each strain value, providing second-order accuracy. The result is more accurate, providing a more stable and accurate reflection of the direction and severity of change at each strain value.

[0073] Then, the gradient directions and gradient magnitudes of all strain values ​​of the quasi-static signal at each sampling point are used to form a gradient direction set and a gradient magnitude set. Statistical analysis is performed on the gradient magnitude set to determine a gradient magnitude threshold. Specifically, the standard deviation and mean of the gradient magnitude set are calculated. Using the rule of three standard deviations in statistics, the mean plus three standard deviations is set as the gradient magnitude threshold. Gradient magnitudes greater than this threshold are filtered out to obtain a subset of the remaining gradient magnitudes.

[0074] Obtain the gradient direction subset corresponding to the gradient magnitude subset from the gradient direction subset, and perform a weighted average operation on the gradient magnitude subset using the gradient direction subset as weight to obtain a long-term rate of change.

[0075] The weighted average operation is specifically as follows:

[0076]

[0077] in, It is the first The long-term rate of change of strain values ​​at each sampling point is the total number of gradient magnitudes contained in the gradient magnitude subset (without deduplication), and 'o' is the index of the gradient magnitude in the gradient magnitude subset. It is the 0th gradient magnitude in the subset of gradient magnitudes. It is the gradient direction corresponding to the 0th gradient magnitude. First, the gradient magnitudes are weighted and summed using the gradient direction, and then the average is calculated to achieve the weighted average operation. Physically, it represents the average long-term rate of change of the sampling point, undisturbed by any noise. For example, the strain value at the sampling point changes daily at a rate of -0.005 microstrain values.

[0078] In one embodiment, the long-term strain increment at the sampling point is determined based on the following method:

[0079] The product of the long-term rate of change and the corresponding time length of the static signal is used to determine the long-term strain value increment at the sampling point. For example, if the long-term rate of change is -0.01 strain value / day and the time length is 365 days, then the long-term strain value increment is -3.65 strain value.

[0080] In this way, the total deformation of each sampling point during the entire monitoring period, contributed solely by the long-term trend, is obtained, which is the long-term strain value increment. This long-term strain value increment is the total amount of strain value generated by the settlement of the tower pile foundation and physical corrosion, and is the basis for subsequent spatial analysis.

[0081] S4: By performing a spatiotemporal decoupling operation on the quasi-static signal of all sampling points, the correlation between the temporal evolution behavior of the sampling point and its physically adjacent sampling points is compared to determine the spatial localization index of the sampling point.

[0082] After obtaining a long-term strain value increment that combines the strain values ​​generated by the settlement and physical corrosion of the tower and pile foundation, the signal is decoupled based on the evolution characteristics of the signal throughout the time domain to distinguish between the settlement and physical corrosion of the tower and pile foundation.

[0083] Specifically, the settlement of tower pile foundation is a relatively holistic and global change. If settlement occurs, the sampling points at different locations on the pile foundation are spatially related and usually move together. Therefore, the strain values ​​at the sampling points also have a certain spatial correlation and produce coordinated changes.

[0084] The correlation between the temporal evolution of each sampling point and its physical neighbors is based on the coherence principle of deformation modes. The global settlement of tower pile foundations is a low-degree-of-freedom common-mode motion, which causes multiple sampling points in the local area to exhibit highly synchronous fluctuation trends on the time axis. That is, their temporal evolution has a very strong linear correlation. In contrast, local corrosion damage is a high-degree-of-freedom differential-mode motion, which only causes abnormal expansion of specific sampling points, while the impact on its neighboring points has lag or attenuation, resulting in the signal in the local area exhibiting uncorrelated characteristics on the time axis.

[0085] Therefore, by constructing local spatiotemporal observation vectors and analyzing their covariance matrix, we are essentially evaluating in high-dimensional space whether the distribution direction of signal energy is concentrated in the main direction representing common-mode motion or diverges to the secondary direction representing differential motion, thereby achieving spatiotemporal decoupling.

[0086] In one embodiment, the method for determining the spatial localization index of each sampling point through spatiotemporal decoupling operations includes:

[0087] For each sampling point, a local spatiotemporal observation vector is constructed. This vector consists of the quasi-static signal of that sampling point and the quasi-static signals of its physically adjacent sampling points. Specifically, the... Local spatiotemporal observation vector of each sampling point Its composition is as follows:

[0088] ,in, For the first Quasi-static signal at one sampling point For the first The quasi-static signal of the next sampling point after the previous sampling point. For the first The quasi-static signal of the previous sampling point of each sampling point.

[0089] calculate The local covariance matrix changes over time, and the time length corresponding to this change is the time length corresponding to the quasi-static signal. Since the acquisition time of each sampling point is consistent, the length of the quasi-static signal is also consistent.

[0090] Perform eigenvalue decomposition on the local covariance matrix to obtain at least three eigenvalues. Take the first three eigenvalues, and their magnitude relationships are as follows: Calculate the first Spatial localization index of each sampling point :

[0091]

[0092] in, To prevent parameters from having a denominator of 0, they are usually set to a very small positive number. Eigenvalue decomposition is essentially principal component analysis of the local covariance matrix. These are the eigenvalues ​​of the covariance matrix. Represents the first The sampling point, the first The sampling point and the first The total energy of the quasi-static signal in a local area composed of sampling points is used to decouple the signal based on its evolution characteristics (covariance) over the entire time domain, in order to distinguish between global settlement and local physical corrosion of tower pile foundations. Since settlement is spatially dependent, the first... The sampling point, the first The sampling point and the first The energy of the quasi-static signal at each sampling point will be concentrated in the first principal component. Above, while localized physical corrosion is spatially uncorrelated, energy will leak into the second principal component. and the third principal component .

[0093] In the above eigenvalue decomposition, It is the largest eigenvalue, corresponding to the first principal component of the local covariance matrix. Physically, the first principal component captures the direction of maximum variance in the local observation vector, i.e., the common and strongly correlated trend of change among the three sampling points. Therefore, It characterizes the energy intensity of global settling (common mode component). and These are secondary eigenvalues, corresponding to directions orthogonal to the first principal component. Physically, they capture the inconsistent, asynchronous differences between the three sampling points. It characterizes the energy intensity of local anomalies (differential mode components).

[0094] The spatial localization index proposed in this invention is essentially a differential energy proportion coefficient. When pure global settlement occurs, the first... Each sampling point is highly synchronized with its neighboring sampling points, and the local covariance matrix tends to be a rank-1 matrix, with almost all energy concentrated in the area between the sampling points and the neighboring sampling points. ,at this time The index approaches 0, causing the spatial localization index to approach 0. When localized corrosion occurs, the first... The behavior of each sampling point is decoupled from its neighboring sampling points, disrupting the correlation and causing energy to leak from the first principal component to the secondary principal components. The increase leads to an increase in the spatial localization index, which then tends towards 1.

[0095] This spatiotemporal decoupling operation based on characteristic spectrum analysis not only qualitatively identifies damage types but also quantitatively provides a continuous physical confidence level (ranging from 0 to 1). It can adaptively separate mixed strain increments into settlement and corrosion components without prior knowledge, thus improving the accuracy of test results under complex working conditions.

[0096] Therefore, the spatial localization index has become a highly reliable physical confidence level. The smaller the value, the closer it is to 0, the more likely the strain value at the sampling point is to include the strain value generated by global settlement. The larger the value, the more likely the strain value at the sampling point is to include the strain value generated by local corrosion.

[0097] S5: Using the spatial localization index as a confidence weight, perform a weighted allocation operation on the long-term strain value increment to determine the adaptive global trend component and local anomaly component of the sampling point respectively.

[0098] In one embodiment, the weighted allocation operation includes:

[0099] For the Each sampling point will be used to determine the spatial localization index. As the confidence weight and the first The long-term strain increments at each sampling point are multiplied to achieve a weighted calculation, and the product is used as the local anomaly component of the strain value at that sampling point; With the The long-term strain value increments of each sampling point are multiplied to achieve a weighted calculation, and the product is used as the first... The adaptive global trend component of the strain value at each sampling point.

[0100] The reason for adopting this method is that the long-term strain increment at each sampling point is a mixed value, containing strain values ​​caused by tower and pile foundation settlement and strain values ​​caused by physical corrosion. The spatial localization index, which reflects the probability of actual physical corrosion, is a value in the range of 0 to 1. Using it as a confidence weight, a reasonable weighting operation is performed on the long-term strain increment. If the strain value at a sampling point is more likely to contain strain values ​​caused by actual physical corrosion, then the spatial localization index of that sampling point is larger, the confidence weight is larger, and it plays a dominant role in the long-term strain increment at that sampling point. The local anomaly component of the strain value at that sampling point will be larger because the local anomaly component reflects the characteristic of physical corrosion. At the same time, the adaptive global trend component of the strain value at that sampling point will be smaller because the adaptive global trend component reflects the global strain influence of tower and pile foundation settlement, and vice versa.

[0101] This method enables signal separation with clear physical meaning.

[0102] S6: Integrate the adaptive global trend component and local anomaly component of all sampling points to obtain two deformation curves. Analyze the deformation state of the tower pile foundation based on the deformation curves to complete the deformation test.

[0103] In one embodiment, adaptive global trend components of all sampling points are obtained to form an adaptive global trend component sequence. Spatial integration is performed on the adaptive global trend component sequence to obtain a global deformation curve that can represent the overall deformation state of the tower and pile foundation. At the same time, a local anomaly component sequence composed of local anomaly components of all sampling points is obtained. Spatial integration is performed on the local anomaly component sequence to obtain a local deformation curve that can represent the deformation of the tower and pile foundation caused by local damage (corrosion). Decoupled diagnosis is performed based on these two curves to jointly assess the deformation state of the tower and pile foundation.

[0104] Specifically, it includes:

[0105] Obtain the distribution along the pile foundation axis The adaptive global trend components at each sampling point constitute an adaptive global trend component sequence. , The total number of sampling points. For the adaptive global trend component of the first sampling point, For the first An adaptive global trend component for each sampling point.

[0106] Since the sampling points are discretely distributed, the first spatial integration operation is calculated using the discrete cumulative summation formula. For the... sampling points ( The corresponding global cumulative deformation is:

[0107]

[0108] in, For the first The global cumulative deformation at each sampling point, i.e., from the starting point to the nth sampling point. The cumulative displacement at each sampling point location, For the front The sequence number of each sampling point, for example... hour, For the front The index of each sampling point is any integer within the interval [1, 5]. For the first The adaptive global trend component (dimensionless strain value) of each sampling point. For the first The physical distance (in meters) between each sampling point and the previous sampling point.

[0109] Using the index of each sampling point as the x-axis, and the index of each sampling point as the y-axis... Using the vertical axis as the ordinate, all sampling points By connecting them sequentially, a global deformation curve is obtained. This curve represents the common motion mode of the pile foundation and is used to analyze the macroscopic deformation state of the pile foundation. By observing this curve, the macroscopic deformation state of the pile foundation can be clearly analyzed without any interference from local noise or corrosion points, so as to determine the extent of the global settlement of the pile foundation and whether the pile body has tilted as a whole.

[0110] At the same time, obtain the distribution along the pile foundation axis. The local anomaly components at each sampling point constitute a local anomaly component sequence. , The total number of sampling points. This refers to the local anomaly component of the first sampling point. For the first Local anomaly components at each sampling point.

[0111] Regarding the first sampling points ( The corresponding local cumulative deformation is:

[0112]

[0113] in, For the first Local cumulative deformation at each sampling point For the front The serial number of each sampling point For the first Local anomaly components (dimensionless strain values) at each sampling point. For the first The physical distance (in meters) between each sampling point and the previous sampling point.

[0114] Using the index of each sampling point as the x-axis, and the index of each sampling point as the y-axis... Using the vertical axis as the ordinate, all sampling points By connecting them sequentially, a local deformation curve is obtained. This curve represents the differential motion mode of the pile foundation and is used to reflect the deformation state caused by local damage to the pile foundation.

[0115] Since the massive global settlement component has been separated into the global deformation curve, the value of the local deformation curve is close to zero in most healthy areas. If the local deformation curve shows an isolated, sharp displacement abrupt change at a specific location, it is determined that there is local damage caused by steel corrosion expansion or cracks at that location, and the amplitude of the abrupt change characterizes the degree of deformation caused by the local damage.

[0116] In summary, by analyzing the global deformation curve to assess the settlement of the tower and pile foundation, and by analyzing the local deformation curve to determine the degree of local damage to the tower and pile foundation with high sensitivity, this decoupled diagnostic method solves the problems of signal submersion and low sensitivity, and improves the accuracy of deformation testing.

[0117] A second aspect of this invention provides a distributed pile foundation deformation testing system, comprising a processor and a memory, wherein the memory stores a computer program, for example, instructions for time-domain filtering, spatial decoupling, and state assessment. When the processor executes the computer program stored in the memory, it implements steps such as those in a distributed pile foundation deformation testing method. The distributed pile foundation deformation testing system also includes a data interface for connecting to a fiber optic sensor demodulator such as a BOTDR to acquire a time-series strain value sequence in step S1.

[0118] The processor can be a central processing unit, a digital signal processor, or a field-programmable gate array, and the memory can be volatile or non-volatile, such as read-only memory (ROM), hard disk drive (HDD), or solid-state drive (SSD).

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A distributed testing method for pile foundation deformation, characterized in that, include: By using fiber optic sensors deployed on the surface of the tower pile foundation, strain values ​​at multiple sampling points along the pile foundation axis at multiple times are obtained to form a time-series strain value sequence for each sampling point. For each sampling point, the dynamic disturbance index of the sampling point is calculated based on the time-series strain value sequence of the sampling point. The quasi-static signal of the sampling point is extracted from the time-series strain value sequence using the dynamic disturbance index. Time-domain trend analysis is performed on the quasi-static signal to determine a long-term rate of change. Combined with the time length corresponding to the time-series strain value sequence, the long-term strain value increment of the sampling point is determined. Spatiotemporal decoupling is performed on the quasi-static signals of all sampling points. The correlation between the strain value of each sampling point and the strain value of its physical neighboring sampling points over time is compared to determine the spatial localization index of that sampling point. Using the spatial localization index as a confidence weight, a weighted allocation operation is performed on the long-term strain value increment to determine the adaptive global trend component and local anomaly component of the sampling point, respectively. The deformation curve is determined based on the adaptive global trend component and local anomaly component of all sampling points to achieve distributed deformation testing of tower pile foundations.

2. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, The dynamic disturbance index of the sampling points is determined based on the following method: Obtain the time-series strain value sequence for each sampling point, and use the normalized standard deviation of the time-series strain value sequence as the time volatility index for that sampling point. Obtain the first time-series strain value sequence of the previous adjacent sampling point and the second time-series strain value sequence of the next adjacent sampling point; The first Pearson correlation coefficient and the second Pearson correlation coefficient between the time-series strain value sequence of the sampling point and the first time-series strain value sequence and the second time-series strain value sequence are calculated respectively. The average value of the first Pearson correlation coefficient and the second Pearson correlation coefficient is determined as the local spatial correlation index of the sampling point. The time fluctuation index and the local spatial correlation index are nonlinearly fused to obtain the dynamic disturbance index of the sampling point.

3. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, The method for extracting the quasi-static signal of a sampling point from a time-series strain value sequence using the dynamic perturbation exponent is as follows: A negatively correlated nonlinear mapping is performed on the product of the normalized root mean square of the time-series strain value sequence at each sampling point and the dynamic perturbation index at that sampling point to determine a suppression factor. The suppression factor is multiplied element-wise with the time-series strain value sequence of the sampling point to obtain a new time-series strain value sequence, which is used as the quasi-static signal of the sampling point.

4. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, The method for performing time-domain trend analysis on a static signal to determine a long-term rate of change is as follows: In the quasi-static signal at each sampling point, the gradient direction and gradient magnitude of each strain value are determined based on each strain value and its adjacent strain values ​​before and after it, so as to obtain a set of gradient directions and a set of gradient magnitudes. Statistical analysis is performed on the gradient magnitude set to determine a gradient magnitude threshold. The gradient magnitude set is then filtered using this threshold to obtain a subset of gradient magnitudes. In the set of gradient directions, determine the gradient direction subset corresponding to the gradient magnitude subset, and perform a weighted average operation on the gradient magnitude subset using the gradient direction subset as the weight to obtain a long-term rate of change.

5. The distributed testing method for pile foundation deformation according to claim 4, characterized in that, The gradient direction and gradient magnitude of each strain value are determined based on each strain value and its preceding and following strain values, including: For each strain value, calculate the difference between the next adjacent strain value and the previous adjacent strain value. Determine the sign of the difference as the gradient direction of the strain value, and determine the absolute value of the difference as the gradient magnitude of the strain value.

6. The distributed testing method for pile foundation deformation according to claim 4, characterized in that, The long-term strain increment at the sampling point is determined based on the following method: The product of the long-term rate of change and the time length corresponding to the time-series strain value sequence is determined as the long-term strain value increment at the sampling point.

7. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, Spatiotemporal decoupling is performed on the quasi-static signal at all sampling points. The correlation between the strain value of each sampling point and the strain values ​​of its physical neighboring sampling points over time is compared to determine the spatial localization index of that sampling point, including: for the ... The sampling points are used to construct the first sampling point. Local spatiotemporal observation vector of each sampling point ,in, , and The first The quasi-static signal of the previous sampling point of the sampling point, the first sampling point, The quasi-static signal at the sampling point and the... The quasi-static signal of the next sampling point after the sampling point; calculate The local covariance matrix varies with time, where the time length corresponds to the time length of the quasi-static signal at that sampling point. Eigenvalue decomposition is performed on this local covariance matrix to obtain three eigenvalues. , and ,and , No. Spatial localization index of each sampling point ,in, To prevent parameters with a denominator of 0.

8. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, A weighted allocation operation is performed on the long-term strain value increments to determine the adaptive global trend component and local anomaly component of the sampling point, including: The local anomaly component of the sampling point is calculated by weighting the long-term strain value increment with confidence weights; the adaptive global trend component of the sampling point is calculated by weighting the long-term strain value increment with the value obtained by subtracting the confidence weights from 1.

9. The distributed testing method for pile foundation deformation according to claim 1, characterized in that, Based on the adaptive global trend component and local anomaly component of all sampling points, the deformation curve is determined to achieve distributed deformation testing of the tower pile foundation, including: The adaptive global trend components of all sampling points are obtained to form an adaptive global trend component sequence. A first spatial integration operation is performed on the adaptive global trend component sequence to obtain a global deformation curve representing the overall deformation state of the tower and pile foundation. The local anomaly components of all sampling points are obtained to form a local anomaly component sequence. A second spatial integration operation is performed on the local anomaly component sequence to obtain a local deformation curve representing the deformation caused by local damage to the tower. The global deformation curve and the local deformation curve are decoupled and diagnosed to jointly determine the deformation state of the tower and pile foundation.

10. A distributed testing system for pile foundation deformation, characterized in that, The distributed pile foundation deformation testing system includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the distributed pile foundation deformation testing method as described in any one of claims 1-9.

Images