Intelligent analysis method and system for pile foundation vibration monitoring data

Abstract

This invention relates to the field of data analysis technology, specifically to an intelligent analysis method and system for pile foundation vibration monitoring data. The method includes: acquiring a current vibration subsequence and a sample vibration subsequence; acquiring the energy intensity characteristic value and main frequency characteristic value of the vibration subsequence; obtaining a reference subsequence set corresponding to the current vibration subsequence based on the difference in influence data between the current vibration subsequence and the sample vibration subsequence; obtaining the relative amplitude anomaly index value and the relative frequency anomaly index value of the current vibration subsequence based on the characteristic value difference, time distance, and difference in influence data between the current vibration subsequence and the reference subsequences in the reference subsequence set; and performing pile foundation vibration anomaly early warning based on the relative amplitude anomaly index value and the relative frequency anomaly index value of the current vibration subsequence. Furthermore, this invention can improve the accuracy and reliability of pile foundation vibration anomaly early warning.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to an intelligent analysis method and system for pile foundation vibration monitoring data. Background Technology

[0002] Currently, in order to ensure the structural integrity and long-term safety of bridge pile foundations, it is usually necessary to monitor and analyze the vibration of bridge pile foundations and to issue early warnings of abnormal pile foundation vibration based on the monitoring and analysis results. In other words, monitoring and analyzing pile foundation vibration is the key to ensuring the structural integrity and long-term safety of bridge pile foundations.

[0003] In existing technologies, the monitoring and analysis of pile foundation vibration typically do not consider the impact of different working environments on vibration. For example, they directly judge whether the monitored vibration data exceeds a threshold, and if so, issue an early warning of abnormal pile foundation vibration. However, pile foundation vibration is affected by changes in the working environment. Therefore, if the monitoring and analysis of pile foundation vibration ignores the differences in the working environment, the reliability of the analysis results will be low, leading to false alarms. For example, pile foundation vibration data collected under different loads, wind speeds, and temperatures have significant differences in temporal characteristics, amplitude distribution, and fluctuation characteristics. If the monitoring and analysis method based on the comparison of vibration data with thresholds is still used for early warning, the probability of false alarms will increase. Therefore, how to monitor and analyze pile foundation vibration to improve the accuracy and reliability of early warning of abnormal pile foundation vibration has become an urgent problem to be solved. Summary of the Invention

[0004] To address the above problems, this invention provides an intelligent analysis method and system for pile foundation vibration monitoring data, the specific technical solution of which is as follows: In a first aspect, embodiments of the present invention provide an intelligent analysis method for pile foundation vibration monitoring data, comprising the following steps: Obtain the pile foundation vibration data sequence to be divided and the influence data sequence to be divided, wherein the influence data sequence to be divided includes the pile foundation load, pile foundation temperature and pile foundation wind load influence data sequence to be divided; A local window of a preset time length is constructed with each data point in each data sequence to be divided as the center, and is denoted as the local window of the corresponding data. Based on the correlation between the data on both sides of the data center of the local window and the difference variation coefficient of the local window, the window fitness evaluation value of each data point in each data sequence to be divided is obtained. Based on the window fitness evaluation value, the data sequences to be divided are divided and merged to obtain the target segmentation time. Based on the target segmentation time, the vibration data sequence of the pile foundation to be segmented is divided and identified to obtain the current vibration subsequence and the sample vibration subsequence. The energy intensity feature value and main frequency feature value of the vibration subsequence are obtained. Based on the difference in influence data between the current vibration subsequence and the sample vibration subsequence, the reference subsequence set corresponding to the current vibration subsequence is obtained. Based on the feature value difference, time distance and influence data difference between the current vibration subsequence and the reference subsequence in the reference subsequence set, the amplitude relative anomaly index value and frequency relative anomaly index value of the current vibration subsequence are obtained. Based on the relative abnormality values ​​of amplitude and frequency of the current vibration subsequence, early warning of pile foundation vibration anomalies is carried out.

[0005] Secondly, embodiments of the present invention provide an intelligent analysis system for pile foundation vibration monitoring data, including a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the aforementioned intelligent analysis method for pile foundation vibration monitoring data.

[0006] Beneficial effects: This invention first acquires the pile foundation vibration data sequence to be divided and the influence data sequence to be divided; then, it constructs a local window of a preset time length centered on each data in each influence data sequence to be divided, and records it as the local window of the corresponding data. Based on the correlation between the data on both sides of the data at the center of the local window and the difference coefficient of variation of the local window, it obtains the window fitness evaluation value of each data in each influence data sequence to be divided. Based on the window fitness evaluation value, it divides and merges the influence data sequences to be divided to obtain the target segmentation time; then, based on the target segmentation time, it divides and identifies the pile foundation vibration data sequence to be divided to obtain the current vibration subsequence and the sample vibration subsequence, obtains the energy intensity feature value and the main frequency feature value of the vibration subsequence, obtains the reference subsequence set corresponding to the current vibration subsequence based on the influence data difference value between the current vibration subsequence and the sample vibration subsequence, obtains the reference subsequence set corresponding to the current vibration subsequence, obtains the amplitude relative anomaly index value and the frequency relative anomaly index value of the current vibration subsequence based on the feature value difference, time distance and influence data difference value between the current vibration subsequence and the reference subsequence in the reference subsequence set, and finally obtains the pile foundation vibration anomaly early warning based on the amplitude relative anomaly index value and the frequency relative anomaly index value of the current vibration subsequence; finally, it performs pile foundation vibration anomaly early warning based on the amplitude relative anomaly index value and the frequency relative anomaly index value of the current vibration subsequence. Furthermore, the amplitude relative anomaly index value and frequency relative anomaly index value obtained by the present invention based on the analysis of both time domain and frequency domain can improve the ability to capture hidden and slowly evolving pile foundation vibration anomalies or improve the monitoring and analysis accuracy under different working environments. In turn, it can improve the accuracy and reliability of early warning of pile foundation vibration anomalies, thereby enhancing the reference value for actual engineering safety assessment. Attached Figure Description

[0007] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, 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.

[0008] Figure 1 This is a flowchart of an intelligent analysis method for pile foundation vibration monitoring data according to the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the protection scope of the embodiments of the present invention.

[0010] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0011] This embodiment provides an intelligent analysis method for pile foundation vibration monitoring data, which is described in detail below: like Figure 1 As shown, this intelligent analysis method for pile foundation vibration monitoring data includes the following steps: Step S001: Obtain the vibration data sequence and influence data sequence of the target pile foundation.

[0012] Since the bridge pile foundation bears the bridge structure above, when the bridge operates under vehicle loads, wind loads, and structural ambient temperature, it transmits dynamic forces to the lower pile foundation, causing pile vibration response. That is, vehicle loads, wind loads, and structural ambient temperature affect pile foundation vibration. However, because the vibration response of the pile foundation may show significant differences in amplitude distribution and frequency structure under different vehicle loads, wind loads, and structural ambient temperature conditions, these differences can interfere with the judgment of vibration status, resulting in low accuracy and reliability of pile foundation abnormal vibration early warning. In this embodiment, when monitoring and analyzing the vibration of bridge pile foundation, this embodiment improves the ability to capture hidden and slowly evolving pile foundation vibration anomalies by combining the pile foundation working environment, or improves the monitoring and analysis accuracy under different working environments. In other words, by combining the pile foundation working environment, it can more accurately characterize the vibration behavior characteristics of the pile foundation under complex conditions, thereby improving the accuracy and reliability of pile foundation vibration anomaly early warning, so as to ensure the integrity and long-term safety of the pile foundation structure.

[0013] Based on the above analysis, this embodiment needs to first monitor and acquire the impact data and pile foundation vibration data that affect the pile foundation vibration. The impact data that affects the pile foundation vibration also reflects the data of the pile foundation's operating environment. The impact data that affects the pile foundation vibration collected in this embodiment includes, but is not limited to, pile foundation load impact data, pile foundation temperature impact data, and pile foundation wind load impact data. Furthermore, for ease of analysis and understanding, this embodiment will subsequently describe the monitoring and analysis process of vibration data for any pile foundation of any bridge as an example, or in other words, it will subsequently describe the vibration anomaly early warning process for any pile foundation of any bridge as an example, and this pile foundation will be referred to as the target pile foundation. The specific process for acquiring the impact data and vibration data of the target pile foundation at each monitoring time is as follows: During the monitoring of the target pile foundation used for the bridge, the vibration data of the target pile foundation collected by the high-sensitivity vibration sensor at each monitoring time is normalized and recorded as the pile foundation vibration data at the corresponding monitoring time. The wind speed data of the target pile foundation working environment collected by the wind speed sensor at each monitoring time is normalized and recorded as the pile foundation wind load impact data at the corresponding monitoring time. The structural temperature data of the target pile foundation working environment collected by the temperature sensor at each monitoring time is recorded as the pile foundation wind load impact data. The result of normalizing the temperature data is recorded as the pile foundation temperature influence data at the corresponding monitoring time. The result of normalizing the total vertical force on the bridge deck collected by the pressure sensor or weighbridge system at each monitoring time is recorded as the pile foundation load influence data at the corresponding monitoring time. The structural temperature data of the working environment of the target pile foundation refers to the thermal environment experienced by the internal structure and interface of the concrete target pile foundation. The total vertical force on the bridge deck at a certain time is the bridge traffic vehicle load data collected by the pressure sensor or weighbridge system at that time, that is, the load data is caused by the traffic vehicles on the bridge.

[0014] In this embodiment, the high-sensitivity vibration sensor used to collect vibration data of the target pile foundation is arranged based on the bridge structure and stress characteristics. For example, in this embodiment, it can be placed at the top of the pile foundation, the critical stress section of the pile body, or near the pile cap. The pile cap is located at the top of the pile foundation, directly supporting the pile head, and not connected to the bridge pier or abutment. The vibration data of the target pile foundation collected by the high-sensitivity vibration sensor is also the vibration data of the target pile foundation under vehicle load, environmental excitation, and structural response. Wind load is an important excitation source for the vibration of the pile foundation structure; that is, the pile foundation vibration is affected by wind load. Therefore, the sensor is deployed to collect vibration data from the target pile foundation. When installing sensors to collect wind speed data in the working environment, it is generally necessary to avoid structural obstructions and place them close to the airflow inlet. For example, wind speed sensors that collect wind load or target pile foundation working environment wind speed can be placed at the top of the bridge tower or the end of the main beam. The temperature of the pile foundation environment structure is also an important excitation source for pile foundation structure vibration, that is, the vibration of the pile foundation is affected by wind load. When deploying sensors to collect the temperature of the target pile foundation environment structure structure, it is necessary to focus on the key interface of the thermal response of the structure itself, rather than simply the ambient air temperature. For example, temperature sensors that collect the temperature of the pile foundation environment structure structure can be placed at the joint surface between the pile top and the pile cap, or in the middle section of the pile body. Bridge vehicle load data is a crucial excitation source for pile foundation vibration, meaning pile foundation vibration is influenced by load. When deploying weighbridge systems or pressure sensors to collect bridge vehicle load data, it's generally necessary to avoid locations such as expansion joints or bridge abutment transition sections. For example, the weighbridge system or pressure sensors can be placed directly below the bridge deck's driving lanes. In practical applications, implementers need to deploy sensors based on the actual conditions of the bridge structure. In this embodiment, all sensors collect data synchronously, and the collected data will be uniformly transmitted to the monitoring platform for storage and management through a data acquisition terminal to ensure data consistency and correlation in time sequence. This embodiment can use maximum-minimum normalization to normalize the data collected by the sensors. For example, for any data collected by any sensor, the ratio of the data minus the sensor's minimum measurement value to the sensor's range is the result of normalization. The sensor's range is the difference between the sensor's maximum and minimum measurement values. The purpose of normalizing the collected data is to eliminate differences in data dimensions and improve the comparability of multi-source data.

[0015] After obtaining the impact data and pile foundation vibration data at each monitoring time, this embodiment needs to construct a time series sequence from the start of vibration anomaly monitoring of the target pile foundation to the current monitoring time to ensure the accuracy and reliability of monitoring and early warning. After constructing the time series sequence, it is divided. The purpose of this division is to identify vibration data segments under stable working conditions or segments with smaller fluctuations in image data, thereby reducing the false alarm rate and improving the sensitivity of subsequent vibration anomaly detection. The specific process for constructing the time series sequence from the start of monitoring to the current monitoring time is as follows: the time period from the start of vibration monitoring of the target pile foundation to the current monitoring time is recorded as the current cumulative monitoring time period. The pile foundation vibration data at each monitoring time within the aforementioned current cumulative monitoring time period is then analyzed. The time series of pile foundation vibration data is denoted as the pile foundation vibration data sequence to be divided at the current monitoring time. The time series of pile foundation wind load influence data at each monitoring time in the current cumulative monitoring period is denoted as the pile foundation wind load influence data sequence to be divided at the current monitoring time. The time series of pile foundation load influence data at each monitoring time in the current cumulative monitoring period is denoted as the pile foundation load influence data sequence to be divided at the current monitoring time. The time series of pile foundation temperature influence data at each monitoring time in the current cumulative monitoring period is denoted as the pile foundation temperature influence data sequence to be divided at the current monitoring time. That is, the influence data sequence to be divided includes the pile foundation load influence data sequence, the pile foundation temperature influence data sequence, and the pile foundation wind load influence data sequence.

[0016] Therefore, this embodiment can obtain the pile foundation vibration data sequence to be divided and the influence data sequence to be divided through the above process.

[0017] Step S002: Construct a local window of a preset time length centered on each data in each data sequence to be divided, and record it as the local window of the corresponding data. Based on the correlation between the data on both sides of the central data of the local window and the difference variation coefficient of the local window, obtain the window fitness evaluation value of each data in each data sequence to be divided. Based on the window fitness evaluation value, obtain the target segmentation time.

[0018] Because the working environment of pile foundations exhibits dynamic fluctuations and phased changes—that is, the load, temperature, wind load, and other influencing factors on the pile foundation vary dynamically and phasedly over different time periods—and the vibration response of the pile foundation is the result of the coupling effect between the inherent dynamic characteristics of the structure and external excitation, different combinations of working conditions directly affect the structural distribution of vibration amplitude and frequency. Therefore, if the global time series data is analyzed as a whole, it is easy to misjudge differences in normal working conditions as structural anomalies or differences in normal working environments as structural anomalies, thereby masking the true vibration characteristics under local stable conditions and affecting the accuracy of pile foundation vibration anomaly monitoring, analysis, and early warning. In order to ensure the accuracy of subsequent pile foundation vibration anomaly monitoring, analysis, and early warning, this embodiment will first adaptively segment the influencing data sequence based on the stability of the influencing data changes, and obtain stable sub-segments on the influencing data sequence to be divided for pile foundation load, pile foundation temperature, and pile foundation wind load, i.e., divide the sub-sequences. Then, continuous sub-sequences with similar characteristics will be merged to form a continuous sub-sequence. Longer segments of stable operating conditions or stable influence data with longer durations are also called target subsequences. Subsequently, the pile foundation vibration data sequence to be divided is segmented based on the time corresponding to the boundary data of the target subsequence (target segmentation time). This extracts pile foundation vibration data segments (vibration subsequences) under stable operating conditions or stable influencing factors, thereby improving the accuracy and rationality of subsequent abnormal vibration identification and early warning. Furthermore, in this embodiment, continuous subsequences with similar characteristics are merged because excessive segmentation leads to fragmentation of reference samples, increases subsequent computational load, and reduces the reliability of subsequent comparisons with the same operating conditions. Merged segments of stable operating conditions or stable working environments improve the robustness of subsequent vibration characteristic estimation and the feasibility of spectral analysis while ensuring the stability of the operating conditions or influence data. This provides a more reliable statistical basis for constructing a reference set (reference subsequence set) based on similar influencing factors and conducting relative anomaly (relative anomaly index value) assessment, thereby improving the accuracy and rationality of anomaly identification and early warning.

[0019] Based on the above analysis, it can be seen that the next step in this embodiment is to determine the target segmentation time. The specific process for obtaining the target segmentation time is as follows: First, a preset time length is obtained. This preset time length is the longest possible length for subsequent adaptive partitioning or the longest possible time for dividing subsequences. The subsequent adaptive partitioning is not based on a static preset time length or a fixed time window. Instead, it is achieved by analyzing the stability of data changes within a local window constructed using the preset time length as the window length. This process adjusts the local window to complete the partitioning, allowing for the extraction of stable data segments that influence data changes. This forms the basis for determining the appropriate segmentation of the parent school. In practical applications, the implementer needs to set the preset time length based on actual conditions such as the time interval between adjacent monitoring times or the monitoring frequency. However, the time interval between adjacent monitoring times must be much shorter than the preset time length. For example, if the sensor's monitoring frequency is set to 200 Hz, then in this embodiment, the preset time length is set to 1 minute. Then, a local window of the preset time length is constructed centered on each data point in each data sequence to be partitioned, and this window is denoted as the local window for the corresponding data. For example, if the acquisition time corresponding to a certain data point in any data sequence to be partitioned is r0, then the local window for that data is defined by the time interval r0. At the time The data within the local window of this data consists of data whose acquisition time falls within that window in the data sequence to be segmented; and in this embodiment, the cumulative duration on the left or right side is not less than... The data is used to construct a local window, that is, for any data in the data sequence to be divided, if the time span between the first data in the data sequence and this data is less than [a certain value], then [the window is defined]. Or the time span between the last data point in the data sequence to be partitioned and this data is less than [a certain value]. If the data is in a certain range, then no local window is constructed for that data, and T0 is the preset time length. The time span between the first data in the data sequence to be divided and the data refers to the time interval between the original data of the first data and the original data of the data. The time span between other data in the data sequence to be divided is similar. The original data of any data in the data sequence to be divided refers to the data before normalization. The acquisition time corresponding to any data in the data sequence to be divided or the time of acquisition of any data in the data sequence to be divided are the acquisition time corresponding to the original data of that data or the time of acquisition of the original data of that data. Then, based on the correlation between the data on both sides of the local window center data and the difference coefficient of variation of the local window, the window fitness evaluation value of each data in each data in the data to be divided is obtained. The window fitness evaluation value is the key basis for the subsequent adaptive division of the data to be divided or the key to dividing stable segments. For ease of understanding, this embodiment will describe the specific process of obtaining the window fitness evaluation value of the b-th data in any data to be divided as an example, and the data to be divided will be denoted as the data sequence B1, and the time span between the first data and the b-th data in the data sequence B1 is not less than The time span between the last data point and the b-th data point in the data sequence B1 is not less than [a certain value]. The specific process for obtaining the window fitness evaluation value of the b-th data in the data sequence B1, based on the correlation between the data on both sides of the center data of the local window affecting the b-th data in the data sequence B1 and the difference variation coefficient of the local window of the b-th data, is as follows: In the data sequence B1, the time-series data segment consisting of all data to the left of the center data of the local window of the b-th data is denoted as the left-side data segment of the b-th data window, and the time-series data segment consisting of all data to the right of the center data of the local window of the b-th data is denoted as the right-side data segment of the b-th data window. The center data of the local window of the b-th data is the b-th data. The Spearman rank correlation coefficient between the left-side and right-side data segments is calculated, and the result of normalizing the absolute value of the Spearman rank correlation coefficient is denoted as the correlation characterization value. Here, the Sigmoid() function is used for normalization. Sigmoid() is an S-shaped normalization function, whose core function is to map any real number... Within the interval (0,1); in the data sequence B1, the data time series consisting of all data collected within the local window of the b-th data is denoted as the original sequence. The original sequence is second-order differencing to obtain the second-order differencing sequence. The coefficient of variation of the second-order differencing sequence is calculated and denoted as the difference coefficient of variation of the local window of the b-th data. A negative correlation mapping is performed on the difference coefficient of variation, and the mapping result is denoted as the stability characterization value. The coefficient of variation of the sequence is the ratio of the standard deviation to the mean of the sequence multiplied by 100%. Here, a negative exponential function with a base of constant e is used for mapping. The product of the correlation characterization value and the stability characterization value is calculated and denoted as the window fitness evaluation value of the b-th data. The expression for the window fitness evaluation value of the b-th data is: Let be the window fit evaluation value for the b-th data point, sigmoid() be the normalization function, and exp() be the exponential function with base e. Let be the coefficient of variation of the local window for the b-th data point. Let be the Spearman rank correlation coefficient between the left and right data segments of the window for the b-th data point; The larger the value, the more similar or stable the change patterns of the data in the local window of the b-th data are. It is more appropriate to divide the data at the b-th data point using the local window of the b-th data point. It also indicates that dividing the data at the b-th data point in the data sequence B1 to be divided using the local window of the b-th data point is more helpful in subsequently dividing the stable continuous data in the sequence B1 together. The smaller the value, the more stable the changes in the local window of the b-th data point. Therefore, it's more appropriate to divide the sequence at the b-th data point using a local window closer to the range of the b-th data point. This also indicates that dividing the sequence at the b-th data point using a local window closer to the b-th data point is more helpful in subsequently grouping stable, continuous data points together in sequence B1. In other words, the more stable the changes in the local window, the more likely it is that dividing the sequence at the b-th data point using a local window closer to the b-th data point or using a time length closer to the preset time length will extract more stable segments. The greater the sum The smaller, The larger, therefore A larger value indicates that dividing the data at the b-th data point in sequence B1 with a time length closer to the preset time length is more conducive to the extraction of stable segments, and vice versa. The smaller the value, the more difficult it is to extract a stable segment at the b-th data point in sequence B1 by dividing it with a time length close to the preset time length. Alternatively, it indicates that dividing the b-th data point with a local window of the b-th data point is less appropriate. In this case, it is more necessary to adjust the local window size of the b-th data point to a greater extent in order to extract a stable segment.

[0020] After obtaining the window fitness evaluation values ​​of each data point in the data sequence to be segmented, the corresponding data sequence is divided and merged based on these values ​​to obtain the target segmentation time. The specific process is as follows: First, based on the window fitness evaluation values ​​of each data point in the data sequence to be segmented, the corresponding data sequence is divided. The resulting data segments are denoted as subsequences, thus obtaining the subsequences for each data sequence to be segmented, completing the subdivision of the data sequence. Furthermore, the specific process of dividing the corresponding data sequence based on the window fitness evaluation values ​​of each data point in the data sequence to be segmented to obtain the subsequences corresponding to each data sequence is as follows: For the data sequence B1, if the time span from the first data point to the a-th data point in the data sequence B1 to be divided is initially greater than or equal to... Then, the a-th data point is taken as the first data point to be adjusted in the data sequence B1, and the local window length of the first data point to be adjusted is adjusted based on its fitness evaluation value, thus obtaining the adjustment time length corresponding to the first data point to be adjusted. The product of the fitness evaluation value of the first data point to be adjusted and T0 is the adjustment time length corresponding to the first data point to be adjusted. The time interval between the collection time and the time in the data sequence B1 is then considered. and The most recent data are denoted as the first and second initial segmentation data on the influence data sequence B1, respectively. t1 is the time when the original data of the first data to be adjusted was collected, and T1 is the adjustment time length corresponding to the first data to be adjusted. The next data after the second initial segmentation data on the influence data sequence B1 is taken as the second data to be adjusted on the influence data sequence B1. The local window length of the second data to be adjusted is further adjusted based on its fitness evaluation value, resulting in the adjustment time length corresponding to the second data to be adjusted. The product of the fitness evaluation value of the second data to be adjusted and T0 is the adjustment time length corresponding to the second data to be adjusted. The time interval between the collection time and the time interval on the influence data sequence B1 is then considered. The most recent data is denoted as the third initial segmentation data of the data sequence B1 to be divided. t2 is the time when the original data of the second data to be adjusted was collected, and T2 is the adjustment time length corresponding to the second data to be adjusted. This process continues until the time span from the next data of the initial segmentation data to the last data in the data sequence B1 is less than [a certain value]. The iteration stops at a certain point, and all the initial segmentation data on the acquired influence data sequence B1 are used to divide the influence data sequence B1 to be divided. Each segment obtained is recorded as a subsequence on the influence data sequence B1 to be divided, with each pair of adjacent initial segmentation data forming a subsequence. In this embodiment, the data segment formed from the first data point on the influence data sequence B1 to the data point preceding the first initial segmentation data on the influence data sequence B1 is taken as the first subsequence on the influence data sequence B1. The data segment formed from the next data point of the last initial segmentation data on the influence data sequence B1 to the last data point on the sequence B1 is taken as the last subsequence on the influence data sequence B1 to be divided. Except for the endpoints of the subsequences formed by the first two initial segmentation data points on sequence B1, which are the initial segmentation data, the other subsequences formed by two adjacent initial segmentation data points on sequence B1 are formed from the next data point of the earlier initial segmentation data to the later initial segmentation data. The larger the fitness evaluation value of the data to be adjusted, the smaller the adjustment degree of the corresponding local window of the data to be adjusted; conversely, the smaller the fitness evaluation value of the data to be adjusted, the larger the adjustment degree of the corresponding local window of the data to be adjusted. The original data affecting any data in data sequence B1 is unnormalized data measured by the sensor. For example, if the time span from the next data after the c-th initial segmentation data affecting data sequence B1 to the last data in sequence B1 is less than... Then stop iterating.

[0021] Example of determining the first initial segmentation data on the data sequence B1 to be divided: Obtaining the acquisition time of the original data for each data point on the data sequence B1. The time interval between them affects the acquisition time of the original data of data a0 in data sequence B1. If the time interval between the two is the shortest, then data a0 is the first initial segmentation data, and the process of determining other initial segmentation data is the same.

[0022] Then, iterative self-organizing clustering is performed on all subsequences corresponding to each data sequence to be partitioned. The results of iterative self-organizing clustering for all subsequences corresponding to each data sequence to be partitioned are denoted as the subsequence clusters corresponding to the data sequence to be partitioned. That is, iterative self-organizing clustering of all subsequences corresponding to a data sequence to be partitioned yields all subsequence clusters corresponding to that data sequence. The purpose of clustering is to group subsequences of the same type into one class, and the iterative self-organizing clustering process is a well-known technique. Then, adjacent and continuous subsequences belonging to the same cluster on each data sequence to be partitioned are merged. The merged subsequences are denoted as the target subsequences on the corresponding data sequence to be partitioned. That is, if the first and second subsequences on a data sequence to be partitioned belong to the same subsequence cluster, the second and third subsequences on the same data sequence to be partitioned do not belong to the same subsequence cluster. For the same subsequence cluster, the sequence resulting from merging the first and second subsequences is the first target subsequence on the data sequence to be divided. Specifically, for any data sequence to be divided, iterative self-organizing clustering is performed on all corresponding subsequences to obtain the subsequence clusters. Then, adjacent or consecutive subsequences belonging to the same cluster are merged, and the merged subsequences are all recorded as target subsequences on the data sequence to be divided. Next, the acquisition time corresponding to the tail boundary data of each target subsequence or the last data in each target subsequence is recorded as the target segmentation time. In this embodiment, the initial time in the current cumulative monitoring period is also used as the target segmentation time. Subsequently, based on the target segmentation time, the pile foundation vibration data sequence to be divided is used to extract pile foundation vibration data segments under stable working conditions or stable influencing factors.

[0023] Therefore, this embodiment can obtain the target segmentation time through the above process, and then the vibration data sequence of the pile foundation to be segmented will be segmented based on the target time.

[0024] Step S003: Based on the target segmentation time, the vibration data sequence of the pile foundation to be segmented is segmented and identified to obtain the current vibration subsequence and the sample vibration subsequence. The energy intensity feature value and main frequency feature value of the vibration subsequence are obtained. Based on the difference in influence data between the current vibration subsequence and the sample vibration subsequence, the reference subsequence set corresponding to the current vibration subsequence is obtained. Based on the feature value difference, time distance and influence data difference between the current vibration subsequence and the reference subsequence in the reference subsequence set, the amplitude relative anomaly index value and frequency relative anomaly index value of the current vibration subsequence are obtained.

[0025] In this embodiment, after obtaining the target segmentation time, all target segmentation times are used to divide the pile foundation vibration data sequence to be divided, and the resulting subsequences are all recorded as vibration subsequences. It is required that the acquisition time corresponding to the original data of the start and end boundary data of the first vibration subsequence on the pile foundation vibration data sequence to be divided is the target segmentation time, and the acquisition time corresponding to the original data of the tail boundary data of other vibration subsequences is the target segmentation time.

[0026] After dividing the pile foundation vibration data sequence to be divided, the vibration subsequences are identified to obtain the current vibration subsequence and sample vibration subsequences. The sample vibration subsequences are mainly used for anomaly assessment of the current vibration subsequence. The specific process for the current vibration subsequence and sample vibration subsequences is as follows: Among all vibration subsequences, the vibration subsequence containing the current monitoring time is recorded as the current vibration subsequence, that is, the acquisition time of the original data of the last data in the current vibration subsequence is the current monitoring time; a preset sample monitoring time period is constructed. In specific applications, the implementer needs to set the preset sample monitoring time period according to the accuracy of anomaly assessment of the current vibration subsequence. For example, in this embodiment, at least the continuous monitoring time period of the first six months can be used as the preset sample monitoring time period. That is, the length of the preset sample monitoring time period in this embodiment is at least six months, and it is accumulated from the time when the target pile foundation is first monitored; in addition, if the length of the current accumulated monitoring time period is not greater than the length of the preset sample monitoring time period, then the current accumulated monitoring time period is used as the length of the preset sample monitoring time period. Then, all vibration subsequences belonging to the preset sample monitoring time period are recorded as sample vibration subsequences. That is, if the time period formed by the collection time of the original data in a certain sample vibration subsequence belongs to the preset sample monitoring time period, then the vibration subsequence is a sample vibration subsequence.

[0027] Since the characteristics of the influencing factors corresponding to different vibration sub-segments are highly similar when the data of each influencing factor are in a relatively stable state, their corresponding vibration responses should theoretically maintain consistent statistical characteristics under a healthy structural condition. Conversely, when the characteristics of the influencing factors are similar but the vibration responses differ significantly, it can be preliminarily determined that a certain comparative vibration sub-segment may have abnormal vibration or structural state changes, so as to achieve initial screening of abnormal vibrations based on comparison under the same working conditions. Based on the above description, this embodiment will first use the KNN idea and the similarity or difference between the influencing data segments on the influencing data sequence to be divided and the same time period as the current vibration sub-sequence and the sample vibration sub-sequence to obtain the reference sub-sequence set corresponding to the current vibration sub-sequence based on the similarity or difference between the influencing data segments on the influencing data sequence to be divided and the same time period as the current vibration sub-sequence and the sample vibration sub-sequence. The vibration sub-sequences in the reference sub-sequence set are similar to the working conditions or working environment of the current vibration sub-sequence. And the subsequent evaluation of the abnormality of the current vibration sub-sequence is mainly based on the sub-sequences in the reference sub-sequence set. So the specific process of obtaining the reference sub-sequence set corresponding to the current vibration sub-sequence based on the similarity or difference between the influencing data segments on the influencing data sequence to be divided and the same time period as the current vibration sub-sequence and the sample vibration sub-sequence is as follows: First, extract the influence data segments that correspond to the time period of the current vibration subsequence from each influence data sequence to be divided, and record them as the current influence data subsequence corresponding to the current vibration subsequence. Then, extract the influence data segments that correspond to the time period of each sample vibration subsequence from each influence data sequence to be divided, and record them as the sample influence data subsequence corresponding to the corresponding sample vibration subsequence. The time period corresponding to any subsequence refers to the time period from the time when the original data of the first data in the subsequence was collected to the time when the original data of the last data in the subsequence was collected, or the time period formed by the two times when the subsequence was divided. If the time period corresponding to a certain influence data segment in any influence data sequence to be divided is the same as the time period corresponding to a certain vibration subsequence, then the influence data segment is the influence data segment in the influence data sequence to be divided that corresponds to the time period corresponding to the vibration subsequence. Each influence data subsequence corresponding to any vibration subsequence belongs to a different influence data sequence to be divided, and the time period corresponding to each influence data subsequence corresponding to any vibration subsequence is consistent with the time period corresponding to the vibration subsequence. Moreover, the number of influence data subsequences corresponding to each vibration subsequence is the number of influence data sequences to be divided.

[0028] Then, based on the DTW distance between the current influence data subsequence corresponding to the current vibration subsequence and the sample influence data subsequence corresponding to each sample vibration subsequence, the DTW distance set between the current vibration subsequence and each sample vibration subsequence is obtained. Furthermore, the h-th DTW distance in the DTW distance set between the current vibration subsequence and any sample vibration subsequence is the DTW distance between the h-th current influence data subsequence corresponding to the current vibration subsequence and the h-th sample influence data subsequence corresponding to that sample vibration subsequence, calculated using the dynamic programming algorithm. The h-th current influence data subsequence corresponding to the current vibration subsequence and the h-th sample influence data subsequence corresponding to any sample vibration subsequence belong to the same influence data sequence to be divided. Next, the mean of the DTW distance set between the current vibration subsequence and each sample vibration subsequence is recorded as the influence data difference value between the current vibration subsequence and the corresponding sample vibration subsequence. The sample vibration subsequences are sorted in ascending order of influence data difference values, and the set of the first predetermined number of sample vibration subsequences is recorded as the reference subsequence set corresponding to the current vibration subsequence. The smaller the influence data difference value, the more similar they are. In practical applications, implementers need to set the preset quantity value according to the actual situation such as the level of statistical stability requirements. If the requirement for statistical stability is high, a larger preset quantity value can be set, but the preset quantity cannot be greater than the total number of sample vibration subsequences. For example, in this embodiment, the preset quantity can be set to 100.

[0029] Since pile foundation vibration anomalies often manifest as changes in frequency structure and amplitude energy anomalies, this embodiment aims to improve the accuracy of subsequent vibration anomaly identification and early warning. This embodiment will extract features from the vibration subsequence in both the time and frequency domains, and analyze the anomalies of the current vibration subsequence in both domains based on the extracted features. Specifically, this embodiment will extract the feature values ​​of the vibration subsequence, including energy intensity feature values ​​and main frequency feature values. Energy intensity feature values ​​are time-domain features. The anomalies of the current vibration subsequence will be assessed based on the difference in feature values ​​between the current vibration subsequence and the reference subsequence in the reference subsequence set. The specific process for obtaining the energy intensity feature values ​​and main frequency feature values ​​of the vibration subsequence is as follows: For any vibrational subsequence: the mean of the squares of all data in the vibrational subsequence is denoted as the energy intensity characteristic value of the vibrational subsequence, that is, the energy intensity characteristic value of the vibrational subsequence is the square mean of the vibrational subsequence, and the expression for the energy intensity characteristic value of the vibrational subsequence is: , where m is the total amount of data in the vibrational subsequence. Let m be the square of the m-th data point in the oscillating subsequence. Perform a short-time Fourier transform on the oscillating subsequence to obtain its spectrum. Based on the spectrum, obtain the centroid of the oscillating subsequence and denote it as the principal frequency characteristic value of the oscillating subsequence. The centroid of the oscillating subsequence is the sum of the products of each frequency in the spectrum and the normalized amplitude of the corresponding frequency. The normalized amplitude of each frequency in the spectrum is the ratio of the amplitude corresponding to that frequency to the sum of the amplitudes corresponding to all frequencies in the spectrum. The amplitude used when determining the centroid is the amplitude on the spectrum.

[0030] After obtaining the feature values ​​of the vibrational subsequence, the relative amplitude anomaly index value of the current vibrational subsequence is obtained based on the differences in energy intensity feature values, time distance, and influence data between the current vibrational subsequence and the reference subsequences in the reference subsequence set; the relative frequency anomaly index value of the current vibrational subsequence is obtained based on the differences in the main frequency feature values, time distance, and influence data between the current vibrational subsequence and the reference subsequences in the reference subsequence set. The methods for obtaining the relative frequency anomaly index value and the relative amplitude anomaly index value of the current vibrational subsequence are not described in detail hereafter; therefore, the specific process for obtaining the relative amplitude anomaly index value of the current vibrational subsequence is as follows: The energy intensity difference between the current vibrational subsequence and each reference subsequence in the reference subsequence set is calculated. The energy intensity difference between the current vibrational subsequence and the j-th reference subsequence in the reference subsequence set is the absolute value of the difference between the energy intensity feature value of the current vibrational subsequence and the energy intensity feature value of the j-th reference subsequence. The confidence weight of each reference subsequence is also calculated. The confidence weight of the j-th reference subsequence is the proportionally normalized result of multiplying the influence data similarity value between the current vibrational subsequence and the j-th reference subsequence by the normalized time distance between the current vibrational subsequence and the j-th reference subsequence. The influence data similarity value between the current vibrational subsequence and the j-th reference subsequence is the ratio of the influence data similarity value between the current vibrational subsequence and the j-th reference subsequence. The result of inverse normalization of the impact data difference values ​​is as follows: Inverse normalization refers to first normalizing the data and then subtracting the normalized result from the constant 1. The inverse normalization result reflects the similarity between the impact data subsequence corresponding to the current vibration subsequence and the impact data subsequence corresponding to the j-th reference subsequence; the larger the value, the greater the similarity. The weighted energy intensity difference value between the current vibration subsequence and each reference subsequence in the reference subsequence set is multiplied by the corresponding confidence weight value of the reference subsequence. The sum of the weighted energy intensity difference values ​​between the current vibration subsequence and all reference subsequences in the reference subsequence set is denoted as the amplitude relative anomaly index value of the current vibration subsequence. The specific expression for the amplitude relative anomaly index value of the current vibration subsequence is: Where Q1 is the relative amplitude anomaly index value of the current vibrational subsequence, N is the total number of reference subsequences in the set of reference subsequences corresponding to the current vibrational subsequence, and E0 is the energy intensity characteristic value of the current vibrational subsequence. Let j be the energy intensity eigenvalue of the j-th reference subsequence in the set of reference subsequences. Let J be the confidence weight value of the j-th reference subsequence. This represents the difference in influence data between the current vibrational subsequence and the j-th reference subsequence, where Norm() is the normalization function. The interval between the current vibration subsequence and the j-th reference subsequence is the time interval between the original acquisition time of the first data in the two sequences; The larger the value, the more inconsistent the working conditions or working environment are between the current vibration subsequence and the j-th reference subsequence. Therefore, when analyzing the relative anomaly of the current vibration subsequence based on the j-th reference subsequence, the reliability or participation of the j-th reference subsequence is lower. This is to enhance cross-timeframe comparison capabilities, reduce the impact of nearest-neighbor data assimilation effects when anomalies evolve slowly, and thus improve the accuracy of anomaly assessment. Specifically, to improve the accuracy of anomaly assessment, the reference value of a reference subsequence closer in time to the current vibrational subsequence should be greater. The smaller the value, the lower the credibility or participation of the j-th reference subsequence; while smaller and When it is larger, The smaller, therefore The smaller the value, the lower the confidence or participation of the j-th reference subsequence when evaluating the relative anomaly index value of the amplitude of the current vibrational subsequence, and vice versa. When the value of Q1 is larger, the confidence or participation of the j-th reference subsequence is higher when evaluating the relative anomaly index value of the amplitude of the current vibrational subsequence. A larger difference between the energy intensity eigenvalue of the current vibrational subsequence and that of the reference subsequence, along with a larger confidence weight value for the reference subsequence, indicates a greater likelihood of anomalous vibration in the current vibrational subsequence in terms of amplitude. Conversely, a larger difference between the energy intensity eigenvalue of the previous vibrational subsequence and that of the reference subsequence, along with a larger confidence weight value for the reference subsequence, results in a larger Q1. Therefore, a larger Q1 indicates a greater likelihood of anomalous vibration in the current vibrational subsequence in terms of amplitude, and vice versa.

[0031] Therefore, this embodiment obtains the relative amplitude anomaly index value and the relative frequency anomaly index value of the current vibration subsequence through the above process.

[0032] Step S004: Based on the relative abnormality index values ​​of amplitude and frequency of the current vibration subsequence, perform early warning of pile foundation vibration anomalies.

[0033] After obtaining the relative anomaly index values ​​of the current oscillator sequence, this embodiment further combines historical oscillator data for statistical analysis, extracting the data distribution characteristics with the highest proportion and representing stable operating conditions. The corresponding central tendency is used as a reference benchmark under normal operation or as a threshold for determining whether to issue a warning. This involves obtaining frequency anomaly warning values ​​and amplitude anomaly warning values. These anomaly warning values ​​are the basis for subsequent judgments on whether a vibration anomaly warning is needed. The specific process for obtaining the warning values ​​is as follows: First, the relative amplitude and relative frequency anomaly indices of the sample vibration subsequences are obtained. In this embodiment, the methods for obtaining the relative amplitude and relative frequency anomaly indices of other vibration subsequences are the same as those for obtaining the current vibration subsequence. However, the reference subsequence for the sample vibration subsequence is selected from all remaining sample vibration subsequences excluding the corresponding sample vibration subsequence. Then, the median of the set formed by the relative amplitude anomaly indices of all sample vibration subsequences is used as the representative amplitude index value, and the median of the set formed by the relative frequency anomaly indices of all sample vibration subsequences is used as the representative frequency index value. Afterward, based on the relative amplitude anomaly indices of each sample vibration subsequence, the difference between the relative amplitude anomaly indices of each sample vibration subsequence and the representative amplitude index value, and the time interval between each sample vibration subsequence and the initial monitoring time, an amplitude anomaly warning measurement value is obtained. Finally, based on the relative frequency anomaly indices of each sample vibration subsequence... The frequency anomaly warning measure is obtained by taking the abnormal index value, the difference between the frequency of each sample vibration subsequence relative to the abnormal index value and the representative frequency index value, and the time interval between each sample vibration subsequence and the initial monitoring time. Alternatively, as another implementation, the median of the set of amplitude relative to the abnormal index values ​​of all historical vibration subsequences can be used as the representative amplitude index value. Based on the amplitude relative to the abnormal index values ​​of each historical vibration subsequence, the difference between the amplitude relative to the abnormal index values ​​of each historical vibration subsequence and the representative amplitude index value, and the time interval between each historical vibration subsequence and the initial monitoring time, an amplitude anomaly warning measure is obtained. Historical vibration subsequences are all vibration subsequences other than the current vibration subsequence. The frequency anomaly warning measure is obtained similarly. The initial monitoring time is the initial time of vibration anomaly monitoring of the target pile foundation. The method for obtaining the frequency anomaly warning measure is the same as the method for obtaining the amplitude anomaly warning measure; therefore, the specific process of obtaining the frequency anomaly warning measure will not be described in this embodiment.

[0034] The specific process for obtaining the amplitude anomaly warning measure value based on the amplitude relative anomaly index value of each sample vibration subsequence, the difference between the amplitude relative anomaly index value of each sample vibration subsequence and the representative amplitude index value, and the time interval between each sample vibration subsequence and the initial monitoring time is as follows: The time-domain reference weight of the corresponding sample vibration subsequence is defined as the product of the inversely normalized result of the absolute value of the difference between the relative anomaly index value of the amplitude of each sample vibration subsequence and the value representing the amplitude index, and the inversely normalized result of the product of the time interval between the corresponding sample vibration subsequence and the initial monitoring time. The expression for the time-domain reference weight of the v-th sample vibration subsequence is: Where V is the total number of sample vibrational subsequences, exp() is the exponential function with base e, and Norm() is the normalization function. Let be the absolute value of the difference between the relative anomaly index value of the amplitude of the v-th sample vibration subsequence and the representative amplitude index value. The time interval between the v-th sample vibration subsequence and the initial monitoring time refers to the time interval between the original acquisition time of the first data in the subsequence and the time corresponding to the initial monitoring time. Since the probability of structural vibration anomalies in the early stages of pile foundation use is relatively low, or the pile foundation is in a stable operating phase during the early stages of use, this embodiment ensures the reliability of the calculated amplitude anomaly warning measure by making the sample vibration subsequence closer to the initial monitoring time contribute more to the calculation of the amplitude anomaly warning measure. This reduces the interference of data from the later deterioration period on the results. The smaller the value, the greater the contribution or participation of the v-th sample vibration subsequence in calculating the amplitude anomaly early warning measure. The smaller the value, the closer the vibration subsequence of the v-th sample is to the center of the normal distribution of the population. Therefore, its reference weight in the amplitude anomaly warning measure should be increased to improve the robustness of the amplitude anomaly warning measure. The smaller the value, the greater the contribution or participation of the v-th sample vibration subsequence in calculating the amplitude anomaly early warning measure; while smaller and The smaller the value, the greater the time-domain reference weight of the v-th sample vibration subsequence. A greater time-domain reference weight indicates a greater contribution or participation of the v-th sample vibration subsequence's amplitude relative to the anomaly index value when calculating the amplitude anomaly warning measure. The sum of the product of the time-domain reference weights of all sample vibration subsequences and the corresponding amplitude relative anomaly index values ​​is denoted as the amplitude anomaly warning measure. The expression for the amplitude anomaly warning measure is: QC is the early warning value for abnormal amplitude. Let be the relative anomaly index value of the amplitude of the v-th sample vibrational subsequence. The time-domain reference weight is the v-th sample vibration subsequence; the larger the relative amplitude anomaly index value and the time-domain reference weight of the sample vibration subsequence, the larger the amplitude anomaly warning measure value.

[0035] In this embodiment, after obtaining the amplitude anomaly warning measurement value and the frequency anomaly warning measurement value, the amplitude warning index value and the frequency warning index value are obtained based on the amplitude anomaly warning measurement value and the frequency anomaly warning measurement value of the current vibration subsequence, as well as the amplitude relative anomaly index value and the frequency relative anomaly index value of the current vibration subsequence. The specific acquisition process is as follows: the normalized result of the ratio of the amplitude relative anomaly index value of the current vibration subsequence to the amplitude anomaly warning measurement value is recorded as the current amplitude warning index value, and the normalized result of the ratio of the frequency relative anomaly index value of the current vibration subsequence to the frequency anomaly warning measurement value is recorded as the current frequency warning index value. Here, the normalization function Sigmoid() is used to achieve normalization. The Sigmoid() function is a classic S-shaped nonlinear normalization function, which is often used to map any real number to the interval between 0 and 1. The closer the relative anomaly index value is to the corresponding anomaly warning measurement value, the larger the ratio, that is, the larger the warning index value, indicating that the pile foundation has a greater possibility of pile foundation vibration anomaly or that the pile foundation is about to have a pile foundation vibration anomaly.

[0036] Next, based on the aforementioned warning index values, it is determined whether to issue a warning for abnormal pile foundation vibration. Specifically, it is determined whether there is an index value greater than the preset warning threshold among the current amplitude warning index value and the current frequency warning index value. If so, it indicates that the vibration response of the pile foundation has deviated from the normal fluctuation range, and a warning prompt needs to be triggered to conduct further verification or on-site testing. That is, a vibration abnormality warning should be issued immediately. If neither the current amplitude warning index value nor the current frequency warning index value is greater than the preset warning threshold, it indicates that the pile foundation is operating normally and no vibration abnormality warning is required. In specific applications, implementers need to set the preset warning threshold according to monitoring accuracy requirements, pile foundation type, experimental statistics, engineering operation and maintenance experience, and the three-standard-deviation principle. In this embodiment, it is set to 0.6.

[0037] Thus, this embodiment completes the intelligent analysis of pile foundation vibration monitoring data or the early warning of abnormal pile foundation vibration. Furthermore, the amplitude relative anomaly index value and frequency relative anomaly index value obtained by this embodiment based on the analysis in both the time domain and frequency domain can improve the stability and adaptability of identifying abnormal pile foundation vibration, thereby improving the accuracy and reliability of early warning of abnormal pile foundation vibration, and thus enhancing the reference value for safety assessment of actual engineering projects.

[0038] This embodiment of an intelligent analysis system for pile foundation vibration monitoring data includes a memory and a processor. The processor executes a computer program stored in the memory to implement the aforementioned intelligent analysis method for pile foundation vibration monitoring data.

[0039] In summary, this embodiment first obtains the pile foundation vibration data sequence to be segmented and the influence data sequence to be segmented; then, it constructs a local window of a preset time length centered on each data point in each influence data sequence to be segmented, and records this window as the local window for the corresponding data. Based on the correlation between the data on both sides of the central data point of the local window and the difference coefficient of variation of the local window, it obtains the window fitness evaluation value for each data point in each influence data sequence to be segmented. Based on the window fitness evaluation value, it segments and merges the influence data sequences to be segmented to obtain the target segmentation time; finally, it segments and identifies the pile foundation vibration data sequence to be segmented based on the target segmentation time. The process involves obtaining the current vibration subsequence and sample vibration subsequences, acquiring their energy intensity and main frequency characteristic values, and, based on the difference in influence data between the current and sample vibration subsequences, obtaining a set of reference subsequences corresponding to the current vibration subsequence. Furthermore, based on the difference in characteristic values, time distance, and influence data between the current and reference subsequences in the reference set, the relative amplitude and frequency anomaly indices of the current vibration subsequence are obtained. Finally, based on these values, a pile foundation vibration anomaly early warning system is implemented. The amplitude and frequency anomaly indices obtained from both time and frequency domain analyses improve the ability to detect concealed and slowly evolving pile foundation vibration anomalies, and enhance the monitoring and analysis accuracy under different working conditions. This, in turn, improves the accuracy and reliability of pile foundation vibration anomaly early warning, thereby increasing its reference value for actual engineering safety assessments.

[0040] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 this application, and should all be included within the protection scope of this application.

Claims

1. An intelligent analysis method for pile foundation vibration monitoring data, characterized in that, The method includes the following steps: Obtain the pile foundation vibration data sequence to be divided and the influence data sequence to be divided, wherein the influence data sequence to be divided includes the pile foundation load, pile foundation temperature and pile foundation wind load influence data sequence to be divided; A local window of a preset time length is constructed with each data point in each data sequence to be divided as the center, and is denoted as the local window of the corresponding data. Based on the correlation between the data on both sides of the data center of the local window and the difference variation coefficient of the local window, the window fitness evaluation value of each data point in each data sequence to be divided is obtained. Based on the window fitness evaluation value, the data sequences to be divided are divided and merged to obtain the target segmentation time. Based on the target segmentation time, the vibration data sequence of the pile foundation to be segmented is divided and identified to obtain the current vibration subsequence and the sample vibration subsequence. The energy intensity feature value and main frequency feature value of the vibration subsequence are obtained. Based on the difference in influence data between the current vibration subsequence and the sample vibration subsequence, the reference subsequence set corresponding to the current vibration subsequence is obtained. Based on the feature value difference, time distance and influence data difference between the current vibration subsequence and the reference subsequence in the reference subsequence set, the amplitude relative anomaly index value and frequency relative anomaly index value of the current vibration subsequence are obtained. Based on the relative abnormality values ​​of amplitude and frequency of the current vibration subsequence, early warning of pile foundation vibration anomalies is carried out.

2. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 1, characterized in that, Methods for obtaining window fit evaluation values ​​include: For any data in any data sequence to be divided, the normalized result of the absolute value of the Spearman rank correlation coefficient between the data segments on both sides of the central data in the local window of the data is denoted as the correlation characterization value. The negative correlation mapping result of the coefficient of variation of the second difference sequence of the local window of the data is denoted as the stability characterization value. The product of the correlation characterization value and the stability characterization value is denoted as the window fitness evaluation value of the data.

3. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 1, characterized in that, Methods for obtaining the target segmentation time include: Based on the window fitness evaluation value of each data in each data sequence to be divided, the corresponding data sequence to be divided is divided to obtain the subsequences of each data sequence to be divided. Iterative self-organizing clustering is performed on all subsequences of each data sequence to be divided to obtain the subsequence clusters corresponding to each data sequence to be divided. The subsequences that belong to the same cluster and are continuous are merged, and the merged subsequences are all recorded as the target subsequences of the corresponding data sequence to be divided. The acquisition time corresponding to the tail boundary data of each target subsequence is recorded as the target segmentation time.

4. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 3, characterized in that, Methods for obtaining subsequences include: For any data sequence to be divided, if the time span from the first data point to the a-th data point in the data sequence to be divided is initially greater than or equal to... Then select the a-th data as the first data to be adjusted, and... and The nearest neighbor data are denoted as the first initial segmentation data and the second initial segmentation data, respectively. The next data after the second initial segmentation data is selected as the second data to be adjusted, and... The nearest neighbor data is denoted as the 3rd initial segment data. The process of determining the initial segment data is iterated until the maximum time span of the remaining data is less than 3. Stop at a certain time, and use the obtained initial segmentation data to divide the data sequence to be divided, to obtain the subsequences on the data sequence to be divided; T0 is the preset time length, t1 is the time when the first data to be adjusted is collected, T1 is the product of the window fitness evaluation value of the first data to be adjusted and T0, t2 is the time when the second data to be adjusted is collected, T2 is the product of the window fitness evaluation value of the second data to be adjusted and T0.

5. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 1, characterized in that, Current methods for obtaining vibrational subsequences and sample vibrational subsequences include: All subsequences obtained by dividing the pile foundation vibration data sequence to be divided at all target segmentation times are denoted as vibration subsequences. Vibration subsequences containing the current monitoring time are denoted as current vibration subsequences. All vibration subsequences belonging to the preset sample monitoring time period are denoted as sample vibration subsequences.

6. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 5, characterized in that, Methods for obtaining the reference subsequence set include: For each data sequence to be divided, extract the data segments with the same time period as the current vibration subsequence and record them as the current impact data subsequence corresponding to the current vibration subsequence. Extract the data segments with the same time period as the sample vibration subsequence and record them as the sample impact data subsequence corresponding to the corresponding sample vibration subsequence. Based on the DTW distance between the current impact data subsequence corresponding to the current vibration subsequence and the sample impact data subsequence corresponding to each sample vibration subsequence, a set of DTW distances between the current vibration subsequence and each sample vibration subsequence is obtained. The h-th DTW distance in the set of DTW distances between the current vibration subsequence and the g-th sample vibration subsequence is the DTW distance between the h-th current impact data subsequence corresponding to the current vibration subsequence and the h-th sample impact data subsequence corresponding to the g-th sample vibration subsequence. The h-th current impact data subsequence and the h-th sample impact data subsequence corresponding to the g-th sample vibration subsequence belong to the same impact data sequence to be divided. The mean of the set of DTW distances between the current vibration subsequence and each sample vibration subsequence is recorded as the impact data difference value between the current vibration subsequence and the corresponding sample vibration subsequence. The sample vibration subsequences are sorted in ascending order of impact data difference value, and the set of the first preset number of sample vibration subsequences is recorded as the reference subsequence set corresponding to the current vibration subsequence.

7. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 6, characterized in that, The methods for obtaining the relative anomaly values ​​of amplitude and frequency of the current vibrational subsequence include: Based on the differences in energy intensity characteristic values, temporal distance, and influence data between the current vibrational subsequence and the reference subsequences in the reference subsequence set, the relative amplitude anomaly index value of the current vibrational subsequence is obtained; based on the differences in main frequency characteristic values, temporal distance, and influence data between the current vibrational subsequence and the reference subsequences in the reference subsequence set, the relative frequency anomaly index value of the current vibrational subsequence is obtained using the same method as the relative amplitude anomaly index value of the current vibrational subsequence. The method for obtaining the relative anomaly index value of the amplitude of the current vibration subsequence includes: multiplying the absolute value of the difference in energy intensity characteristic values ​​between the current vibration subsequence and each reference subsequence in the reference subsequence set by the confidence weight value of the corresponding reference subsequence, and recording the result as the weighted energy intensity difference value between the current vibration subsequence and the corresponding reference subsequence; summing the weighted energy intensity difference values ​​between the current vibration subsequence and all reference subsequences in the reference subsequence set, and recording the relative anomaly index value of the amplitude of the current vibration subsequence; the confidence weight value of the j-th reference subsequence is the proportional normalized result of multiplying the inverse normalized result of the difference in influence data between the current vibration subsequence and the j-th reference subsequence by the normalized time distance between the current vibration subsequence and the j-th reference subsequence.

8. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 1, characterized in that, A method for early warning of pile foundation vibration anomalies based on the relative amplitude and relative frequency anomalies of the current vibration subsequence includes: The median of the set of relative amplitude anomaly index values ​​for all sample vibration subsequences is used as the representative amplitude index value. An amplitude anomaly warning measure is obtained based on the amplitude relative anomaly index values ​​for each sample vibration subsequence, the difference between the amplitude relative anomaly index values ​​for each sample vibration subsequence and the representative amplitude index value, and the time interval between each sample vibration subsequence and the initial monitoring time. Similarly, the median of the set of relative frequency anomaly index values ​​for all sample vibration subsequences is used as the representative frequency index value. A frequency anomaly warning measure is obtained based on the frequency relative anomaly index values ​​for each sample vibration subsequence, the difference between the frequency relative anomaly index values ​​for each sample vibration subsequence and the representative frequency index value, and the time interval between each sample vibration subsequence and the initial monitoring time. The initial monitoring time is the initial time of vibration anomaly monitoring of the target pile foundation. The method for obtaining the frequency anomaly warning measure is the same as that for obtaining the amplitude anomaly warning measure. The normalized result of the ratio of the relative amplitude anomaly index value of the current vibration subsequence to the amplitude anomaly warning measure value is recorded as the amplitude warning index value. The normalized result of the ratio of the relative frequency anomaly index value of the current vibration subsequence to the frequency anomaly warning measure value is recorded as the frequency warning index value. When the warning index value is greater than the preset warning threshold, the pile foundation vibration anomaly warning is immediately issued.

9. The intelligent analysis method for pile foundation vibration monitoring data as described in claim 8, characterized in that, Methods for obtaining amplitude anomaly early warning measurement values ​​include: The proportional normalization result of the product of the absolute value of the difference between the relative abnormality index value of the amplitude of each sample vibration subsequence and the value representing the amplitude index, and the product of the inverse normalization result of the time interval between the corresponding sample vibration subsequence and the initial monitoring time, is denoted as the time domain reference weight of the corresponding sample vibration subsequence. The sum of the product of the time-domain reference weights of all sample vibration subsequences and the relative amplitude anomaly index values ​​of the corresponding sample vibration subsequences is recorded as the amplitude anomaly warning measure.

10. An intelligent analysis system for pile foundation vibration monitoring data, comprising a memory and a processor, characterized in that, The processor executes the computer program stored in the memory to implement an intelligent analysis method for pile foundation vibration monitoring data as described in any one of claims 1-9.

Images