Bridge monitoring time series data similarity measurement method based on pattern discrimination

Abstract

The application discloses a bridge monitoring time series data similarity measurement method based on mode division, acquires environment data, structure response data and time series data of an abnormal measuring point under normal load working conditions of a bridge health monitoring system and carries out pretreatment, then carries out slope calculation of bridge monitoring time series data in each time period unit; the similarity degree of bridge monitoring time series data in each time period unit is considered through two judgment modes respectively; mode division calculation of bridge monitoring time series data is carried out based on two judgment results; the similarity degree of bridge monitoring time series data is measured based on mode division distance, and bridge monitoring time series data similarity measurement is carried out according to a high-dimensional data similarity measurement function. The application provides the possibility of direct comparison of the similarity between bridge monitoring time series data of different attributes, reduces the conversion process of the different attribute data, and fully reflects the dynamic characteristics of the time series.

Description

Technical Field

[0001] This invention relates to the fields of bridge health monitoring (SHM system), engineering project management, and bridge risk assessment, specifically to a method for measuring the similarity of bridge monitoring time series data based on pattern differentiation. Background Technology

[0002] my country has achieved remarkable success in bridge construction, moving from a major bridge-building nation to a leading bridge-building power. Intelligent construction technology has become indispensable in the field of bridge construction engineering, experiencing rapid development in bridge monitoring, detection, evaluation, and management technologies. A range of sensors and monitoring products are widely used in projects such as bridge dynamic monitoring. Existing sophisticated monitoring systems have accumulated a large amount of data that urgently needs to be fully interpreted. The main data generated is time-series data. Bridge monitoring time-series data is closely related to time, representing the results of recording and observing phenomena in chronological order, and can be used to describe how phenomena change over time. Due to the high dimensionality and complexity of bridge monitoring time-series data, data mining is challenging. Therefore, current research on clustering analysis of bridge monitoring time-series data often focuses on feature representation, similarity measurement, and clustering algorithms. The theoretical basis for distance measurement in the time-series similarity search process mainly includes Euclidean distance and dynamic time curvature theory.

[0003] Similarity analysis of bridge monitoring time series data is of great significance. Conventional data similarity analysis mainly employs Euclidean distance or clustering methods. However, Euclidean distance is highly sensitive to small distortions along the time axis, which can lead to it failing to provide an intuitively accurate measure of similarity between two sequences. Furthermore, most commonly used methods only consider numerical differences between identical sampling points on the curve, neglecting changes in curve displacement; existing dynamic matching methods have excessive computational costs and may result in excessive bending.

[0004] Therefore, a method for measuring the similarity of bridge monitoring time series data based on pattern discrimination is invented. This method uses morphological similarity as the primary similarity criterion to determine the degree of similarity between different sequences, making it particularly important. On the one hand, this method is more likely to find data periods with high similarity than distance-based methods; on the other hand, it aligns well with actual conditions. Applying this method to bridge data monitoring and management systems can improve work efficiency. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a similarity measurement method for bridge monitoring time series data based on pattern differentiation. By effectively combining different sequence data with morphological features for pattern differentiation, and then performing comparative analysis of the morphological similarity of data change trends in the classification, the similarity measurement of different data sequences is optimized.

[0006] The technical solution of the present invention is as follows: A method for measuring the similarity of bridge monitoring time series data based on pattern differentiation, comprising the following steps:

[0007] (1) Obtain environmental data, structural response data and bridge monitoring time series data for several time periods of different measurement points under normal load conditions through the bridge monitoring system, including reference bridge monitoring time series data S and comparison bridge monitoring time series data S′ for similarity measurement. Preprocess the obtained bridge monitoring time series data, and divide the preprocessed bridge monitoring time series data into segments according to time period, and calculate the slope of the straight line segment in each time period unit.

[0008] (2) Make trend judgments. By judging whether the slope of each segment of the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′ is the same, divide each segment of bridge monitoring time series data into two categories: same-direction type and opposite-direction type.

[0009] (3) Make a similarity judgment, calculate the angle difference between the two straight line segments of the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′, and compare it with the maximum angle difference of the data sample to conduct a similarity comparison analysis;

[0010] (4) Based on the trend judgment and the similarity judgment results, the similarity of bridge monitoring time series data in each time period unit is considered, and the product of the two results is used as the pattern discrimination distance;

[0011] (5) The similarity of bridge monitoring time series data is measured based on the high-dimensional data similarity measurement function based on pattern discrimination distance.

[0012] Furthermore, bridge monitoring time series data under normal load conditions are obtained through the bridge monitoring system, and the bridge monitoring time series data is defined as Y, as shown in the following formula:

[0013] Y={(y1,T1),(y2,T2),(y3,T3),...,(y i ,T i )}

[0014] In the formula: y i For the i-th time period T i Numerical values ​​of bridge monitoring time series data within the region.

[0015] Further, the bridge monitoring time series data described in step (1) is preprocessed, including: performing dimensionless processing on the data and differential operations to eliminate the influence of long-term trends, and dividing the i-th time period unit T i The trend of value change within a given time period is represented by the slope k of the straight line segment of that time period. i Indicates that the slope k i As shown in the following formula, the time series data points are separated by one hour, in hours, i.e., T. i+1 -T i =1;

[0016]

[0017] In the formula: k i For time period T i To T i+1 The slope of the straight line segment in the bridge monitoring time series data; i = 1, 2, 3...n, where n is the number of segments in the entire bridge monitoring time series data.

[0018] Furthermore, the bridge monitoring time series data Y is redefined and denoted by M, as follows:

[0019] M={(ju1,u1,k1),(ju2,u2,k2),...(ju n ,u n ,k n )}

[0020] In the formula: ju i Represented as T i To T i+1 The trend judgment results of the time series, u i Represented as T i To T i+1 The similarity judgment results of time series.

[0021] Further, the trend judgment in step (2) specifically involves: calculating the slope of the straight line segment in each time period unit of the selected reference bridge monitoring time series data S and the comparison bridge monitoring time series data S' for similarity measurement, respectively, with the slope of the straight line segment in the i-th time period unit denoted as k. i and k i ', and perform a main trend comparison to determine the consistency of the changing trends. The trend judgment result is recorded as ju. i The judgment criteria are as follows:

[0022]

[0023] In the formula: ju iFor the discrimination result; k i For reference, bridge monitoring time series data S in T i To T i+1 The slope of the straight line segment within the period; k i 'To compare time series data S' in T i To T i+1 The slope of the straight line segment within the period.

[0024] By classifying the reference bridge monitoring time series data S and the comparative bridge monitoring time series data S' for similarity measurement, ju i =1 indicates the first type of same direction, ju i =-1 represents the second type of anisotropic type.

[0025] Further, the similarity judgment in step (3) specifically involves: calculating the slope of the straight line segments in each time period unit of the selected reference bridge monitoring time series data S and the comparison bridge monitoring time series data S' for similarity measurement, and using the slopes of the two straight line segments and trigonometric function formulas to calculate the angle difference between the two straight line segments, as shown in the following formula:

[0026] k i =tanαk i =tanβ

[0027]

[0028]

[0029] In the formula: k i For reference, bridge monitoring time series data S in T i To T i+1 The slope of the straight line segment within the period; k i 'To compare time series data S' in T i To T i+1 The slope of the straight line segment within the period; α is the reference bridge monitoring time series data S in T i To T i+1 The angle between the straight line segment and the horizontal line segment within the period; β is the angle between the bridge monitoring time series data S' and T. i To T i+1 The angle between the straight line segment and the horizontal line segment within the period; Δk i It represents the angle difference between two straight line segments.

[0030] Further, in step (4), based on the trend judgment result and the similarity judgment result, the similarity of the bridge monitoring time series data within each time period unit is considered respectively. Specifically, based on the trend judgment, the slopes of the straight line segments of the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′ within each time period unit are compared for trend, and ju is defined as follows: i =1 indicates the first type of same direction, ju i =-1 indicates the second type of anisotropic type;

[0031] By judging the trend and determining the degree of similarity, when ju i =1 for the first type of same direction, when |Δk i |Exceeds the maximum angular difference Δk of the data samples max If the similarity is 50%, then it is not included in the calculation process during the similarity assessment, and the similarity assessment value u is taken. i =0; when |Δk i |The maximum angular difference Δk in the data sample max 20% of the maximum angle difference Δk between the data samples max When the similarity is between 50% and 50%, a certain reduction is considered in the similarity judgment process, and the similarity judgment value is taken. When |Δk i |The maximum angle difference Δk in the data samples was not exceeded. max When the similarity is 20%, no reduction is considered in the similarity judgment process, and the similarity judgment value u is taken. i =1;

[0032] When ju i When -1 represents the second type of heterogeneous case, for k i 'Through symmetry processing, if |Δk i |The maximum angle difference Δk in the data samples was not exceeded. max When the similarity is 20%, its negative impact on the similarity is not considered in the similarity judgment process, and the similarity judgment value u is taken. i =0; when |Δk i |The maximum angular difference Δk in the data sample max 20% of the maximum angle difference Δk between the data samples max If the similarity is between 50% and 50%, then a certain reduction should be considered in the similarity judgment process, and the similarity judgment value should be taken. When |Δk i |Exceeds the maximum angular difference Δk of the data samples max When the similarity is 50%, its negative impact on the degree of similarity needs to be considered in the similarity judgment process, and the similarity judgment value u is taken. i =1.

[0033] By considering the similarity of bridge monitoring time series data within each time period unit based on the two judgment results, the result is denoted as d. si As shown in the following formula:

[0034] ds i =ju i ×u i

[0035] In the formula: d si To distinguish distances for patterns; ju i The result is based on trend judgment; u i The result of the similarity judgment;

[0036] Furthermore, the pattern distinguishing distance d si A similarity comparison analysis was performed to obtain the similarity measurement judgment dataset U:

[0037] U = {x | -1 ≤ x ≤ 1}

[0038] Where x is d si The value of is [0,1]. When the value is closer to 1, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar. When the value is [-1,0], the value is closer to 0, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar.

[0039] Furthermore, the high-dimensional data similarity measure function based on pattern discrimination distance in step (5) is denoted as Hs(S,S'), as follows:

[0040]

[0041] In the formula: Hs(S,S') is a high-dimensional data similarity measurement function; d si Distinguish the distance for the pattern; Δk i The difference in angle between two line segments is denoted as Hs. High-dimensional similarity measurement methods calculate the degree of similarity based on the trend and shape of the data. The similarity measurement value ranges from [0,1]. The closer the similarity is, the closer the similarity measurement value Hs is to 1, i.e., Hs≈1; conversely, the more the similarity is different, the closer the similarity measurement value Hs is to 0, i.e., Hs≈0. This is used to assess the degree of similarity between high-dimensional data.

[0042] The beneficial effects of this invention are:

[0043] 1) It improves the static measurement problem of Euclidean distance to a certain extent and fully reflects the dynamic characteristics of time series.

[0044] 2) It has the ability to express opposite similarities between different sequence data.

[0045] 3) The pattern distance focuses more on the changing trend of the data, which matches the monitoring time series data of the bridge better. It is more suitable for the similarity characteristics of the monitoring time series data of the bridge. Compared with the Euclidean distance method, it is easier to find data periods with high similarity, which facilitates more intelligent management in actual engineering.

[0046] 4) It can quantitatively describe the trend differences between different modes of bridge monitoring time series data, more effectively measure the similarity of different sequences, and the results are more intuitive, clear and easy to operate.

[0047] 5) By reducing the standardization processing steps of measurement data, the pattern-distinguishing distance can be calculated quickly, improving calculation efficiency and effectively reflecting the similarity of bridge monitoring time series data.

[0048] 6) By distinguishing different modes, bridge monitoring time series data have more distinguishable and discriminative characteristics, effectively reflecting the similarity of bridge monitoring time series data under different modes. Attached Figure Description

[0049] Figure 1 This is a flowchart of the method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to the present invention;

[0050] Figure 2 This is a flowchart of the calculation process for the dual-judgment method of the present invention;

[0051] Figures 3 to 6 This is a comparative analysis chart of the similarity between deflection and temperature data in application example calculations;

[0052] Figures 7 to 10 This is a schematic diagram comparing the slope of the reference sequence and the comparison sequence. Detailed Implementation

[0053] The present invention will now be further described with reference to the accompanying drawings.

[0054] This invention provides a method for measuring the similarity of time-series data based on pattern differentiation, primarily applicable to the fields of bridge health monitoring data analysis (SHM system), engineering project management data analysis, and bridge risk assessment. For example... Figure 1 As shown, the method of the present invention includes the following steps:

[0055] 1) The bridge monitoring system acquires environmental data, structural response data, and bridge monitoring time series data from several time periods at different monitoring points under normal load conditions. This includes reference bridge monitoring time series data S and comparison bridge monitoring time series data S′ for similarity measurement. The obtained bridge monitoring time series data undergoes preprocessing. The main basis for measuring the similarity between different series is the similarity metric, which is commonly used with a distance function. Assuming two time series Y and Z, the distance function judgment formula is as follows:

[0056] d(y,z)≤ε

[0057] In the formula: d(y,z) is a similarity distance metric function; ε is a given similarity threshold.

[0058] If the judgment formula is satisfied, it indicates that the two sequences are similar. Based on the theory of using distance functions to measure the similarity of different sequence data, and in order to solve the static measurement problem of the commonly used Euclidean distance function and fully reflect the dynamic characteristics of time series, a similarity measurement method for pattern discrimination is proposed. First, the time series is defined as Y, and the time series is defined as follows:

[0059] Y={(y1,T1),(y2,T2),(y3,T3),...,(y i ,T i )}

[0060] In the formula: y i For the i-th time period T i The numerical values ​​of time series data within the range.

[0061] When performing similarity judgments based on patterns, complete data consistency is not required; the primary consideration is the data's trend. Therefore, differencing the data eliminates the influence of long-term trends. Furthermore, the purpose of differencing is to approximate the derivative of a discrete function; the derivative represents the rate of change, i.e., the slope k. i .

[0062] The preprocessed bridge monitoring time series data is segmented according to time periods. The trend of value change within each time period cell can be represented by the slope k of the straight line segment of that cell. i The time series is represented as follows. In this invention, the time series interval between two data points is one hour, expressed in hours, i.e., T. i+1 -T i =1.

[0063]

[0064] In the formula: k i Represented as time period T i To Ti+1 The slope of the straight line segment in the bridge monitoring time series data, i = 1, 2, ... n, where n is the number of segments in the entire bridge monitoring time series data.

[0065] At the same time, the interval sequence Y is redefined and denoted by M, as follows:

[0066] M={(ju1,u1,k1),(ju2,u2,k2),...(ju n ,u n ,k n )}

[0067] In the formula: ju i Represented as T i To T i+1 The trend judgment results of the time series, u i Represented as T i To T i+1 The similarity judgment results of time series.

[0068] 2) Determine trends in the data, such as... Figure 2 As shown, the main trend of the data sequence pattern is distinguished by whether their changing trends are consistent. This is achieved by comparing the slope k of the straight line segment within the same period of the reference bridge monitoring time series data S and the comparative bridge monitoring time series data S′. i and k′ i Perform a main trend comparison, and denote the result as ju. i The criteria for judgment are as follows.

[0069]

[0070] In the formula: ju i For the discrimination result; k i For reference, bridge monitoring time series data S in T i To T i+1 The slope of the straight line segment within the period; k′ i To compare bridge monitoring time series data S′ at T i To T i+1 The slope of the straight line segment within the period.

[0071] 3) Based on the trend judgment results, the reference bridge monitoring time series data S and the comparative bridge monitoring time series data S', which are used for similarity measurement, are classified and processed. i =1 indicates the first type of same direction, ju i =-1 represents the second type of anisotropic type.

[0072] 4) Perform a similarity assessment, such as... Figure 2As shown, based on the slopes of the straight line segments in each time period unit of the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′, the angle difference between the two straight line segments is calculated using the slopes of the two straight line segments and trigonometric function formulas, as follows:

[0073] k i =tanαk i =tanβ

[0074]

[0075]

[0076] In the formula: k i For reference, bridge monitoring time series data S in T i To T i+1 The slope of the straight line segment within the period; k i ' is for comparing time series data S' in T i To T i+1 The slope of the straight line segment within the period; α is the reference bridge monitoring time series data S in T i To T i+1 The angle between the straight line segment and the horizontal line segment within the period; β is the angle between the bridge monitoring time series data S′ and T. i To T i+1 The angle between the straight line segment and the horizontal line segment within the period; Δk i It represents the angle difference between two straight line segments.

[0077] The slope difference between the classified reference bridge monitoring time series data S and the comparison time series data S′ was compared to determine the degree of similarity. This comparison was used as the criterion for calculating the degree of similarity. The comparison results are shown in Table 1.

[0078] Table 1. Comparison of slope levels between reference and comparison sequences.

[0079]

[0080]

[0081] Considering the threshold selection method in Fragment Alignment Distance (FAD), the threshold ε can be used to determine the range of feature values ​​for the sequence. This range can be taken as the top 20% of the maximum slope difference between the training data samples, i.e., ε1 = |Δk|. max |×20%, ε1'=|Δk max |×20%; The value of ε2 is 50% of the maximum slope difference of the data samples, that is, ε2=|Δk max |×50%, ε'2=|Δk max |×50%. Based on the trend judgment results, when ju iWhen = 1 represents the first type of unidirectional type, based on actual engineering requirements, when |Δk i Line segments exceeding the threshold ε2 are not included in the similarity calculation; when ε1≤|Δk i When |≤ε2, a certain reduction is considered in the similarity judgment process; when |Δk i When |≤ε1, no reduction is considered in the similarity judgment process. When ju i When -1 represents the second type of heterogeneous case, for k i 'Through symmetric processing with k i ' symm Express it as follows: if |Δk i If the threshold ε1' is not exceeded, its negative impact on the similarity is not considered in the similarity judgment process; if ε1' ≤ |Δk i When |≤ε'2, a certain reduction is considered in the similarity judgment process; when ε'2≤|Δk i In this case, the negative impact of the similarity factor on the degree of similarity needs to be considered during the similarity assessment process.

[0082] 5) By considering the similarity of bridge monitoring time series data within each time period unit based on the two judgment results, the straight line segment pattern is distinguished by distance d according to the magnitude of the slope difference. si The model is divided into two levels, and the pattern distinguishing distance ds is calculated by the following formula. i

[0083] ds i =ju i ×u i

[0084] In the formula: ds i To distinguish patterns, the similarity value between the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′ is used; i The result is based on trend judgment; u i The result is the similarity assessment.

[0085] The pattern differentiation similarity comparison results between the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′ are represented as a similarity metric judgment dataset U, as shown in the following formula:

[0086] U = {x | -1 ≤ x ≤ 1}

[0087] Where x is d siThe value of is [0,1]. When the value is closer to 1, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar. When the value is [-1,0], the value is closer to 0, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar.

[0088] 6) The similarity between the reference bridge monitoring time series data S and the comparative bridge monitoring time series data S′ is measured using the pattern discrimination distance method, taking into account the correlation between different time periods. The high-dimensional data similarity metric function Hs(S,S′) is used to calculate the similarity based on the trend pattern between the data, as shown in the following formula:

[0089]

[0090] In the formula: Hs(S,S') is a high-dimensional data similarity measurement function; d si Distinguish the distance for the pattern; Δk i The difference in angle between two line segments is denoted as Hs. High-dimensional similarity measurement methods calculate the degree of similarity based on the trend and shape of the data. The similarity measurement value ranges from [0,1]. The closer the similarity is, the closer the similarity measurement value Hs is to 1, i.e., Hs≈1; conversely, the more the similarity is different, the closer the similarity measurement value Hs is to 0, i.e., Hs≈0. This is used to assess the degree of similarity between high-dimensional data.

[0091] When the high-dimensional data similarity value Hs≥0.8, the two sets of data are considered to have high similarity and are considered to be consistent data; conversely, when Hs≤0.2, the two sets of data are considered to have almost no similarity.

[0092] Application Examples

[0093] The measured deflection and temperature data of a small-to-medium-sized bridge in a certain city were used as the experimental data for this study. Since the monitoring data collected during normal bridge operation mainly reflects the combined response of temperature, vehicles, and wind, with the vehicle effect exhibiting a stable and predictable response over a stable time period, and wind load having a relatively small impact on concrete beam bridges, the bridge monitoring data was divided into multiple segments within a 24-hour period (including peak commuting hours) using fixed windows, based on the design concept that similar time series exhibit similar trends. The measured values ​​from each period were then connected by straight lines.

[0094] Step 1: By monitoring the bridge data, obtain the bridge monitoring time series data under normal load conditions of the bridge monitoring system, and record and statistically analyze it.

[0095] Taking one day's 24 hours as the time node for the bridge, four sets of data under normal load conditions were obtained from the bridge monitoring system, using the bridge monitoring system's recorded deflection and temperature data as an example. The details are shown in Table 2.

[0096] Table 2: Actual Measurement Data of Bridge Deflection and Temperature

[0097]

[0098]

[0099] Step 2: Preprocess the bridge monitoring time series data obtained above, and calculate the slope k within each time period unit of the preprocessed bridge monitoring time series data. i The calculations were performed, and the main trend comparisons were made based on the slopes to determine the consistency of the changing trends, as shown in Table 3. Through trend judgment, the obtained bridge monitoring time series data were classified and processed to obtain two types of bridge monitoring time series data sample sets: unidirectional and heterodirectional. The results were recorded with reference to Table 1, which shows the comparison of the slope of the reference and comparison sequences.

[0100] Step 3: Perform similarity assessment. Calculate the angle difference between the two line segments of the reference bridge monitoring time series data S and the comparison bridge monitoring time series data S′, and compare it with the maximum angle difference of the data samples to perform a similarity comparison analysis.

[0101] Step 4: Based on the trend judgment and similarity calculation results, assess the similarity of bridge monitoring time series data within each time period unit, and record the results as d. si The dataset U is formed, and the calculation results are shown in Table 3:

[0102] Table 3: Data Processing of Bridge Deflection and Temperature

[0103]

[0104]

[0105]

[0106] Step 5: The similarity of bridge monitoring time series data is measured using the pattern discrimination distance method. The similarity is then measured using the high-dimensional data similarity metric function Hs(S,S′), and compared with the similarity results calculated using the Euclidean distance formula. The calculation results are shown in Table 4. Figure 3 , Figure 4 , Figure 5 , Figure 6The similarity comparisons of the first, second, third, and fourth groups are presented in the form of graphs, and the comparison results are shown in Table 4.

[0107] Table 4: Similarity Measurement of Bridge Monitoring Time Series Data

[0108] Group number High-dimensional similarity measure Euclidean distance First group 0.803365 0.148602 Second group 0.829675 0.103712 Third group 0.290797 0.259924 Fourth group 0.061657 0.207054

[0109] In summary, to verify the advantages of this pattern-discrimination distance-based similarity identification method, additional example data and the commonly used Euclidean distance similarity testing method were added. The first and second groups used both the pattern-discrimination-based bridge monitoring time series data similarity measurement method and the Euclidean distance formula to calculate the same data similarity results. The two methods yielded consistent data similarity comparisons, demonstrating the reliability of the pattern-discrimination-based bridge monitoring time series data similarity measurement method. In the third and fourth groups, the Euclidean distance method was used to detect time periods considered highly similar. However, analysis of the graphs revealed that while the data were relatively similar, their trends were not similar. The pattern-discrimination-based bridge monitoring time series data similarity measurement method fully reflects the dynamic characteristics of time series data. This method provides a more efficient approach and process for measuring the similarity of bridge monitoring time series data.

[0110] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for measuring the similarity of bridge monitoring time series data based on pattern discrimination, characterized in that, Includes the following steps: (1) Obtain environmental data, structural response data and bridge monitoring time series data for several time periods at different monitoring points under normal load conditions through the bridge monitoring system, including reference bridge monitoring time series data for similarity measurement. Comparison of bridge monitoring time series data The obtained bridge monitoring time series data is preprocessed, and the preprocessed bridge monitoring time series data is segmented according to time period, and the slope of the straight line segment in each time period unit is calculated. (2) Make trend judgments by referring to the bridge monitoring time series data. Comparison of bridge monitoring time series data Whether the signs of the slopes in each segment are the same, each segment of bridge monitoring time series data is divided into two categories: unidirectional and heterodirectional. (3) Make a similarity judgment and calculate the reference bridge monitoring time series data. Comparison of bridge monitoring time series data The angle difference between two line segments is compared with the maximum angle difference in the data sample to perform a similarity comparison analysis. (4) Based on the trend judgment and the similarity judgment results, the similarity of bridge monitoring time series data in each time period unit is considered, and the product of the two results is used as the pattern discrimination distance; (5) Measure the similarity of bridge monitoring time series data according to the high-dimensional data similarity metric function based on pattern distinguishing distance; the high-dimensional data similarity metric function based on pattern distinguishing distance is denoted as... As shown in the following formula: In the formula: For high-dimensional data similarity measurement functions; Distinguish distances for patterns; This represents the angular difference between two line segments; high-dimensional similarity measurement methods calculate the degree of similarity based on the trend and pattern of the data, and the range of similarity measurement values ​​is... The closer the similarity, the higher the similarity metric value. The closer it gets to 1, and Conversely, the greater the deviation in similarity, the lower the similarity measure value. The closer it gets to 0, and This is used to assess the degree of similarity between high-dimensional data.

2. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 1, characterized in that, Bridge monitoring time series data under normal load conditions is obtained through a bridge monitoring system, and the bridge monitoring time series data is defined as follows: As shown in the following formula: In the formula: For the first Time period Numerical values ​​of bridge monitoring time series data within the region.

3. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 2, characterized in that, Step (1) involves preprocessing the bridge monitoring time series data, including: performing dimensionless processing on the data and differential operations to eliminate the influence of long-term trends, and dividing the i-th time period unit... The trend of value change within a given time period is represented by the slope of the straight line segment within that time period. Indicates the slope As shown in the following formula, the time series data points are one hour apart, in hours. ; In the formula: Time period to The slope of the straight line segment in the bridge monitoring time series data; , This refers to the number of segments in the entire bridge monitoring time series data.

4. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 3, characterized in that, Bridge monitoring time series data Redefine the representation to It can be expressed as follows: In the formula: Represented as to The results of trend judgment for the time series. Represented as to The similarity judgment results of time series.

5. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 1, characterized in that, The trend judgment in step (2) specifically involves: calculating the time series data of the selected reference bridge monitoring for similarity measurement. Comparison of bridge monitoring time series data The slope of the straight line segment within each time period unit, and the slope of the straight line segment within the i-th time period unit are respectively denoted as... and And perform a main trend comparison to determine the consistency of the changing trends. The trend judgment result is recorded as follows. The judgment criteria are as follows: In the formula: For the judgment result; For reference, bridge monitoring time series data exist to The slope of the straight line segment within the period; To compare time series data exist to The slope of the straight line segment within the period; By using reference bridge monitoring time series data for similarity measurement Comparison of bridge monitoring time series data Classify and process them. It is the first type of unidirectional type. It belongs to the second type of anisotropic type.

6. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 5, characterized in that, The similarity determination in step (3) specifically involves: calculating the selected reference bridge monitoring time series data for similarity measurement. Comparison of bridge monitoring time series data The slope of the line segment within each time period is calculated using the slopes of the two line segments and trigonometric function formulas to determine the angle difference between them, as shown in the following formula: In the formula: For reference, bridge monitoring time series data exist to The slope of the straight line segment within the period; To compare time series data exist to The slope of the straight line segment within the period; For reference, bridge monitoring time series data exist to The angle between the straight line segment and the horizontal line segment within the period; To compare bridge monitoring time series data exist to The angle between the straight line segment and the horizontal line segment within the period; It represents the angle difference between two straight line segments.

7. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 6, characterized in that, Step (4) involves considering the similarity of bridge monitoring time series data within each time period unit based on both the trend judgment result and the similarity judgment result. Specifically, based on the trend judgment, the similarity of the bridge monitoring time series data will be compared with the reference bridge monitoring time series data. Comparison of bridge monitoring time series data The slope of the straight line segment within each time period unit is compared to determine the trend. It is the first type of unidirectional type. It is the second type of anisotropic type; By judging the results based on trends and determining the degree of similarity, when... In the case of the first type of same direction, when Exceeding the maximum angle difference of the data samples If the similarity score is 50%, it will not be included in the calculation process, and the similarity score will be taken. ;when The maximum angular difference in the data samples 20% of the maximum angle difference from the data sample When the similarity is between 50% and 50%, a certain reduction is considered in the similarity judgment process, and the similarity judgment value is taken. ;when The maximum angle difference in the data sample did not exceed the maximum angle difference. When the similarity score is 20%, no reduction is considered in the similarity judgment process, and the similarity judgment value is taken. ; when When it is the second type of opposite direction, for By symmetry processing, if The maximum angle difference in the data sample did not exceed the maximum angle difference. If the similarity score is 20%, then its negative impact on the similarity score will not be considered during the similarity assessment process, and the similarity score will be taken. ;when The maximum angular difference in the data samples 20% of the maximum angle difference from the data sample If the similarity is between 50% and 50%, then a certain reduction should be considered in the similarity judgment process, and the similarity judgment value should be taken. ;when Exceeding the maximum angle difference of the data samples If the similarity score is 50%, then its negative impact on the overall similarity score needs to be considered during the similarity assessment process, and a suitable similarity score should be used. ; By considering the similarity of bridge monitoring time series data within each time period unit based on the two judgment results, the results are denoted as follows: As shown in the following formula: In the formula: Distinguish distances for patterns; This is based on the result of trend analysis; The result of the similarity judgment; 8. The method for measuring the similarity of bridge monitoring time series data based on pattern differentiation according to claim 7, characterized in that, Distinguish distance between patterns A similarity comparison analysis was performed to obtain a similarity measurement judgment dataset. : in for The value of is within the range of . When the value is closer to 1, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar. When the value range is... When the value is closer to 0, it indicates that the similarity of the bridge monitoring time series data within each time period unit is more similar.

Images