Mountainous highway construction area intelligent monitoring method

Abstract

The present application relates to a kind of mountainous highway construction area intelligent monitoring method, belong to data processing technical field.The method is obtained by the feature difference screening of extreme value and normal extreme value in stress data target extreme value, based on target extreme value with EMD algorithm decomposition stress sequence obtains initial component, then the time series stability of each initial component is further analyzed, it is decomposed to obtain stress component by adaptive repeated decomposition, and interference source is decomposed to obtain interference component, then the change characteristics of interference component and stress component are compared to determine the noise interference degree of each stress component, with the noise interference degree to stress component is filtered to obtain denoised stress data.The present application is combined with the scene of the scene of mountainous highway construction this kind of noise complex and changeable, and stress steady state also changes, and the data adaptive decomposition and component internal noise are targeted to remove, compared with existing fixed decomposition denoising strategy, can significantly improve denoising effect to improve stress monitoring accuracy.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent monitoring method for construction areas of mountainous highways. Background Technology

[0002] During the construction of highways in mountainous areas, structures such as bridges, temporary supports, and hoisting platforms are subjected to the combined effects of terrain undulations, wind loads, and construction loads, resulting in complex stress changes. These stress changes reflect the structural safety of these devices. Therefore, the ability to accurately monitor the stress on these devices is crucial for assessing their structural safety and stability to ensure construction safety.

[0003] Existing stress monitoring methods typically rely on sensors such as strain gauges and stress meters to collect stress data in real time. However, in mountainous construction environments, factors such as strong winds, temperature changes, mechanical vibrations, and electromagnetic interference can easily lead to signal fluctuations and data deviations. For example, thermal expansion and contraction caused by diurnal temperature variations can cause periodic drift in stress readings; vibrations from construction blasting and mechanical operations can cause signal peak distortion; and structural micro-movements under strong winds may also mask the true stress characteristics. These interferences reduce the accuracy and reliability of monitoring data, easily leading to misjudgments of the structural state by the system, and affecting construction safety early warning and decision-making.

[0004] To improve the accuracy of stress monitoring, existing methods use the EMD algorithm to decompose the signal and then denoise the component signals using wavelet thresholding. However, the existing denoising methods based on the EMD algorithm have fixed processing strategies, making it difficult to achieve adaptive decomposition and denoising based on the signal characteristics of different locations and environments. This makes it difficult to distinguish between noise and actual stress changes when facing scenarios such as construction of mountain highways, where noise interference is variable and stress itself is complex. As a result, the denoising results are not ideal, ultimately leading to insufficient accuracy in stress monitoring of construction equipment.

[0005] In other words, current stress monitoring methods based on EMD algorithms for denoising suffer from insufficient monitoring accuracy when facing complex environments such as mountainous highway construction. Summary of the Invention

[0006] In view of this, the present invention provides an intelligent monitoring method for construction areas of mountainous highways to solve the technical problem that current stress monitoring methods based on EMD algorithms cannot accurately monitor equipment stress in mountainous highway construction scenarios.

[0007] The present invention provides an intelligent monitoring method for construction areas of mountainous expressways, comprising:

[0008] Stress is collected at the target monitoring point to form a stress sequence, and interference signals are collected simultaneously. The same interference signals are used to form a corresponding interference sequence. The maximum and minimum values ​​of the stress sequence are classified into different extreme value classes. All extreme values ​​are clustered based on the similarity of changes between any two extreme values ​​in any extreme value class. The retention of a cluster is determined by the number of extreme points, the uniformity of distribution, and the similarity of changes in any cluster. The extreme values ​​in the clusters with a retention greater than the retention threshold are taken as target extreme values. Based on the target extreme values, the stress sequence is decomposed into initial component signals using the EMD algorithm.

[0009] The difference between two adjacent sliding windows is determined by the difference in amplitude and frequency change between any two adjacent sliding windows when the sliding window slides on any initial component signal. The stability of the initial component signal is determined by the dispersion and magnitude of all differences under the initial component signal. The initial component signal with stability less than the stability threshold is decomposed into a new initial component signal by the EMD algorithm until there is no initial component signal with stability less than the stability threshold. The initial component signal at this time is called the stress component signal.

[0010] Each interference sequence is decomposed into interference component signals using the EMD algorithm. The interference degree of the current stress component signal is determined by the feature similarity between the current stress component signal and each interference component signal. The denoising threshold of the current stress component signal is adjusted based on the interference degree of the current stress component signal to obtain the denoised component signal. After reconstructing all the denoised component signals into a denoised stress sequence, stress monitoring is completed.

[0011] Furthermore, the interference signals include ambient temperature signals, wind load signals, mechanical vibration intensity signals, and electromagnetic signals.

[0012] Furthermore, determining the similarity of changes at any two extreme values ​​within any extreme value class includes:

[0013] Determine the angle between any extreme point in any extreme value class and its two adjacent non-extreme extreme points. Calculate the normalized value of the absolute value of the difference between the angle values ​​corresponding to any two extreme points in any extreme value class as the local variation difference degree of the two extreme points. Calculate the normalized value of the absolute value of the difference between the amplitudes of the two extreme points as the amplitude difference degree of the two extreme points.

[0014] The variation similarity between any two extreme points is constructed based on the local variation difference and the amplitude difference, wherein the variation similarity is inversely proportional to both the local variation difference and the amplitude difference.

[0015] Furthermore, determining the retention degree of a cluster based on the number of extreme points, distribution uniformity, and variation similarity in any given cluster includes:

[0016] The normalized value of the number of extreme points in the stress sequence that do not belong to any particular cluster between the time corresponding to the first extreme point of any cluster and the time corresponding to the last extreme point of any cluster is used as the discontinuity of any particular cluster.

[0017] The normalized value of the coefficient of variation of the set of variation similarity formed by the variation similarity between any two extreme points in any cluster is used as the discreteness of any cluster.

[0018] The retention coefficient of any cluster is constructed based on the number of extreme points in any cluster, the discontinuity, and the dispersion. The retention coefficient is directly proportional to the number of extreme points in any cluster, and inversely proportional to both the discontinuity and the dispersion.

[0019] The retention degree of any cluster is determined by comparing its retention coefficient with the retention coefficients of all clusters in the extreme value class to which the cluster belongs.

[0020] Further, determining the retention rate of any of the said clusters includes:

[0021] Calculate the ratio of the retention coefficient of any cluster to the retention coefficient of each cluster in the extreme value class to which the cluster belongs, and use the normalized value of the mean of all the obtained ratios as the retention degree of the cluster.

[0022] Furthermore, determining the difference between two adjacent sliding windows includes:

[0023] The difference between the mean of the maximum and the mean of the minimum values ​​in any sliding window of the two adjacent sliding windows is calculated as the average extreme value difference of the sliding window. The normalized value of the absolute value of the difference of the average extreme value difference between the two adjacent sliding windows is used as the amplitude difference between the two adjacent sliding windows.

[0024] The normalized value of the mean square error between the spectral data corresponding to each of the two adjacent sliding windows is calculated as the frequency change difference between the two adjacent sliding windows.

[0025] The sum of the amplitude difference and the frequency change difference is used as the difference degree between the two adjacent sliding windows.

[0026] Furthermore, determining the stability of the initial component signal includes:

[0027] The coefficient of variation of the difference sequence formed by the difference degree corresponding to any two adjacent sliding windows in any initial component signal is calculated as the dispersion of all differences degree under the given initial component signal.

[0028] The stability of any initial component signal is constructed based on the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal. The stability of any initial component signal is inversely proportional to both the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal.

[0029] Furthermore, the feature similarity is:

[0030] ,

[0031] in, This represents the feature similarity between the current stress component signal S and the current disturbance component signal Y. This represents the mutual information value between the current stress component signal S and the current disturbance component signal Y. This represents the number of frequency intersection points between the current stress component signal S and the current interference component signal Y, that is, the number of frequency points in their respective spectral data that have the same frequency and whose corresponding amplitudes are all non-zero. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current interference component signal Y. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current stress component signal S. This represents the upper frequency limit of the spectral data of the current stress component signal S. This represents the lower frequency limit of the spectral data of the current stress component signal S. This represents the amplitude corresponding to the v-th frequency in the spectral data of the current stress component signal S.

[0032] Furthermore, determining the degree of interference of the current stress component signal includes:

[0033] The sum of the feature similarities between the current stress component signal and each interference component signal is taken as the interference degree of the current stress component signal.

[0034] Furthermore, the adjustment of the denoising threshold for the current stress component signal includes:

[0035] The product of the interference level of the current stress component signal and the preset mapping parameter is used as the denoising threshold of the current stress component signal.

[0036] The preset mapping parameters are obtained by adaptive training and correction using a Bayesian optimization algorithm based on sample data of the interference level and corresponding noise content of the determined stress component signals.

[0037] The advantages of this invention compared to the prior art are:

[0038] This invention first analyzes the consistency between the characteristics of extreme points and normal stress data in the stress data to select target extreme points. Based on the target extreme points, the stress sequence is decomposed using the EMD algorithm to obtain initial component signals. Then, the temporal stability of each initial component signal is further analyzed, and adaptive repeated decomposition is performed to obtain stress component signals. Interference sources are also decomposed to obtain interference component signals. By comparing the variation characteristics of the interference component signals and the stress component signals, the noise interference level of each stress component signal is determined. The stress component signals are then subjected to targeted adaptive filtering based on the noise interference level, and the filtered stress component signals are reconstructed into denoised stress data to achieve more accurate stress monitoring. This invention addresses the scenario of multi-source and diverse noise in mountainous highway construction areas, where the steady-state stress data also varies. It combines adaptive data decomposition with targeted noise removal, significantly improving the denoising effect of stress data and thus enhancing the accuracy of stress monitoring. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating an intelligent monitoring method for construction areas of mountainous highways provided in Embodiment 1 of the present invention. Detailed Implementation

[0041] The overall concept of this invention is as follows:

[0042] Simultaneously with the acquisition of stress sequences, interference sequences that may interfere with the stress sequences are also acquired. Extreme points in the stress sequences are clustered based on their similarity, and the distribution pattern of extreme points in each cluster is analyzed to determine the degree of agreement with the normal stress signal. This process completes the extreme point screening in the stress sequences, first removing transient noise and providing a better basis for subsequent stress sequence decomposition using the EMD algorithm. Then, the stability of the components obtained from the stress sequence decomposition is analyzed, and components with insufficient stability are repeatedly decomposed until the obtained stress component signals are sufficiently singular and pure. Simultaneously, interference sequences are decomposed to obtain interference components. By comparing the similarity of the variation characteristics of each stress component with all interference components, the degree of interference of each stress component is determined. This allows for targeted denoising suppression of stress components mainly containing interference components and stress components mainly containing stress components, achieving adaptive and targeted denoising for stress in mountainous highway construction areas, thus improving the accuracy of stress monitoring in such complex scenarios.

[0043] To further illustrate the technical solution of the present invention, specific embodiments are described below.

[0044] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a particular feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. Furthermore, a particular feature, structure, or characteristic in one or more embodiments may be combined in any suitable form, and the terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically emphasized.

[0045] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0046] Method Implementation Examples:

[0047] See Figure 1 This is a flowchart illustrating an intelligent monitoring method for construction areas of mountainous highways provided in Embodiment 1 of the present invention. Figure 1 As shown, the monitoring method may include the following steps:

[0048] S101, stress is collected at the target monitoring point to form a stress sequence and interference signals are collected simultaneously. The same interference signals are used to form corresponding interference sequences. The maximum and minimum values ​​of the stress sequence are classified into different extreme value classes. All extreme values ​​are clustered based on the similarity of changes between any two extreme values ​​in any extreme value class. The retention degree of any cluster is determined by the number of extreme points, the uniformity of distribution, and the similarity of changes in any cluster. The extreme values ​​in the clusters with a retention degree greater than the retention degree threshold are taken as target extreme values. Based on the target extreme values, the stress sequence is decomposed into initial component signals using the EMD algorithm.

[0049] This invention aims to address the challenges of complex construction environments and significant multi-source interference in mountainous highway construction by proposing an intelligent monitoring method capable of identifying, adaptively decomposing, and filtering interference signals. This method aims to improve the accuracy of stress data acquisition in mountainous highway construction scenarios, thereby enhancing the reliability of structural safety assessments for construction equipment.

[0050] Therefore, this embodiment first deploys stress sensors at key locations on the construction structure, i.e., monitoring points, during the data acquisition phase to collect real-time stress changes during construction and form stress sequences. Simultaneously, considering that the actual collected stress signals are susceptible to interference from multiple sources such as temperature changes, wind loads, mechanical vibrations, and electromagnetic noise, resulting in random fluctuations and deviations in the signals, and that the stress conditions and environments at different monitoring locations vary significantly, their interference characteristics also differ, making it difficult to effectively remove noise and maintain signal authenticity using a fixed denoising method. Therefore, to make subsequent denoising and correction accuracy possible or a prerequisite, this embodiment also synchronously collects data on interference sources that may affect stress measurement results at the monitoring points, including multi-dimensional information such as ambient temperature, wind load, construction machinery vibration intensity, and surrounding electromagnetic noise. Corresponding interference sequences, such as ambient temperature sequences and wind load sequences, are formed for the same type of interference signal. This facilitates the establishment of a correspondence between stress data and various interference sources, providing an accurate input basis for subsequent noise identification and adaptive denoising algorithms based on multi-source information fusion, ultimately improving the reliability and analytical accuracy of stress monitoring data.

[0051] While existing denoising schemes using the EMD algorithm can decompose acquired data signals and thus denoise individual components, this method relies on determining local extrema. However, as mentioned above, noise interference in mountainous highway construction areas is multi-source and complex. If extrema are extracted solely based on mathematical rules without adaptive optimization according to signal characteristics, the system is susceptible to random noise, leading to mode aliasing and affecting decomposition accuracy and denoising performance. This is because the EMD decomposition process relies on maxima and minima to construct the upper and lower envelopes of the signal. If extrema are not accurately identified (e.g., noise fluctuations are misidentified as extrema or true extrema are missed), the envelope fitting will be biased, causing distortion in the amplitude and frequency characteristics of the component signals. This can lead to mode aliasing or energy leakage, ultimately affecting the physical meaning of the IMF components obtained from EMD decomposition and the accuracy of subsequent denoising and feature extraction.

[0052] Therefore, this step divides the extreme points in the stress sequence into two categories: maxima and minima. The similarity of the extreme points in each category is analyzed and clustered to eliminate extreme points corresponding to instantaneous noise. The remaining extreme points are then used as target extreme points to complete the decomposition of the stress sequence. This effectively suppresses random interference, improves the accuracy and stability of the IMF components, and achieves high-precision denoising analysis of stress signals in complex mountainous environments, facilitating the accuracy of subsequent stress data monitoring.

[0053] Specifically, the maximum and minimum points are determined by taking the second derivative of each data point in the stress sequence. Then, the maximum and minimum values ​​of the stress sequence are assigned to different extreme value classes. The similarity of changes between any two extreme values ​​within any extreme value class is determined, including:

[0054] Determine the angle between any extreme point in any extreme value class and its two adjacent non-extreme extreme points. Calculate the normalized value of the absolute value of the difference between the angle values ​​corresponding to any two extreme points in any extreme value class as the local variation difference degree of the two extreme points. Calculate the normalized value of the absolute value of the difference between the amplitudes of the two extreme points as the amplitude difference degree of the two extreme points.

[0055] The variation similarity between any two extreme points is constructed based on the local variation difference and the amplitude difference, wherein the variation similarity is inversely proportional to both the local variation difference and the amplitude difference.

[0056] Preferably, the formula for the similarity of changes is expressed as follows:

[0057]

[0058] in, This represents the similarity of changes between the a-th and b-th extreme points in the current extreme value class. Represents an exponential function with the natural constant as its base; Let represent the angle between the a-th extreme point in the current extreme value class and its two adjacent (non-class) extreme points. Adjacent (non-class) extreme points refer to points where, since maxima and minima alternate, if the current a-th extreme point is a maximum, then its two adjacent (non-class) minimum points are also minimums. The angle between this maximum point and its two adjacent (non-class) minimum points is denoted as . , This represents the angle between the b-th extreme point in the current extreme value class and its adjacent non-class extreme points; , These represent the magnitudes of the a-th and b-th extreme points in the current extreme value class, respectively. This represents the absolute value of the difference between the angle values ​​of the a-th and b-th extreme points in the current extreme value class, which indicates the difference in the local variation characteristics of these two local extreme points. The smaller this value, the more similar the local variation characteristics of the two local extreme points are. The larger the value; This represents the absolute value of the difference between the amplitudes of the a-th and b-th extreme points in the current extreme value class. The smaller this value, the more similar the amplitudes of these two extreme points in the signal, and the more similar the characteristics they represent. Therefore, the similarity between them may be greater. The larger the value, This represents a normalization function, such as linear normalization, norm normalization, etc., whose function is to make... and They are of the same magnitude.

[0059] Based on the above acquisition of the similarity of changes at different extreme points in any extreme value class, the HDBSCAN clustering algorithm (Hierarchical Density-Based Spatial Clustering of Applications with Noise, a density-based hierarchical clustering algorithm that can effectively discover clusters with arbitrary shapes and automatically identify noise points or outliers by processing datasets of different densities; this algorithm is a well-known existing technology and will not be elaborated here) is used to further divide these extreme points in the same category into several clusters according to similar characteristics. After clustering, different clusters can reflect the change pattern characteristics of stress signals in different time periods or under different stress states.

[0060] Under normal circumstances, clusters representing normal stress variation patterns exhibit good continuity and stability of their local extreme points over time, with clear amplitude variation patterns. However, when stress signals are affected by random noise or other disturbances, the distribution of local extreme points becomes unstable or abrupt, and their similarity decreases significantly.

[0061] Therefore, the stability differences between different clusters can be used to measure their reliability in representing the real stress mode. The higher the reliability of the cluster, the more likely its extreme point is to be the target extreme point corresponding to the normal stress data, and it should be retained. On the other hand, the clusters with lower reliability mostly correspond to random interference and should be removed in subsequent decomposition to avoid the aliasing of random noise with the real signal, which would affect the accuracy of subsequent denoising analysis of the collected stress data.

[0062] Since the clusters corresponding to normal stress variation data have a larger data volume than those corresponding to noise interference data, and the data distribution is more uniform with more consistent similarity between different extreme points, the retention coefficient of any cluster can be determined first.

[0063] The normalized value of the number of extreme points in the stress sequence that do not belong to any particular cluster between the time corresponding to the first extreme point of any cluster and the time corresponding to the last extreme point of any cluster is used as the discontinuity of any particular cluster.

[0064] The normalized value of the coefficient of variation of the set of variation similarity formed by the variation similarity between any two extreme points in any cluster is used as the discreteness of any cluster.

[0065] The retention coefficient of any cluster is constructed based on the number of extreme points in any cluster, the discontinuity, and the dispersion. The retention coefficient is directly proportional to the number of extreme points in any cluster, and inversely proportional to both the discontinuity and the dispersion.

[0066] As a further optimization, the retention factor is:

[0067]

[0068] in, This represents the retention coefficient of the r-th cluster in the clustering results corresponding to the current extreme value class. The value represents the number of extreme points in the r-th cluster. The larger the value, the larger the size of the cluster, the lower the probability that it is a random disturbance, and the larger its retention coefficient; e represents the natural constant. This represents the number of extreme points in the original stress sequence that do not belong to the r-th cluster between the time corresponding to the first extreme point in the r-th cluster and the time corresponding to the last extreme point in the r-th cluster. The larger this value, the less continuous and stable the extreme points are in that cluster, and the greater the probability that they are local extreme points caused by random disturbances. The smaller the value; This represents the coefficient of variation of the set of variation similarities formed by the variation similarities between any two extreme points in the r-th cluster. A larger value indicates lower stability of the variation similarity between extreme points in that cluster, and a greater degree of dispersion in that cluster. The corresponding retention coefficient is... The smaller the value, This represents a normalization function, such as linear normalization, norm normalization, etc.

[0069] Then, the retention degree of each cluster can be determined further based on the relative magnitude of the retention coefficients among the clusters obtained under the same extreme value class, including:

[0070] Calculate the ratio of the retention coefficient of any cluster to the retention coefficient of each cluster in the extreme value class to which the cluster belongs, and use the normalized value of the mean of all the obtained ratios as the retention degree of the cluster.

[0071] The formulaic representation of retention is as follows:

[0072]

[0073] in, This represents the retention rate of the r-th cluster in the clustering results corresponding to the current extreme value class, where m represents the number of clusters in the clustering results corresponding to the current extreme value class. This represents the retention coefficient of the r-th cluster in the clustering results corresponding to the current extreme value class. This represents the retention coefficient of the y-th cluster in the clustering results corresponding to the current extreme value class. This represents the average of the sum of the ratios of the retention coefficients of the r-th cluster to all other clusters in the clustering results corresponding to the current extreme value. The larger this value, the greater the degree of retention that the r-th cluster should retain compared to the other clusters, and the larger its retention value. This represents a normalization function, such as linear normalization, norm normalization, etc.

[0074] To achieve effective screening of clusters, a retention threshold is further set. ,Will Greater than Clusters are retained. Retention threshold. It can be empirically set to 0.1 and can be adaptively adjusted according to actual conditions. Specifically, noise-free simulated stress data and noisy simulated data can be used as a comparison, and iterative adjustments can be made through Bayesian or other optimization algorithms. The value of is chosen to minimize the difference between the denoised signal and the original noise-free signal, thereby determining the optimal threshold and applying it to the actual monitoring environment.

[0075] Similarly, the processing of extreme points for another type of extreme value is as described above. Ultimately, all extreme points in the stress sequence are filtered, and the filtered extreme points are recorded as target extreme points. Then, based on the obtained target extreme points, the EMD algorithm is used to decompose the collected stress data into multiple component signals, which are recorded as initial component signals. The method of obtaining component signals using the EMD algorithm is existing technology and will not be described in detail in this embodiment.

[0076] This step serves two purposes: firstly, it suppresses the presence of transient noise in the initial component signal by filtering out extreme points, thereby achieving noise reduction of transient noise; secondly, it ensures the structural stability and physical consistency of EMD decomposition, and prevents mode aliasing to a certain extent.

[0077] S102, determine the difference between two adjacent sliding windows by the difference in amplitude and frequency change between any two adjacent sliding windows when the sliding window slides on any initial component signal, determine the stability of the initial component signal by the dispersion and magnitude of all differences under the initial component signal, decompose the initial component signal with stability less than the stability threshold into new initial component signals by EMD algorithm until there are no initial component signals with stability less than the stability threshold, and denote the initial component signal at this time as the stress component signal.

[0078] By filtering the collected stress data for extreme points before using EMD decomposition, signal disturbances caused by occasional transient random noise in complex mountainous environments can be effectively reduced, resulting in smoother component signals and clearer physical meanings. However, during the construction of mountain highways, on the one hand, external environmental conditions (such as mechanical vibration and wind) often generate stable but large-amplitude interference over short periods (such as several seconds or even minutes). This interference differs from transient random noise and cannot be eliminated through the aforementioned extreme point filtering, resulting in an insufficiently pure and simple composition of the initial component signals. On the other hand, the stress state of the monitored construction equipment at the monitoring points may also change with the construction stage, which also leads to an insufficiently pure and simple composition of the initial component signals. In other words, even after extreme point filtering, the obtained initial component signals may still suffer from poor continuous stability over long periods.

[0079] Therefore, in order to ensure that the component signals can accurately reflect the actual stress change characteristics, it is necessary to further analyze the characteristics of the obtained initial component signals and determine whether further decomposition processing is needed based on their characteristics, so as to improve the accuracy of subsequent stress data correction.

[0080] To further evaluate the stability of each component signal, this invention performs a Short-Time Fourier Transform (STFT) on any initial component signal to obtain its time-frequency distribution characteristics. Subsequently, an analysis window of the same size as that used in the STFT process is selected as the sliding window, with a sliding step size set to half the length of the sliding window to ensure a balance between time and frequency resolution. During continuous sliding, the mean square error (MSE) of the spectral data and the amplitude difference of the component signals within two adjacent windows are calculated to obtain a signal difference index between window segments. The spectral mean square error measures the difference in frequency variation of the signal, reflecting the degree of drift or aliasing of the dominant frequency of the component signal in different time periods; while the amplitude difference characterizes the change in the intensity of signal fluctuations in the time domain, revealing local amplitude instability caused by external interference or stress abrupt changes. By combining the difference characteristics in the frequency and time domains, the dynamic stability of the component signals can be evaluated more comprehensively, providing a basis for subsequent judgments on whether further decomposition is necessary.

[0081] Based on the above, the degree of difference between two adjacent sliding windows is determined, including:

[0082] The difference between the mean of the maximum and the mean of the minimum values ​​in any sliding window of the two adjacent sliding windows is calculated as the average extreme value difference of the sliding window. The normalized value of the absolute value of the difference of the average extreme value difference between the two adjacent sliding windows is used as the amplitude difference between the two adjacent sliding windows.

[0083] The normalized value of the mean square error between the spectral data corresponding to each of the two adjacent sliding windows is calculated as the frequency change difference between the two adjacent sliding windows.

[0084] The sum of the amplitude difference and the frequency change difference is used as the difference degree between the two adjacent sliding windows.

[0085] The formula for the difference between two adjacent sliding windows is as follows:

[0086]

[0087] in, This represents the difference between any two adjacent sliding windows of the current initial component signal. This represents the mean square error of the spectral data between any two adjacent sliding windows of the current initial component signal. The larger this value, the greater the difference in the frequency characteristics of the data changes within the two sliding windows. , Let represent the mean of the maximum points and the mean of the minimum points within the a-th sliding window of the initial component signal being analyzed, respectively. , These represent the mean of the maximum points and the mean of the minimum points within the (a+1)th sliding window of the currently analyzed initial component signal, respectively. This represents the overall range of amplitude variation within the a-th sliding window of the initial component signal being analyzed. This represents the range of amplitude variation within the (a+1)th sliding window of the initial component signal being analyzed. This represents the difference in the overall amplitude variation range within the adjacent a-th and (a+1)-th sliding windows of the initial component signal analyzed previously. The larger this value, the worse the stability between the two windows. This represents a normalization function, such as linear normalization or norm normalization, which aims to unify the magnitude of a quantity.

[0088] Therefore, for any initial component signal, several differences can be obtained according to the time sequence, i.e., the sliding window, and further constitute a difference sequence. This difference sequence can reflect the local fluctuation characteristics of the initial component signal in time, thereby reflecting the stability of the initial component signal. When the difference sequence shows high stability in time, it indicates that the frequency distribution of the initial component signal is stable and no further decomposition is needed. Conversely, if the difference sequence fluctuates greatly, it indicates that there may still be aliasing components or short-term disturbance characteristics in the initial component signal, and secondary decomposition or feature extraction should be performed.

[0089] Based on this, the stability of the initial component signal is determined, including:

[0090] The coefficient of variation of the difference sequence formed by the difference degree corresponding to any two adjacent sliding windows in any initial component signal is calculated as the dispersion of all differences degree under the given initial component signal.

[0091] The stability of any initial component signal is constructed based on the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal. The stability of any initial component signal is inversely proportional to both the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal.

[0092] As a further preferred option, the stability is:

[0093]

[0094] Where B represents the stability of the current initial component signal, This represents the difference sequence corresponding to the current initial component signal. This represents the coefficient of variation of the difference sequence corresponding to the current initial component signal. The larger the value, the greater the dispersion of the difference sequence, the worse the stability of the current initial component signal, and the greater the necessity for further decomposition of the current initial component signal. The reason for this is... Using natural constant mapping serves two purposes: firstly, it enhances the representation of the stability characteristics of changes in the current initial component signal; secondly, it prevents... This part is set to 0 to avoid errors in subsequent calculations. This represents the mean of all differences in the difference sequence corresponding to the current initial component signal. The smaller this value, the greater the overall temporal stability of the current initial component signal.

[0095] Therefore, the stability level of the initial component signal can be adaptively identified, thereby effectively determining the necessity of further decomposition and improving the targeting and reliability of signal processing. To successfully determine whether further decomposition of the initial component signal is necessary, a stability threshold is further set. (The default value here is 0.65, which is an empirical value. The method for determining a further optimized value is the same as described above.) (The method of determining the value), when B is less than the stability threshold If the initial component signal is unstable in terms of time, it indicates that the initial component signal is not stable enough and needs to be further decomposed using the EMD algorithm to obtain several new initial component signals.

[0096] For all initial component signals, a process of determining whether to further decompose based on stability is performed until the stability of all current initial component signals is no less than a stability threshold. If we assume that the signal components of all initial component signals are relatively simple and pure at this point, without obvious mode aliasing, then the accuracy of subsequent denoising processing can be guaranteed. Therefore, all initial component signals at this point can be denoised as stress component signals for use in subsequent denoising.

[0097] S103, each interference sequence is decomposed into interference component signals using the EMD algorithm, the interference degree of the current stress component signal is determined by the feature similarity between the current stress component signal and each interference component signal, the denoising threshold of the current stress component signal is adjusted by the interference degree of the current stress component signal to obtain the denoised component signal, and stress monitoring is completed after reconstructing all the denoised component signals into a denoised stress sequence.

[0098] It is easy to understand that although the signal components in each stress component signal are relatively simple and pure, there are short-term stable external interferences during the stress sequence acquisition process. Therefore, a certain number of component signals in the obtained stress component signals have internal signal components that are not stress data but external interferences. Therefore, it is also necessary to decompose the interference sequence acquired synchronously during stress data acquisition to obtain interference component signals, and compare them with the obtained stress component signals to determine the amount of external interference information contained in each stress component signal, so as to carry out targeted noise reduction processing, improve the accuracy of the final stress data acquisition, and ensure the accuracy of stress monitoring.

[0099] Stress data from construction equipment in mountainous highway construction areas are often affected by environmental interference. The primary purpose of extreme point selection and accurate signal extraction is to improve the decomposition accuracy of signals heavily influenced by complex interference or random noise, obtaining single-component components. However, external noise sources, such as wind-induced vibrations and mechanical vibrations, typically have relatively singular signal sources with concentrated frequency and energy characteristics, making them less prone to complex aliasing with other signals. Therefore, when analyzing such interference signals, EMD (Enhanced Motion Decomposition) can be directly used without the aforementioned extreme point selection and other processing. In other words, for interference sequences composed of simple structures and clearly defined fluctuation patterns, direct decomposition yields relatively stable and physically meaningful IMF (Integrated Motion Component) components, which can be used for subsequent comparison with stress signal component characteristics to complete interference identification and analysis. Thus, each interference sequence is directly decomposed to obtain several interference component signals.

[0100] Based on the obtained interference component signals, the feature similarity of each interference component signal to each stress component signal can be determined:

[0101]

[0102] in, This represents the feature similarity between the current stress component signal S and the current disturbance component signal Y. This represents the mutual information value between the current stress component signal S and the current interference component signal Y. The calculation of the mutual information value is a prior art technique and will not be elaborated in this embodiment. The larger the mutual information value, the more information is shared between the current interference component signal and the current stress component signal, which also indicates that the interference of the current interference component signal on the current stress component signal may be greater. This represents the number of frequency intersection points between the current stress component signal S and the current interference component signal Y (obtained through the aforementioned short-time Fourier transform), which is also the number of frequency points in their respective spectral data that have the same frequency and whose corresponding amplitudes are all non-zero. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current interference component signal Y. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current stress component signal S. This represents the upper frequency limit of the spectral data of the current stress component signal S. This represents the lower frequency limit of the spectral data of the current stress component signal S. This represents the amplitude corresponding to the v-th frequency in the spectral data of the current stress component signal S. This represents the weight of the amplitude corresponding to the u-th intersection frequency point in the spectrum data of the current stress component signal S, representing the energy proportion of the spectrum data of the current stress component signal S. The larger this value, the greater the contribution of that frequency to the current stress component signal S, and the greater the influence of external noise on the current stress component signal. This represents the ratio of the amplitude of the u-th frequency intersection point in the current noise component signal spectrum data to the amplitude of its corresponding amplitude in the current stress component signal spectrum data. The larger this value, the greater the influence of that frequency point in the current interference component signal on that frequency point in the current stress component signal, and the greater the contribution weight of that frequency point to the current stress component signal, indicating a greater degree of noise impact. Therefore, This represents the weighted average value of the influence of the current interference component signal Y on the frequency points of the current stress component signal S. The larger this value is, the greater the noise interference of the interference source component signal on the stress component signal, and the greater the feature similarity between the two component signals.

[0103] Therefore, for any stress component signal, the feature similarity between it and each interfering component signal can be obtained. Since the feature similarity characterizes the degree of noise interference of each interfering component signal to the stress component signal, the degree of interference of the current stress component signal can be determined based on the obtained feature similarity, including:

[0104] The sum of the feature similarities between the current stress component signal and each interference component signal is taken as the interference degree of the current stress component signal.

[0105] The formulaic characterization of the interference level of the current stress component signal is as follows:

[0106]

[0107] in, The current stress component signal S represents its final degree of interference, and L represents the total number of interfering components. This represents the sum of the characteristic similarities between the current stress component signal S and all interfering component signals. It also characterizes the sum of the degree of interference of the current stress component signal S with all interfering component signals. The larger this value is, the greater the degree of interference it is subjected to by the interference source.

[0108] The above method yields the interference level of any component signal of any stress sequence at any monitoring point. This interference level is used as the basis for adaptive threshold adjustment in the wavelet threshold denoising algorithm for stress component signals. The denoising threshold for the current stress component signal is adjusted, including:

[0109] The product of the interference level of the current stress component signal and the preset mapping parameter is used as the denoising threshold of the current stress component signal.

[0110] The formula for the denoising threshold is expressed as follows:

[0111]

[0112] in, This represents the denoising threshold of the wavelet threshold denoising algorithm for the current stress component signal. This represents the preset mapping parameters. The optimal value can be obtained by combining existing experimental data, that is, by using Bayesian optimization algorithm through adaptive training and correction based on the sample data of the interference degree of the stress component signal and the corresponding noise content, or by manual adjustment. There are no restrictions here.

[0113] This achieves adaptive and targeted adjustment of the stress component denoising threshold based on the noise characteristics of the stress component signal, resulting in higher suppression of noisy stress components and preservation of more true stress variation characteristics for less noisy stress components. Subsequently, all component signals after adaptive denoising are reconstructed to obtain the overall denoised stress signal, i.e., the denoised stress sequence, which is then input into a trained Support Vector Machine (SVM) model for stress anomaly detection. By combining adaptive denoising with artificial intelligence monitoring algorithms, the accuracy and stability of stress signal anomaly monitoring for relevant construction equipment in mountainous highway construction areas are improved, enhancing the reliability of structural safety assessment in complex mountainous construction environments.

[0114] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent monitoring of construction areas of mountainous expressways, characterized in that, include: Stress is collected at the target monitoring point to form a stress sequence, and interference signals are collected simultaneously. The same interference signals are used to form a corresponding interference sequence. The maximum and minimum values ​​of the stress sequence are classified into different extreme value classes. All extreme values ​​are clustered based on the similarity of changes between any two extreme values ​​in any extreme value class. The retention of a cluster is determined by the number of extreme points, the uniformity of distribution, and the similarity of changes in any cluster. The extreme values ​​in the clusters with a retention greater than the retention threshold are taken as target extreme values. Based on the target extreme values, the stress sequence is decomposed into initial component signals using the EMD algorithm. The difference between two adjacent sliding windows is determined by the difference in amplitude and frequency change between any two adjacent sliding windows when the sliding window slides on any initial component signal. The stability of the initial component signal is determined by the dispersion and magnitude of all differences under the initial component signal. The initial component signal with stability less than the stability threshold is decomposed into a new initial component signal by the EMD algorithm until there is no initial component signal with stability less than the stability threshold. The initial component signal at this time is called the stress component signal. Each interference sequence is decomposed into interference component signals using the EMD algorithm. The interference degree of the current stress component signal is determined by the feature similarity between the current stress component signal and each interference component signal. The denoising threshold of the current stress component signal is adjusted based on the interference degree of the current stress component signal to obtain the denoised component signal. After reconstructing all the denoised component signals into a denoised stress sequence, stress monitoring is completed.

2. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1, characterized in that, The interference signals include ambient temperature signals, wind load signals, mechanical vibration intensity signals, and electromagnetic signals.

3. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1, characterized in that, Determining the similarity of changes at any two extreme values ​​in any extreme value class includes: Determine the angle between any extreme point in any extreme value class and its two adjacent non-extreme extreme points. Calculate the normalized value of the absolute value of the difference between the angle values ​​corresponding to any two extreme points in any extreme value class as the local variation difference degree of the two extreme points. Calculate the normalized value of the absolute value of the difference between the amplitudes of the two extreme points as the amplitude difference degree of the two extreme points. The variation similarity between any two extreme points is constructed based on the local variation difference and the amplitude difference, wherein the variation similarity is inversely proportional to both the local variation difference and the amplitude difference.

4. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1 or 3, characterized in that, The determination of the retention degree of a cluster based on the number of extreme points, distribution uniformity, and variation similarity in any given cluster includes: The normalized value of the number of extreme points in the stress sequence that do not belong to any particular cluster between the time corresponding to the first extreme point of any cluster and the time corresponding to the last extreme point of any cluster is used as the discontinuity of any particular cluster. The normalized value of the coefficient of variation of the set of variation similarity formed by the variation similarity between any two extreme points in any cluster is used as the discreteness of any cluster. The retention coefficient of any cluster is constructed based on the number of extreme points in any cluster, the discontinuity, and the dispersion. The retention coefficient is directly proportional to the number of extreme points in any cluster, and inversely proportional to both the discontinuity and the dispersion. The retention degree of any cluster is determined by comparing its retention coefficient with the retention coefficients of all clusters in the extreme value class to which the cluster belongs.

5. The intelligent monitoring method for construction areas of mountainous expressways according to claim 4, characterized in that, Determining the retention rate of any of the aforementioned clusters includes: Calculate the ratio of the retention coefficient of any cluster to the retention coefficient of each cluster in the extreme value class to which the cluster belongs, and use the normalized value of the mean of all the obtained ratios as the retention degree of the cluster.

6. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1, characterized in that, Determining the difference between two adjacent sliding windows includes: The difference between the mean of the maximum and the mean of the minimum values ​​in any sliding window of the two adjacent sliding windows is calculated as the average extreme value difference of the sliding window. The normalized value of the absolute value of the difference of the average extreme value difference between the two adjacent sliding windows is used as the amplitude difference between the two adjacent sliding windows. The normalized value of the mean square error between the spectral data corresponding to each of the two adjacent sliding windows is calculated as the frequency change difference between the two adjacent sliding windows. The sum of the amplitude difference and the frequency change difference is used as the difference degree between the two adjacent sliding windows.

7. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1 or 6, characterized in that, Determining the stability of the initial component signal includes: The coefficient of variation of the difference sequence formed by the difference degree corresponding to any two adjacent sliding windows in any initial component signal is calculated as the dispersion of all differences degree under the given initial component signal. The stability of any initial component signal is constructed based on the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal. The stability of any initial component signal is inversely proportional to both the dispersion of all differences under any initial component signal and the mean of all differences under any initial component signal.

8. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1, characterized in that, The feature similarity is: ， in, This represents the feature similarity between the current stress component signal S and the current disturbance component signal Y. This represents the mutual information value between the current stress component signal S and the current disturbance component signal Y. This represents the number of frequency intersection points between the current stress component signal S and the current interference component signal Y, that is, the number of frequency points in their respective spectral data that have the same frequency and whose corresponding amplitudes are all non-zero. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current interference component signal Y. This represents the amplitude corresponding to the intersection point with the u-th frequency in the spectral data of the current stress component signal S. This represents the upper frequency limit of the spectral data of the current stress component signal S. This represents the lower frequency limit of the spectral data of the current stress component signal S. This represents the amplitude corresponding to the v-th frequency in the spectral data of the current stress component signal S.

9. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1 or 8, characterized in that, Determining the degree of interference of the current stress component signal includes: The sum of the feature similarities between the current stress component signal and each interference component signal is taken as the interference degree of the current stress component signal.

10. The intelligent monitoring method for construction areas of mountainous expressways according to claim 1, characterized in that, The adjustment of the denoising threshold for the current stress component signal includes: The product of the interference level of the current stress component signal and the preset mapping parameter is used as the denoising threshold of the current stress component signal. The preset mapping parameters are obtained by adaptive training and correction using a Bayesian optimization algorithm based on sample data of the interference level and corresponding noise content of the determined stress component signals.

Images