CN118035772B - Intelligent analysis method for civil engineering detection data based on machine learning - Google Patents
Intelligent analysis method for civil engineering detection data based on machine learning Download PDFInfo
- Publication number
- CN118035772B CN118035772B CN202410430485.1A CN202410430485A CN118035772B CN 118035772 B CN118035772 B CN 118035772B CN 202410430485 A CN202410430485 A CN 202410430485A CN 118035772 B CN118035772 B CN 118035772B
- Authority
- CN
- China
- Prior art keywords
- wave band
- data
- detected
- band
- wave
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 86
- 238000001514 detection method Methods 0.000 title claims abstract description 20
- 238000010801 machine learning Methods 0.000 title claims abstract description 18
- 230000002159 abnormal effect Effects 0.000 claims abstract description 52
- 230000005856 abnormality Effects 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 13
- 238000010606 normalization Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000007405 data analysis Methods 0.000 description 9
- 230000008859 change Effects 0.000 description 5
- 230000001174 ascending effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012876 topography Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention relates to the technical field of engineering data supervision, in particular to an intelligent analysis method of civil engineering detection data based on machine learning. The method comprises the steps of obtaining a data wave band of an engineering structure; determining initial abnormality degree according to the vibration amplitude, the number of data points, the difference of wave band duration and the second value difference of the wave band to be detected and the adjacent wave bands; determining the time sequence similarity according to the time sequence fluctuation characteristics between the to-be-detected wave band and the adjacent wave band, and determining the importance degree of the to-be-detected wave band according to the time sequence similarity and the initial abnormality degree; and determining a clustering characteristic value by combining the importance degree and the vibration amplitude, carrying out hierarchical clustering on all the data wave bands to obtain a cluster, and carrying out characteristic analysis on all the data wave bands in the cluster to obtain an analysis result. The invention classifies through the abnormal fluctuation and the actual vibration amplitude, can greatly reduce the error caused by the abnormal influence, and improves the reliability of the analysis result obtained by the feature analysis.
Description
Technical Field
The invention relates to the technical field of engineering data supervision, in particular to an intelligent analysis method of civil engineering detection data based on machine learning.
Background
Civil engineering generally has continuous vibration characteristics, monitors and analyzes vibration data of the civil engineering, can help engineers ensure the safety of the structure of the civil engineering, timely discovers potential problems possibly existing, timely maintains the civil engineering, improves the safety of the engineering structure, and avoids traffic accidents caused by potential safety hazards of the engineering structure.
In the related art, data analysis is directly performed according to the vibration amplitude of vibration corresponding to civil engineering, in this way, due to the fact that the structural difference of the civil engineering is large, certain vibration errors can be inevitably generated in the process of collecting and transmitting vibration data corresponding to the civil engineering, the vibration amplitude can be affected by the vibration errors, and accordingly the data analysis errors are large and the reliability is insufficient.
Disclosure of Invention
In order to solve the technical problems of large error and insufficient reliability of data analysis caused by directly carrying out data analysis according to vibration amplitude of vibration corresponding to civil engineering in the related art, the invention provides an intelligent analysis method of civil engineering detection data based on machine learning, which adopts the following technical scheme:
The invention provides an intelligent analysis method of civil engineering detection data based on machine learning, which comprises the following steps:
Obtaining vibration data of an engineering structure in a preset time period, and dividing the vibration data according to a first extreme value of a vibration amplitude in the vibration data to obtain a data wave band;
Taking any data wave band as a wave band to be measured, taking other data wave bands in a preset neighborhood range with the wave band to be measured as a center as adjacent wave bands, and determining abnormal indexes of the wave band to be measured according to the difference of vibration amplitudes, data point numbers and wave band duration of all data points in the wave band to be measured and all adjacent wave bands; determining the initial abnormality degree of the to-be-detected wave band according to the second-level difference of the vibration amplitude values in the to-be-detected wave band and all the adjacent wave bands and the abnormality index of the to-be-detected wave band, and updating the to-be-detected wave band to obtain the initial abnormality degree of each data wave band;
Determining the time sequence similarity of the to-be-detected wave band and each adjacent wave band according to the time sequence fluctuation characteristics between the to-be-detected wave band and the adjacent wave bands, and determining the importance degree of the to-be-detected wave band according to the time sequence similarity between the to-be-detected wave band and all the adjacent wave bands and the initial abnormality degree of all the adjacent wave bands;
And determining a clustering characteristic value of each data wave band according to the importance degree and the vibration amplitude of each data wave band, carrying out hierarchical clustering on all the data wave bands according to the clustering characteristic values of all the data wave bands to obtain clustered clusters, and carrying out characteristic analysis on all the data wave bands in different clustered clusters to obtain an analysis result.
Further, determining the abnormal index of the to-be-measured wave band according to the difference of the vibration amplitude, the number of data points and the wave band duration of all data points in the to-be-measured wave band and all adjacent wave bands, including:
Calculating the absolute value of the difference between the average value of the vibration amplitudes of all the data points in the to-be-measured wave band and the average value of the vibration amplitudes of all the data points in each adjacent wave band to obtain the amplitude difference between the to-be-measured wave band and each adjacent wave band, calculating the average value of the amplitude difference between the to-be-measured wave band and all the adjacent wave bands, and carrying out normalization treatment to obtain the amplitude influence coefficient of the to-be-measured wave band;
Calculating the average value of the number of the data points of the to-be-detected wave band and all adjacent wave bands to obtain the average value of the data points, and taking the absolute value of the difference value between the number of the data points of the to-be-detected wave band and the average value of the data points as a quantity influence coefficient by normalization processing;
determining a duration influence coefficient of the to-be-detected wave band according to the duration distribution of the to-be-detected wave band and all adjacent wave bands;
and calculating a normalized value of the product of the amplitude influence coefficient, the quantity influence coefficient and the duration influence coefficient to obtain an abnormal index of the wave band to be detected.
Further, the determining the duration influence coefficient of the to-be-measured band according to the duration distribution of the to-be-measured band and all adjacent bands includes:
Calculating variances of duration time of the to-be-measured wave band and all adjacent wave bands as first time variances;
calculating the variance of the duration time of all adjacent wave bands as a second time variance;
and normalizing the absolute value of the difference value of the first time variance and the second time variance to obtain a duration influence coefficient.
Further, the second extreme value is a maximum value, and determining the initial abnormality degree of the to-be-measured wave band according to the second extreme value difference of the vibration amplitude values in the to-be-measured wave band and all adjacent wave bands and the abnormality index of the to-be-measured wave band includes:
Calculating the average value of the absolute value of the difference between the maximum value of the vibration amplitude in each adjacent wave band and the maximum value of the vibration amplitude in each wave band to be detected, and obtaining the maximum value difference average value coefficient;
and calculating the product of the maximum difference mean coefficient and the abnormal index of the wave band to be measured, and normalizing to obtain the initial abnormal degree of the wave band to be measured.
Further, the determining the time sequence similarity between the to-be-measured wave band and each adjacent wave band according to the time sequence fluctuation characteristics between the to-be-measured wave band and the adjacent wave bands includes:
And calculating dtw values of vibration data of the to-be-detected wave band and each adjacent wave band based on a dynamic time warping algorithm, and mapping and normalizing the dtw values in a negative correlation manner to obtain the time sequence similarity of the to-be-detected wave band and each adjacent wave band.
Further, the determining the importance degree of the to-be-measured band according to the time sequence similarity between the to-be-measured band and all adjacent bands and the initial abnormality degree of all adjacent bands includes:
optionally selecting an adjacent wave band as an analysis wave band, and taking the difference value of the time sequence similarity between the wave band to be detected and the analysis wave band and the initial abnormality degree of the analysis wave band as the time sequence important coefficient between the wave band to be detected and the analysis wave band;
And calculating the average value of the time sequence importance coefficients of the to-be-measured wave band and all adjacent wave bands, and carrying out normalization processing to obtain the importance degree of the to-be-measured wave band.
Further, the determining the clustering feature value of each data band according to the importance degree and the vibration amplitude of each data band includes:
calculating the average value of vibration amplitudes of all data points in a data wave band to obtain a vibration coefficient;
And taking the product of the importance degree and the vibration coefficient of the same data wave band as the clustering characteristic value of the corresponding data wave band.
Further, the hierarchical clustering is performed on all the data bands according to the clustering feature values of all the data bands to obtain a cluster, including:
hierarchical clustering is carried out on the clustering characteristic values of all the data wave bands based on a hierarchical clustering algorithm, and all the data wave bands are clustered into at least two clustering clusters.
Further, the performing feature analysis on all the data bands in different clusters to obtain analysis results includes:
And carrying out feature extraction on all data wave bands in different clusters based on a pre-trained feature analysis model to obtain feature data, and taking the feature data corresponding to all the clusters respectively as an analysis result.
Further, the first extremum is a minimum value, the dividing the vibration data according to the first extremum of the vibration amplitude in the vibration data to obtain a data band includes:
and dividing the vibration data between two minima in the vibration data into a data wave band.
The invention has the following beneficial effects:
According to the method, the data analysis is directly carried out according to the vibration amplitude in the related technology, so that the problem of larger error and insufficient reliability of the data analysis is solved, the data wave bands corresponding to the vibration data of the engineering structure are obtained, the abnormal analysis is carried out on the wave bands to be detected according to the differences of the vibration amplitude, the number of data points, the corresponding wave band duration and the like of adjacent wave bands and the central wave band to be detected in a local range, the initial abnormal degree of the wave bands to be detected is determined by combining the extreme value difference of the vibration amplitude, and as the influence of the vibration of the civil engineering on the adjacent wave bands in time sequence is larger, namely, the corresponding abnormal condition is represented when the data wave bands are suddenly changed, the abnormal condition of each data wave band is represented by the initial abnormal degree. And then, determining the time sequence similarity by combining the time sequence fluctuation characteristics between the to-be-detected wave band and the adjacent wave band, wherein the time sequence similarity is higher in waveform periodicity and also higher in time sequence similarity under the normal fluctuation condition of civil engineering. The method and the device for clustering the data bands in the frequency domain determine the importance degree of the bands to be detected by combining the time sequence similarity and the initial abnormality degree, determine the clustering characteristic value based on the importance degree and the vibration amplitude, realize clustering processing and obtain the clustering clusters, and effectively combine the waveform characteristics of each data band in a local range so as to classify according to the abnormality condition and the actual vibration amplitude, enable the analysis result in the subsequent feature analysis to greatly reduce errors caused by abnormal influence, and improve the reliability of the analysis result obtained by the feature analysis.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an intelligent analysis method for civil engineering detection data based on machine learning according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of the intelligent analysis method for civil engineering detection data based on machine learning according to the invention with reference to the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the intelligent analysis method of civil engineering detection data based on machine learning.
Referring to fig. 1, a flowchart of an intelligent analysis method for civil engineering detection data based on machine learning according to an embodiment of the present invention is shown, where the method includes:
s101: and obtaining vibration data of the engineering structure in a preset time period, and dividing the vibration data according to a first extreme value of the vibration amplitude in the vibration data to obtain a data wave band.
The engineering structure in the embodiment of the invention can be specifically a bridge structure, a tunnel structure and the like, and the bridge engineering has certain self-vibration characteristics, the tunnel engineering can generate certain vibration effect due to the running of vehicles, the vibration of the bridge engineering and the tunnel engineering can influence the safety of the engineering structure, the dangers such as cracks and collapse and the like are easy to occur in the engineering structure, the vibration data can be used for monitoring the health condition of the civil engineering structure due to the influence of different terrains, and the damage and deformation of the structure can be found in time through the monitoring and analysis of the vibration data generated by the structure, so that the data analysis is needed.
The concrete implementation scene of the invention is that vibration data in bridge engineering and tunnel engineering structures are analyzed, so that the vibration characteristics of the engineering structures are determined, and related staff can conveniently realize operations such as vibration prediction, abnormality detection and the like based on the vibration characteristics.
In the embodiment of the invention, continuous vibration data of 1 minute can be acquired, and the vibration data can be acquired every 1 minute in the preset time period or according to actual detection requirements, and the vibration data in the embodiment of the invention is waveform data, and vibration data of an engineering structure in 1 minute can be directly acquired by using a vibration sensor or can be acquired from other data sources, which is not limited.
After the vibration data is acquired, the data is in a waveform, and different wave bands have the characteristics of corresponding wave crests, wave troughs and the like, so that the vibration data can be divided for the convenience of analysis.
Further, in some embodiments of the present invention, dividing the vibration data according to a first extremum of a vibration amplitude in the vibration data to obtain a data band includes: vibration data between two minima in the vibration data is divided into a data band.
In the embodiment of the invention, the first extreme value is taken as the minimum value, the second extreme value is taken as the maximum value, the minimum value and the maximum value respectively correspond to turning points of waveforms, the waveforms are converted from a descending trend to an ascending trend at the moment corresponding to the minimum value, and the waveforms are converted from the ascending trend to the descending trend at the moment corresponding to the maximum value. That is, in the embodiment of the present invention, the vibration data between two minima is used as one data band, and the transformation trend of each data band is that the whole is ascending and then descending.
S102: taking any data wave band as a wave band to be measured, taking other data wave bands in a preset neighborhood range with the wave band to be measured as a center as adjacent wave bands, and determining abnormal indexes of the wave band to be measured according to the difference of vibration amplitudes, data point numbers and wave band duration of all data points in the wave band to be measured and all adjacent wave bands; according to the second-level difference of the vibration amplitude values in the to-be-measured wave band and all the adjacent wave bands and the abnormal index of the to-be-measured wave band, determining the initial abnormal degree of the to-be-measured wave band, and updating the to-be-measured wave band to obtain the initial abnormal degree of each data wave band.
In the embodiment of the invention, one data wave band can be selected as a wave band to be detected, other data wave bands in a preset neighborhood range with the wave band to be detected as the center are adjacent wave bands, and then, according to the fluctuation characteristics between the wave band to be detected and the adjacent wave bands in the adjacent time sequence local range, the abnormal condition of the wave band to be detected is analyzed.
It should be noted that, the engineering structure generally has a certain waveform abnormal phenomenon, for example, abnormal changes of weather and topography, or abnormal vibration sources, or noise influence in the transmission process, which can affect the analysis of the waveform, so that the abnormal condition of each data band needs to be determined.
Further, in some embodiments of the present invention, determining an anomaly index for a band to be measured based on a difference in vibration amplitude, number of data points, and band duration between the band to be measured and all data points in all adjacent bands includes: calculating the absolute value of the difference between the average value of the vibration amplitudes of all the data points in the to-be-measured wave band and the average value of the vibration amplitudes of all the data points in each adjacent wave band to obtain the amplitude difference between the to-be-measured wave band and each adjacent wave band, calculating the average value of the amplitude difference between the to-be-measured wave band and all the adjacent wave bands, and carrying out normalization treatment to obtain the amplitude influence coefficient of the to-be-measured wave band; calculating the average value of the number of data points of the to-be-measured wave band and all adjacent wave bands, obtaining the average value of the data points, and taking the absolute value of the difference value between the number of the data points of the to-be-measured wave band and the average value of the data points as a quantity influence coefficient by normalization processing; according to the duration distribution of the wave band to be measured and all adjacent wave bands, determining the duration influence coefficient of the wave band to be measured; and calculating a normalized value of the product of the amplitude influence coefficient, the quantity influence coefficient and the duration influence coefficient to obtain an abnormal index of the wave band to be detected.
In the embodiment of the invention, the average value of the vibration amplitudes of all the data points in the to-be-detected wave bands can be calculated as the to-be-detected average value, then the average value of the vibration amplitudes of all the data points in each adjacent wave band is respectively calculated as the adjacent average value, and the absolute value of the difference between the to-be-detected average value and the adjacent average value is used as the amplitude difference between the to-be-detected wave band and the corresponding adjacent wave band. It can be understood that the average value of the amplitude values of all the data points in each data band represents the amplitude characteristics in the data band, the distance between the adjacent bands and the band to be measured is closer, and the more abnormal the corresponding band to be measured can be represented when the amplitude characteristics change greatly in the similar data bands, therefore, the amplitude influence coefficient of the band to be measured is obtained through calculating the average value of the amplitude differences between the band to be measured and all the adjacent bands by normalization processing, and the larger the amplitude influence coefficient is, the more abnormal the amplitude characteristics of the band to be measured are.
The method comprises the steps of calculating the absolute value of the difference between the number of data points of a to-be-measured wave band and the mean value of the data points, wherein the sampling interval of vibration data is unchanged, the number of the data points in adjacent data wave bands is smaller under normal conditions, and when the number of the data points is larger, the waveform change possibly caused by abnormal vibration is caused, so that the normalization processing is used as a quantity influence coefficient, and the larger the quantity influence coefficient is, the larger the number difference between the to-be-measured wave band and the data points of all adjacent wave bands is, namely the more the fluctuation condition of the to-be-measured wave band is abnormal.
Further, in some embodiments of the present invention, determining a duration influence coefficient of a band to be measured according to a distribution of durations of the band to be measured and all adjacent bands includes: calculating the variance of the duration time of the wave band to be measured and all adjacent wave bands as a first time variance; calculating the variance of the duration time of all adjacent wave bands as a second time variance; and normalizing the absolute value of the difference between the first time variance and the second time variance to obtain a duration influence coefficient.
The variance of the duration represents the distribution discrete feature of the duration, the waveform duration complexity in the total time period corresponding to all adjacent wave bands after the wave bands to be detected are not present is the second time variance, the waveform duration complexity of the wave bands to be detected is the first time variance, and the larger the difference between the first time variance and the second time variance is, the larger the influence of the wave bands to be detected in the corresponding total time period is, namely, the more the feature of the wave bands to be detected is prominent, the more abnormal the wave bands to be detected are.
In summary, the method combines the amplitude influence coefficient, the quantity influence coefficient and the duration influence coefficient to determine the abnormality degree of the wave band to be detected, calculates the normalized value of the product of the amplitude influence coefficient, the quantity influence coefficient and the duration influence coefficient, and obtains the abnormality index of the wave band to be detected.
The abnormal index is index information obtained by combining waveform characteristics, data point characteristics and duration characteristics corresponding to adjacent wave bands and wave bands to be detected, and the larger the numerical value of the abnormal index is, the larger the difference between the wave bands to be detected and all the adjacent wave bands is, and the more the abnormal characteristics of the wave bands to be detected are highlighted.
Further, in some embodiments of the present invention, determining an initial abnormality degree of the to-be-measured band according to a second polar value difference of vibration amplitudes in the to-be-measured band and all adjacent bands and an abnormality index of the to-be-measured band includes: calculating the average value of the absolute value of the difference between the maximum value of the vibration amplitude in each adjacent wave band and the maximum value of the vibration amplitude in each wave band to be detected, and obtaining the maximum value difference average value coefficient; and calculating the product of the maximum difference mean coefficient and the abnormal index of the wave band to be measured, and normalizing the product to serve as the initial abnormal degree of the wave band to be measured.
Since the second extreme value is the maximum value and each data wave band is a wave band between two minimum values, each data wave band has and only has one maximum value, in the normal engineering vibration process, the waveform change is a continuous process, the maximum value change of the adjacent data wave bands is smaller, and when the waveform is abnormal or has abnormal influence, the maximum value can be caused to generate mutation, therefore, the maximum value difference mean value coefficient is obtained by calculating the difference of the maximum values, the larger the value of the maximum value difference mean value coefficient is, the product of the maximum value difference mean value coefficient and the abnormal index of the wave band to be measured is calculated, and the normalization processing is used as the initial abnormal degree of the wave band to be measured.
The initial abnormality degree of the embodiment of the invention combines the amplitude characteristic, the local distribution characteristic and the extremum change characteristic of each waveform, thereby being capable of effectively analyzing the abnormality of the wave band to be tested, avoiding the occurrence of larger errors caused by directly determining the abnormality degree according to the influence of the amplitude, and affecting the accuracy and the reliability of final abnormality analysis.
S103: according to the time sequence fluctuation characteristics between the to-be-measured wave band and the adjacent wave bands, determining the time sequence similarity between the to-be-measured wave band and each adjacent wave band, and according to the time sequence similarity between the to-be-measured wave band and all the adjacent wave bands and the initial abnormal degree of all the adjacent wave bands, determining the importance degree of the to-be-measured wave band.
Further, in some embodiments of the present invention, determining the timing similarity between the band under test and each adjacent band according to the timing fluctuation characteristics between the band under test and the adjacent bands includes: based on a dynamic time warping algorithm, dtw values of vibration data of the to-be-detected wave band and each adjacent wave band are calculated, and dtw values are mapped and normalized in a negative correlation mode to obtain the time sequence similarity of the to-be-detected wave band and each adjacent wave band.
It can be understood that the time sequence features represent the waveform difference condition of two data wave bands on time sequence, the waveform can be analyzed by using a dynamic time warping algorithm, and the larger the dtw value is, the larger the difference between the wave band to be measured and vibration data corresponding to adjacent wave bands is, so that negative correlation mapping and normalization are needed, and the time sequence similarity between the wave band to be measured and each adjacent wave band is obtained. The negative correlation map may specifically be, for example, a negative number of dtw values and then normalized, or may also be calculated and normalized as the inverse of dtw values, without limitation.
Further, in some embodiments of the present invention, determining the importance level of the to-be-measured band according to the time sequence similarity between the to-be-measured band and all adjacent bands and the initial abnormality level of all adjacent bands includes: optionally selecting an adjacent wave band as an analysis wave band, and taking the difference value of the time sequence similarity between the wave band to be detected and the analysis wave band and the initial abnormality degree of the analysis wave band as the time sequence important coefficient between the wave band to be detected and the analysis wave band; and calculating the average value of the time sequence importance coefficients of the to-be-measured wave band and all adjacent wave bands, and carrying out normalization processing to obtain the importance degree of the to-be-measured wave band.
In the embodiment of the invention, as the noise data only affects the vibration data at the current moment, the corresponding normal wave band to be detected can avoid the influence of abnormality on the whole analysis, and the importance is higher.
In the embodiment of the invention, the higher the time sequence similarity is, the higher the similarity between an analysis wave band and a wave band to be detected is, and the higher the initial abnormality degree of the analysis wave band is, the higher the abnormality degree of data of the analysis wave band is, and the higher the time sequence importance coefficient of the wave band to be detected and the analysis wave band is used for representing the normal degree of the wave band to be detected, the more the time sequence importance coefficient of the wave band to be detected is similar to the analysis wave band, and the higher the importance degree of the wave band to be detected is, so that the difference value between the time sequence similarity between the wave band to be detected and the analysis wave band and the initial abnormality degree of the analysis wave band is calculated as the time sequence importance coefficient of the wave band to be detected and the analysis wave band. And updating the analysis wave band, calculating the average value of time sequence importance coefficients of the wave band to be detected and all adjacent wave bands, and normalizing to obtain the importance degree of the wave band to be detected.
S104: and determining a clustering characteristic value of each data wave band according to the importance degree and the vibration amplitude of each data wave band, carrying out hierarchical clustering on all the data wave bands according to the clustering characteristic values of all the data wave bands to obtain clustered clusters, and carrying out characteristic analysis on all the data wave bands in different clustered clusters to obtain an analysis result.
Further, in some embodiments of the present invention, determining the cluster feature value of each data band according to the importance level and the vibration amplitude of each data band includes: calculating the average value of vibration amplitudes of all data points in a data wave band to obtain a vibration coefficient; and taking the product of the importance degree and the vibration coefficient of the same data wave band as the clustering characteristic value of the corresponding data wave band.
In the embodiment of the invention, the mean value of the vibration amplitude in each data wave band is used as the vibration coefficient, and the clustering analysis can be realized based on the vibration amplitude.
In the embodiment of the invention, the product of the importance degree and the vibration coefficient of the same data wave band is used as the clustering characteristic value of the corresponding data wave band, and the clustering characteristic value is the amplitude characteristic coefficient obtained by the abnormal analysis, so that the abnormal amplitude characteristic is clustered into one type due to the lower importance degree, and the normal amplitude characteristic is clustered into one type due to the higher importance degree, thereby being convenient for the subsequent amplitude analysis.
Further, in some embodiments of the present invention, hierarchical clustering is performed on all data bands according to the cluster feature values of all data bands to obtain a cluster, including: hierarchical clustering is carried out on the clustering characteristic values of all the data wave bands based on a hierarchical clustering algorithm, and all the data wave bands are clustered into at least two clustering clusters.
The hierarchical clustering algorithm is a clustering algorithm well known in the art, and can effectively cluster the clustering characteristic values of all data wave bands through hierarchical clustering, in the embodiment of the invention, the hierarchical clustering can be divided into two clustering clusters, one is a normal clustering cluster with analysis value, the clustering cluster is higher in importance degree, the vibration amplitude is large, the vibration characteristic is obvious, the other is a clustering cluster with lower analysis value, the clustering cluster is lower in importance degree, the vibration amplitude is small, the vibration characteristic is not obvious, and of course, the number of the clustering clusters can be adjusted according to actual clustering requirements, and the method is not limited in this respect.
After the clustering is carried out to obtain the clustering cluster, the embodiment of the invention can realize the feature analysis according to the data wave band contained in the clustering cluster. Further, in some embodiments of the present invention, feature analysis is performed on all data bands in different clusters, to obtain analysis results, including: and carrying out feature extraction on all data wave bands in different clusters based on a pre-trained feature analysis model to obtain feature data, and taking the feature data corresponding to all the clusters as an analysis result.
The feature analysis model may specifically be, for example, an SVM model, which is a supervised learning model, and performs feature analysis on each cluster through training and learning, so as to obtain corresponding feature data, and performs overall analysis on abnormal conditions, normal fluctuations and the like in civil engineering by analyzing the feature data in each cluster.
It should be noted that, the characteristic data in the embodiment of the invention can be used as auxiliary reference data when vibration analysis is performed in civil engineering, and the characteristic data can be analyzed by combining with various factors such as actual topography, weather and the like where the civil engineering is located.
According to the method, the data analysis is directly carried out according to the vibration amplitude in the related technology, so that the problem of larger error and insufficient reliability of the data analysis is solved, the data wave bands corresponding to the vibration data of the engineering structure are obtained, the abnormal analysis is carried out on the wave bands to be detected according to the differences of the vibration amplitude, the number of data points, the corresponding wave band duration and the like of adjacent wave bands and the central wave band to be detected in a local range, the initial abnormal degree of the wave bands to be detected is determined by combining the extreme value difference of the vibration amplitude, and as the influence of the vibration of the civil engineering on the adjacent wave bands in time sequence is larger, namely, the corresponding abnormal condition is represented when the data wave bands are suddenly changed, the abnormal condition of each data wave band is represented by the initial abnormal degree. And then, determining the time sequence similarity by combining the time sequence fluctuation characteristics between the to-be-detected wave band and the adjacent wave band, wherein the time sequence similarity is higher in waveform periodicity and also higher in time sequence similarity under the normal fluctuation condition of civil engineering. The method and the device for clustering the data bands in the frequency domain determine the importance degree of the bands to be detected by combining the time sequence similarity and the initial abnormality degree, determine the clustering characteristic value based on the importance degree and the vibration amplitude, realize clustering processing and obtain the clustering clusters, and effectively combine the waveform characteristics of each data band in a local range so as to classify according to the abnormality condition and the actual vibration amplitude, enable the analysis result in the subsequent feature analysis to greatly reduce errors caused by abnormal influence, and improve the reliability of the analysis result obtained by the feature analysis.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (8)
1. An intelligent analysis method of civil engineering detection data based on machine learning, which is characterized by comprising the following steps:
Obtaining vibration data of an engineering structure in a preset time period, and dividing the vibration data according to a first extreme value of a vibration amplitude in the vibration data to obtain a data wave band;
Taking any data wave band as a wave band to be measured, taking other data wave bands in a preset neighborhood range with the wave band to be measured as a center as adjacent wave bands, and determining abnormal indexes of the wave band to be measured according to the difference of vibration amplitudes, data point numbers and wave band duration of all data points in the wave band to be measured and all adjacent wave bands; determining the initial abnormality degree of the to-be-detected wave band according to the second-level difference of the vibration amplitude values in the to-be-detected wave band and all the adjacent wave bands and the abnormality index of the to-be-detected wave band, and updating the to-be-detected wave band to obtain the initial abnormality degree of each data wave band;
Determining the time sequence similarity of the to-be-detected wave band and each adjacent wave band according to the time sequence fluctuation characteristics between the to-be-detected wave band and the adjacent wave bands, and determining the importance degree of the to-be-detected wave band according to the time sequence similarity between the to-be-detected wave band and all the adjacent wave bands and the initial abnormality degree of all the adjacent wave bands;
Determining a clustering characteristic value of each data wave band according to the importance degree and the vibration amplitude of each data wave band, carrying out hierarchical clustering on all the data wave bands according to the clustering characteristic values of all the data wave bands to obtain clustered clusters, and carrying out characteristic analysis on all the data wave bands in different clustered clusters to obtain analysis results;
Determining the importance degree of the to-be-detected wave band according to the time sequence similarity between the to-be-detected wave band and all adjacent wave bands and the initial abnormal degree of all adjacent wave bands, including:
optionally selecting an adjacent wave band as an analysis wave band, and taking the difference value of the time sequence similarity between the wave band to be detected and the analysis wave band and the initial abnormality degree of the analysis wave band as the time sequence important coefficient between the wave band to be detected and the analysis wave band;
Calculating the average value of time sequence importance coefficients of the to-be-detected wave band and all adjacent wave bands, and carrying out normalization processing to obtain the importance degree of the to-be-detected wave band;
the first extremum is a minimum value, the dividing the vibration data according to the first extremum of the vibration amplitude in the vibration data to obtain a data band, including:
and dividing the vibration data between two minima in the vibration data into a data wave band.
2. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 1, wherein the determining the abnormality index of the to-be-detected band according to the difference of vibration amplitude, number of data points and band duration of all data points in the to-be-detected band and all adjacent bands comprises:
Calculating the absolute value of the difference between the average value of the vibration amplitudes of all the data points in the to-be-measured wave band and the average value of the vibration amplitudes of all the data points in each adjacent wave band to obtain the amplitude difference between the to-be-measured wave band and each adjacent wave band, calculating the average value of the amplitude difference between the to-be-measured wave band and all the adjacent wave bands, and carrying out normalization treatment to obtain the amplitude influence coefficient of the to-be-measured wave band;
Calculating the average value of the number of the data points of the to-be-detected wave band and all adjacent wave bands to obtain the average value of the data points, and taking the absolute value of the difference value between the number of the data points of the to-be-detected wave band and the average value of the data points as a quantity influence coefficient by normalization processing;
determining a duration influence coefficient of the to-be-detected wave band according to the duration distribution of the to-be-detected wave band and all adjacent wave bands;
and calculating a normalized value of the product of the amplitude influence coefficient, the quantity influence coefficient and the duration influence coefficient to obtain an abnormal index of the wave band to be detected.
3. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 2, wherein the determining the duration influence coefficient of the band to be detected according to the distribution of the durations of the band to be detected and all adjacent bands includes:
Calculating variances of duration time of the to-be-measured wave band and all adjacent wave bands as first time variances;
calculating the variance of the duration time of all adjacent wave bands as a second time variance;
and normalizing the absolute value of the difference value of the first time variance and the second time variance to obtain a duration influence coefficient.
4. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 1, wherein the second extreme value is a maximum value, and the determining the initial abnormality degree of the to-be-detected band according to the second extreme value difference of the vibration amplitudes in the to-be-detected band and all adjacent bands and the abnormality index of the to-be-detected band comprises:
Calculating the average value of the absolute value of the difference between the maximum value of the vibration amplitude in each adjacent wave band and the maximum value of the vibration amplitude in each wave band to be detected, and obtaining the maximum value difference average value coefficient;
and calculating the product of the maximum difference mean coefficient and the abnormal index of the wave band to be measured, and normalizing to obtain the initial abnormal degree of the wave band to be measured.
5. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 1, wherein the determining the time sequence similarity between the to-be-detected band and each adjacent band according to the time sequence fluctuation characteristics between the to-be-detected band and the adjacent bands comprises:
And calculating dtw values of vibration data of the to-be-detected wave band and each adjacent wave band based on a dynamic time warping algorithm, and mapping and normalizing the dtw values in a negative correlation manner to obtain the time sequence similarity of the to-be-detected wave band and each adjacent wave band.
6. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 1, wherein the determining the cluster feature value of each data band according to the importance level and the vibration amplitude of each data band comprises:
calculating the average value of vibration amplitudes of all data points in a data wave band to obtain a vibration coefficient;
And taking the product of the importance degree and the vibration coefficient of the same data wave band as the clustering characteristic value of the corresponding data wave band.
7. The intelligent analysis method of civil engineering detection data based on machine learning according to claim 1, wherein the hierarchical clustering is performed on all data bands according to the cluster feature values of all data bands to obtain a cluster, and the method comprises the steps of:
hierarchical clustering is carried out on the clustering characteristic values of all the data wave bands based on a hierarchical clustering algorithm, and all the data wave bands are clustered into at least two clustering clusters.
8. The intelligent analysis method for civil engineering detection data based on machine learning according to claim 1, wherein the performing feature analysis on all data bands in different clusters to obtain analysis results includes:
And carrying out feature extraction on all data wave bands in different clusters based on a pre-trained feature analysis model to obtain feature data, and taking the feature data corresponding to all the clusters respectively as an analysis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410430485.1A CN118035772B (en) | 2024-04-11 | 2024-04-11 | Intelligent analysis method for civil engineering detection data based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410430485.1A CN118035772B (en) | 2024-04-11 | 2024-04-11 | Intelligent analysis method for civil engineering detection data based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118035772A CN118035772A (en) | 2024-05-14 |
| CN118035772B true CN118035772B (en) | 2024-06-25 |
Family
ID=90989625
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410430485.1A Active CN118035772B (en) | 2024-04-11 | 2024-04-11 | Intelligent analysis method for civil engineering detection data based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118035772B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119690023B (en) * | 2024-12-18 | 2025-07-08 | 青岛格昌环保科技有限公司 | Intelligent monitoring method for water-based paint production flow based on data analysis |
| CN119416134B (en) * | 2025-01-08 | 2025-06-17 | 曙光信息产业股份有限公司 | Job anomaly detection method and device for intelligent computing network |
| CN119813956B (en) * | 2025-03-13 | 2025-07-18 | 吉林建筑科技学院 | Abnormal state evaluation method used in solar cell operation process |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115263551A (en) * | 2022-09-27 | 2022-11-01 | 山东创乾海洋科技有限公司 | Method for monitoring running health state of marine ship engine |
| CN116821833A (en) * | 2023-08-29 | 2023-09-29 | 好林(威海)新材料有限公司 | Data anomaly detection method for machine filter cloth adhesion |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006126079A (en) * | 2004-10-29 | 2006-05-18 | Nitto Seiko Co Ltd | Sound or vibration waveform determination method and waveform abnormality detection device |
| FR2941049B1 (en) * | 2009-01-13 | 2011-02-11 | Snecma | METHOD AND SYSTEM FOR MONITORING VIBRATION PHENOMENA IN AN AIRCRAFT GAS TURBINE ENGINE IN OPERATION |
| JP5428994B2 (en) * | 2010-03-26 | 2014-02-26 | 新日鐵住金株式会社 | Anomaly diagnosis method, apparatus and program for machine equipment |
| CN112183344B (en) * | 2020-09-28 | 2021-06-01 | 广东石油化工学院 | Large unit friction fault analysis method and system based on waveform and dimensionless learning |
| CN116842348B (en) * | 2023-08-31 | 2023-12-01 | 安徽省云鹏工程项目管理有限公司 | Bridge health monitoring system based on artificial intelligence |
| CN117848488A (en) * | 2024-01-09 | 2024-04-09 | 河北新天科创新能源技术有限公司 | Online processing method for equipment vibration data |
-
2024
- 2024-04-11 CN CN202410430485.1A patent/CN118035772B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115263551A (en) * | 2022-09-27 | 2022-11-01 | 山东创乾海洋科技有限公司 | Method for monitoring running health state of marine ship engine |
| CN116821833A (en) * | 2023-08-29 | 2023-09-29 | 好林(威海)新材料有限公司 | Data anomaly detection method for machine filter cloth adhesion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118035772A (en) | 2024-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118035772B (en) | Intelligent analysis method for civil engineering detection data based on machine learning | |
| CN107610464B (en) | A Trajectory Prediction Method Based on Gaussian Mixture Time Series Model | |
| CN109124625B (en) | A classification method of driver fatigue state level | |
| CN116842348B (en) | Bridge health monitoring system based on artificial intelligence | |
| CN109754110B (en) | Early warning method and system for traction converter faults | |
| CN102247143B (en) | Integratable fast algorithm for denoising electrocardiosignal and identifying QRS waves | |
| CN111623868A (en) | Convolutional neural network construction method for rail corrugation identification | |
| CN116380445A (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| CN113947130B (en) | Training and using method and device for regression prediction AI model of polygonal wear waveform of wheel | |
| CN112668612A (en) | Partial discharge signal clustering analysis method based on grids | |
| WO2022141623A1 (en) | Train compartment vibration monitoring method, vibration signal feature library establishment method and vibration signal feature library application method | |
| CN115982606B (en) | Bridge structure health detection method and system based on fuzzy C-means clustering algorithm | |
| CN108549847B (en) | A crack damage identification method for beam structures without datum data | |
| CN118091341A (en) | Cable partial discharge detection method and system based on multi-voiceprint feature spectrogram recognition | |
| CN113836476B (en) | Pavement evenness measuring method based on smart phone | |
| CN115310499A (en) | Industrial equipment fault diagnosis system and method based on data fusion | |
| CN119577508A (en) | A partial discharge data processing method and system based on big data model | |
| CN119293536A (en) | Real-time vibration data acquisition and analysis method and related system | |
| CN109614850A (en) | The track spectrum approximating method of More algorithm based on L-M | |
| CN112200228A (en) | Epileptic seizure state identification method based on two-dimensional convolutional neural network | |
| CN116712089B (en) | Methods to enrich epileptiform discharges between epileptic seizures and predict pathogenic lesions | |
| CN117409790A (en) | A construction equipment soundprint detection method, system, equipment and medium | |
| CN115452958A (en) | Method for recognizing and early warning pipe body breakage through voiceprint based on convolutional neural network | |
| Zhou et al. | Deep learning-based segment trend removal approach for prognostics and health management signals of rail vehicles | |
| CN110889336A (en) | Quasi-static long-gauge strain extraction method based on EMD decomposition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |