Pipe-roof deformation calculation method and system based on data analysis

Abstract

The present invention relates to the technical field of solid deformation detection, and in particular to a pipe roof deformation calculation method and system based on data analysis. The present invention selects high stress mask moments according to the dramatic time series changes and fluctuations of pipe roof stress data; obtains the number of adjustment interval divisions used in the HBOS algorithm according to the distribution of the mask properties of stress data in a local neighborhood time period of the high stress mask moment, and the time series correlation between the pipe roof stress data and the pipe roof humidity data; makes the stress anomaly degree obtained after further combining the HBOS algorithm more accurate, improves the accuracy of data anomaly detection of small dynamic stress changes, and thus makes the accuracy of pipe roof deformation analysis based on the stress anomaly degree higher.

Description

Technical Field

[0001] The present invention relates to the technical field of solid deformation detection, and in particular to a pipe roof deformation calculation method and system based on data analysis. Background Art

[0002] A pipe shed is a support structure used in underground engineering or tunnel engineering. Before excavation, holes are drilled at predetermined locations and steel pipes or concrete pipes are implanted. Then concrete is filled or grouting is performed inside the pipes to form a solid support structure. Its main function is to protect the stability of the excavation surface and its surrounding areas, and to prevent the loosening or even collapse of soil and rock. Pipe shed deformation refers to the deformation of the pipe shed, such as displacement, bending or cracking, caused by external forces such as ground pressure, groundwater pressure, earthquakes, etc. during construction or use. That is, the deformation of the pipe shed is related to the external forces it is subjected to; therefore, the deformation of the pipe shed can be analyzed by the stress it is subjected to.

[0003] Considering that when the pipe roof is deformed, the stress change will produce anomalies, the stress data of the pipe roof can be detected for anomalies, so as to indirectly analyze the deformation of the pipe roof according to the degree of anomaly. The existing technology usually uses the Histogram-Based Outlier Score (HBOS) algorithm to detect anomalies in the stress data of the pipe roof, and analyzes the deformation of the pipe roof according to the degree of stress anomaly obtained by the detection. The stress data of the pipe roof will increase or decrease as the pipe roof deforms. During the deformation process of the pipe roof, the initial increase or decrease of stress is small, and the corresponding degree of anomaly is small, and usually no abnormal data can be detected; however, as the pipe roof continues to deform until the corresponding stress data is greater than the limit stress, the pipe roof will collapse or break, and other more serious deformations will occur. At this time, the corresponding stress data of the pipe roof will change over a large span in a short period of time, that is, the abnormal deformation of the pipe roof can not be detected until the pipe roof collapses.

[0004] For example, the stress data of the pipe shed will show small dynamic changes due to thermal expansion and contraction caused by humidity changes, and soil strength changes caused by humidity changes. When the pipe shed reaches the limit stress, large-scale deformation will occur, causing the stress data to jump out of the range of small dynamic changes, and a large stress change will occur in a short time, that is, the stress anomaly phenomenon; the corresponding stress data change characteristics will cause the data anomalies before the large changes to be ignored when the HBOS algorithm is used for anomaly detection, resulting in the inability to prevent anomalies in advance, that is, the existing technology uses the HBOS algorithm to detect anomalies in the stress data of the pipe shed, and the accuracy of the stress anomaly degree is low, resulting in inaccurate calculation of the pipe shed deformation. Summary of the invention

[0005] In order to solve the technical problem that the accuracy of the stress anomaly degree obtained by abnormal detection of pipe roof stress data by HBOS algorithm in the prior art is low, resulting in inaccurate calculation of pipe roof deformation, the purpose of the present invention is to provide a pipe roof deformation calculation method and system based on data analysis, and the technical scheme adopted is as follows:

[0006] The present invention proposes a pipe-roof deformation calculation method based on data analysis, the method comprising:

[0007] Obtain the pipe shed stress data and pipe shed humidity data at each sampling moment;

[0008] According to the dramatic time series changes and fluctuations of the pipe-roof stress data, the masking property of the stress data at each sampling moment is obtained; according to the masking property of the stress data at all sampling moments, the masked moments with high stress are screened out;

[0009] According to the mask distribution of stress data in the local neighborhood time period of each high stress mask moment, and the temporal correlation between the pipe shed stress data and the pipe shed humidity data, the number of adjustment interval divisions corresponding to each high stress mask moment is obtained;

[0010] According to the number of adjustment interval divisions combined with the HBOS anomaly detection method, the stress anomaly degree at each stress high mask moment is obtained; and the pipe roof deformation analysis is performed according to the stress anomaly degree.

[0011] Furthermore, the method for obtaining the masked stress data includes:

[0012] The difference between the pipe roof stress data at each sampling moment and the pipe roof stress data at the previous sampling moment is taken as the pipe roof stress change value at each sampling moment;

[0013] In terms of time sequence, all sampling moments within the preset neighborhood time window of each sampling moment are taken as the reference moment of each sampling moment; the maximum pipe-roof stress change value among all reference moments corresponding to each sampling moment is taken as the reference stress change amplitude of each sampling moment;

[0014] The difference between the mean of the pipe-roof stress change values ​​of all reference moments corresponding to each sampling moment and the pipe-roof stress change value at each sampling moment is taken as the reference stress change volatility at each sampling moment;

[0015] The stress data maskability at each sampling moment is obtained according to the reference stress variation amplitude and the reference stress variation volatility; the reference stress variation amplitude and the reference stress variation volatility are both positively correlated with the stress data maskability.

[0016] Furthermore, the method for obtaining the masking property of stress data at each sampling moment according to the reference stress change amplitude and the reference stress change volatility includes:

[0017] The normalized value of the product of the reference stress change amplitude and the reference stress change volatility is used as the stress data mask at each sampling moment.

[0018] Furthermore, the method for obtaining the stress high mask moment includes:

[0019] The sampling moment corresponding to the masked stress data greater than the preset mask threshold is taken as the high stress mask moment.

[0020] Furthermore, the method for obtaining the number of adjustment interval divisions includes:

[0021] Taking each high stress mask moment as the target moment in turn; taking all sampling moments within a preset neighborhood range of the target moment as the neighborhood moments of the target moment in time sequence;

[0022] The normalized value of the negative correlation mapping value of the time interval length between each neighborhood moment and the target moment is used as the time interval weight of each neighborhood moment; the product of the stress data masking property of each neighborhood moment and the time interval weight is used as the weighted data masking property of each neighborhood moment; the mean value of the weighted data masking properties of all neighborhood moments corresponding to the target moment is used as the local mask feature value of the target moment;

[0023] According to the distribution similarity between the pipe shed stress data and the pipe shed humidity data at all neighboring moments corresponding to the target moment, the stress-humidity change correlation at the target moment is obtained;

[0024] The product of the local mask feature value and the stress-humidity change correlation is used as the interval quantity adjustment weight at the target time;

[0025] The preset initial division number in the HBOS anomaly detection method is obtained; the preset initial division number is weighted by the interval number adjustment weight to obtain the adjusted interval division number at the target moment.

[0026] Furthermore, the calculation formula of the stress-humidity change correlation includes:

[0027]

[0028] in, Target time The correlation between stress and humidity changes; Target time the number of corresponding neighborhood moments; Target time The corresponding Pipe shed humidity data at each neighborhood moment; Target time The corresponding Pipe shed stress data at each neighborhood moment; is the absolute value symbol; is the normalization function; is an exponential function with a natural constant as its base.

[0029] Furthermore, the method of weighting the preset initial division quantity by the interval quantity adjustment weight to obtain the adjusted interval division quantity at the target moment includes:

[0030] The product of the positive correlation mapping value of the interval quantity adjustment weight and the preset initial division quantity is rounded up to obtain the adjustment interval division quantity at the target moment.

[0031] Furthermore, the method for obtaining the degree of stress anomaly includes:

[0032] The pipe-roof stress data at the target moment and the corresponding pipe-roof stress data at all neighboring moments are used as samples for HBOS anomaly detection. The degree of anomaly is calculated by the HBOS anomaly detection method according to the number of adjustment interval divisions at the target moment, and the degree of stress anomaly at the target moment is obtained.

[0033] Furthermore, the method for obtaining the pipe roof deformation analysis according to the stress anomaly degree includes:

[0034] When there is a stress abnormality greater than the preset abnormality threshold, a pipe roof deformation warning is issued, and the stress high mask moment corresponding to the stress abnormality greater than the preset abnormality threshold is recorded.

[0035] The present invention also proposes a pipe roof deformation calculation system based on data analysis, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any step of a pipe roof deformation calculation method based on data analysis.

[0036] The present invention has the following beneficial effects:

[0037] The reason why the accuracy of stress anomaly degree obtained by using HBOS algorithm to detect anomaly of pipe-roof stress data is low is that the large change of stress data leads to the masking effect caused by ignoring the features of small changes; therefore, in order to prevent anomalies in advance and more accurately analyze the data anomalies before large changes, it is necessary to analyze the masking effect of each stress change data, so as to further analyze the stress change data with high masking effect, so as to reduce or avoid the influence of data masking on anomaly detection.

[0038] The masking property of stress data is reflected in the fact that the data anomaly of small dynamic stress changes caused by large stress changes in a short period of time is ignored. Therefore, the more rapid or more intense the stress changes in the corresponding stress change data are, the higher the corresponding masking property is, and the easier it is to ignore small dynamic stress changes. Further considering that the causes of normal small dynamic stress changes are usually the thermal expansion and contraction caused by humidity changes and the soil strength changes caused by humidity changes; the stress changes caused by thermal expansion and contraction are relatively regular and stable, and usually no abnormal stress change data will appear. Therefore, the more stable and regular the stress data changes are, the more likely it is that it is only affected by thermal expansion and contraction, that is, the less likely it is to have abnormal stress data, and the less necessary it is to further analyze it, so the masking property is smaller; therefore, combined with the intensity and stability of stress changes, the masking property of stress data at each sampling moment is obtained.

[0039] The HBOS algorithm performs anomaly detection by arranging a set of data and dividing them into intervals. The more intervals are divided, the higher the detection accuracy is, but the corresponding amount of calculation is also greater. The greater the overall masking of stress data in the local neighborhood time period, the greater the impact of data masking, and the more interval divisions are needed. In addition, since the humidity of the pipe shed affects the soil strength, the stronger the correlation between stress changes and humidity data changes in a specific scenario, the greater the impact of the stress changes in the corresponding local neighborhood data segments on objective factors caused by the environment. Therefore, the greater the necessity for more accurate anomaly detection, the greater the corresponding number of interval divisions. Therefore, the number of adjusted interval divisions can be obtained according to the distribution of stress data masking and the temporal correlation between pipe shed stress data and pipe shed humidity data. Therefore, combined with different numbers of adjusted interval divisions, the degree of stress anomaly obtained by the HBOS anomaly detection algorithm is more accurate, that is, the accuracy of data anomaly detection of small dynamic stress changes is improved, so that the accuracy of pipe shed deformation analysis based on the degree of stress anomaly is higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flow chart of a pipe-roof deformation calculation method based on data analysis provided by one embodiment of the present invention;

[0042] Figure 2 A flowchart for obtaining masked stress data provided by one embodiment of the present invention;

[0043] Figure 3 A flowchart for obtaining the number of adjustment interval divisions provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a pipe shed deformation calculation method and system based on data analysis proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0045] 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.

[0046] The following is a detailed description of a pipe-roof deformation calculation method and system based on data analysis provided by the present invention in conjunction with the accompanying drawings.

[0047] See also Figure 1 , which shows a flow chart of a pipe-roof deformation calculation method based on data analysis provided by an embodiment of the present invention, the method comprising:

[0048] Step S1: Obtain the pipe roof stress data and pipe roof humidity data at each sampling time.

[0049] An embodiment of the present invention aims to provide a pipe roof deformation calculation method based on data analysis, which is used to obtain a more accurate number of adjustment interval divisions applied to the HBOS anomaly detection algorithm according to the time series changes of the pipe roof stress data and the distribution similarity between the pipe roof stress data and the pipe roof humidity data, so that the calculated stress anomaly degree is more accurate, thereby improving the accuracy of pipe roof deformation analysis based on the stress anomaly degree.

[0050] Therefore, the embodiment of the present invention obtains the pipe roof stress data and pipe roof humidity data at each sampling moment. Specifically: at each sampling moment, a strain gauge is set on the top of the pipe roof to collect the pipe roof stress as the pipe roof stress data in the embodiment of the present invention; and at the same time, the humidity corresponding to the soil on the surface of the pipe roof is collected by the humidity sensor as the pipe roof humidity data. The implementer can also use the air humidity inside the pipe roof as the pipe roof humidity data. Its essence is to collect the humidity in the environment where the pipe roof is located, which will not be further elaborated here. In the embodiment of the present invention, the time interval between adjacent sampling moments is set to 1 second. It should be noted that the implementer can also use other stress collection devices and other humidity collection devices according to the specific implementation environment; and the interval between the sampling moments can be adjusted according to the specific environment, which will not be further limited or elaborated here.

[0051] Step S2: according to the dramatic time series changes and fluctuations of the pipe-roof stress data, the masking property of the stress data at each sampling moment is obtained; according to the masking property of the stress data at all sampling moments, the high stress masking moment is screened out.

[0052] In the specific environment of pipe sheds, the changes in pipe shed stress data are usually dynamic and large-span. For example, pipe shed stress data will have small dynamic changes due to thermal expansion and contraction caused by humidity changes, soil strength changes caused by humidity changes, etc.; when the pipe shed reaches the limit stress, large-scale deformation will occur, causing the stress data to jump out of the range of small dynamic changes, and a large stress change will occur in a short time, that is, the phenomenon of stress anomaly; the corresponding stress data change characteristics will cause the data anomaly before the large change to be ignored when the HBOS algorithm is used for anomaly detection, resulting in the inability to prevent anomalies in advance; that is, the reason why the accuracy of the stress anomaly degree obtained by using the HBOS algorithm for anomaly detection of pipe shed stress data is low is that the large change of stress data causes the characteristics of small changes to be ignored, resulting in masking; therefore, in order to prevent anomalies in advance and more accurately analyze the data anomaly before the large change, it is necessary to analyze the masking of each stress change data, so as to further analyze the stress change data with high masking, so as to reduce or avoid the impact of data masking on anomaly detection.

[0053] The masking property of stress data is reflected in the fact that the data anomaly of small dynamic stress changes caused by large-amplitude stress changes in a short time is ignored. Therefore, the more rapid or violent the stress changes in the corresponding stress change data are, the higher the corresponding masking property is, and the easier it is to ignore small dynamic stress changes. Further considering that the causes of normal small-amplitude dynamic stress changes are usually caused by thermal expansion and contraction caused by humidity changes and soil strength changes caused by humidity changes; the stress changes caused by thermal expansion and contraction are relatively regular and stable, and usually no abnormal stress change data will appear. Therefore, the more stable and regular the stress data changes are, the more likely it is that it is only affected by thermal expansion and contraction, that is, the less likely it is to have abnormal stress data, and the less necessary it is to further analyze it, so the masking property is also smaller; therefore, combined with the intensity and stability of stress changes, the embodiment of the present invention obtains the masking property of stress data at each sampling moment according to the violent time series changes and the fluctuation of time series changes of pipe-roof stress data. Thus, the stress high masking moments with higher masking property of stress data are further screened out for further analysis, so as to avoid ignoring the data anomalies of small dynamic stress changes as much as possible.

[0054] Preferably, the method for obtaining the masked stress data includes:

[0055] See also Figure 2 , which shows a flowchart of obtaining stress data masking provided by an embodiment of the present invention, the method comprising:

[0056] Step S201: taking the difference between the pipe roof stress data at each sampling moment and the pipe roof stress data at the previous sampling moment, ie, the absolute value of the difference, as the pipe roof stress change value at each sampling moment.

[0057] It should be noted that, in order to make the embodiment complete, the embodiment of the present invention sets the pipe roof stress change value at the first sampling moment to the average value of the pipe roof stress change values ​​at all other sampling moments.

[0058] Step S202: In time sequence, all sampling moments within a preset neighborhood time window of each sampling moment are used as reference moments of each sampling moment; the maximum pipe roof stress change value among all reference moments corresponding to each sampling moment is used as the reference stress change amplitude of each sampling moment.

[0059] When the HBOS algorithm detects abnormalities, the existence of pipe-roof stress data that changes greatly in a short period of time will affect the maskability of the data at each sampling moment, and the larger the change amplitude, the stronger the impact on the maskability of the data. Therefore, the embodiment of the present invention uses the maximum pipe-roof stress change value among all reference moments adjacent to each sampling moment as the reference stress change amplitude; the larger the reference stress change amplitude, the higher the data maskability of the corresponding sampling moment, and the greater the need for further analysis.

[0060] In the embodiment of the present invention, the preset time neighborhood window includes 101 sampling moments, that is, each sampling moment and the 100 sampling moments closest in time sequence are used as reference moments. When there are 50 sampling moments before and after the corresponding sampling moment, the preset time neighborhood window of the sampling moment includes the 50 sampling moments before and 50 sampling moments after the sampling moment, including the sampling moment; when the number of sampling moments before or after the sampling moment is less than 50, it is supplemented from the other side until the number of sampling moments in the preset time neighborhood window is greater than 50; and it should be noted that the total number of sampling moments in the embodiment of the present invention is greater than 101, so there will be no situation where the preset time neighborhood window cannot be obtained.

[0061] Step S203: taking the difference between the mean of the pipe roof stress change values ​​at all reference moments corresponding to each sampling moment and the pipe roof stress change value at each sampling moment as the reference stress change volatility at each sampling moment.

[0062] Since the humidity changes in different seasons are more obvious, the pipe shed will continue to be affected by thermal expansion and contraction. When the pipe shed is only affected by thermal expansion and contraction, the more stable and regular the actual stress change is, rather than abnormal stress mutation, the lower the corresponding masking property. Humidity changes are usually caused by precipitation. Compared with humidity, precipitation is not regular, that is, humidity changes are random; and humidity changes will affect soil strength. For example, when there is heavy rainfall on a certain day, it will cause soil liquefaction or loosening, resulting in irregular and abnormal changes in pipe shed stress. Under normal circumstances, the stress change caused by heavy rainfall is greater than the stress change caused by thermal expansion and contraction. Therefore, in order not to ignore the stress data anomaly in this case, it is necessary to improve its masking property for subsequent analysis. That is, if the stress change fluctuation is small and regular, it means that it is only affected by thermal expansion and contraction, and the corresponding data masking property is smaller; if the stress change fluctuation is large, it means that there may be anomalies caused by humidity changes here, and a greater masking property is required for more accurate analysis. Therefore, in order to characterize the stress change volatility at each sampling moment, the difference between the mean of the pipe-roof stress change values ​​of all reference moments corresponding to each sampling moment and the pipe-roof stress change value at each sampling moment is taken as the reference stress change volatility at each sampling moment. The closer the pipe-roof stress change value at the sampling moment is to the mean of the pipe-roof stress change values ​​of all reference moments, the smaller the reference stress change volatility is, that is, the more likely it is to be affected only by thermal expansion and contraction, and the smaller the corresponding data masking is.

[0063] Step S204: Obtaining the stress data masking property at each sampling moment according to the reference stress change amplitude and the reference stress change volatility.

[0064] The reference stress change amplitude and the reference stress change volatility are further combined, and the stress data masking property at each sampling moment is obtained according to the reference stress change amplitude and the reference stress change volatility; the reference stress change amplitude and the reference stress change volatility are both positively correlated with the stress data masking property. Preferably, the method of obtaining the stress data masking property at each sampling moment according to the reference stress change amplitude and the reference stress change volatility includes:

[0065] Since the larger the reference stress change amplitude and the larger the reference stress change volatility, the larger the corresponding stress data masking property, the embodiment of the present invention uses the normalized value of the product of the reference stress change amplitude and the reference stress change volatility as the stress data masking property at each sampling moment. It should be noted that the implementer may also use other methods to obtain the stress data masking property according to the reference stress change amplitude and the reference stress change volatility according to the specific implementation environment, for example, using the normalized value of the sum of the reference stress change amplitude and the reference stress change volatility as the stress data masking property, which will not be further elaborated here.

[0066] In the embodiment of the present invention, each sampling moment is sequentially used as the first sampling time, then The method for obtaining the masked stress data at each sampling moment is expressed in the formula as follows:

[0067]

[0068] in, For the Masking of stress data at each sampling moment; For the The maximum pipe-roof stress change value among all reference moments corresponding to the sampling moment, that is, The reference stress variation amplitude corresponding to each sampling moment; For the The change value of pipe shed stress at each sampling moment; For the The number of reference moments corresponding to each sampling moment; For the The sampling time corresponds to the The stress variation of the pipe shed at a reference time; For the The fluctuation of reference stress change corresponding to each sampling moment; is the absolute value symbol; is a normalization function. In the embodiment of the present invention, the normalization method adopts linear normalization. The implementer may also adopt other normalization methods according to the specific implementation environment, which will not be further described here.

[0069] Since the embodiment of the present invention needs to further analyze the stress change data with high masking property, the stress high masking moment is further screened out according to the stress data masking property of all sampling moments. Preferably, the method for obtaining the stress high masking moment includes:

[0070] The sampling time corresponding to the stress data masking greater than the preset masking threshold is used as the stress high masking time. In the embodiment of the present invention, the preset masking threshold is set to 0.9, and the implementer can adjust it according to the specific implementation environment, which will not be further described here.

[0071] Step S3: According to the mask distribution of stress data in the local neighborhood time period of each high stress mask moment, and the temporal correlation between the pipe roof stress data and the pipe roof humidity data, the number of adjustment interval divisions corresponding to each high stress mask moment is obtained.

[0072] The HBOS algorithm performs anomaly detection by arranging a set of data and dividing them into intervals. The more intervals are divided, the higher the detection accuracy is, but the corresponding amount of calculation is also greater. Therefore, in order to reduce the amount of calculation while ensuring real-time performance and more accurately calculate the data anomalies of dynamic stress changes, the embodiment of the present invention only improves the detection accuracy of the data segment corresponding to the stress high mask moment, that is, only the number of interval divisions corresponding to each stress high mask moment is adjusted. Since the HBOS algorithm performs anomaly detection on a segment of data, it is necessary to analyze it in combination with the local neighborhood time period corresponding to the stress high mask moment; the greater the overall masking of the stress data in the local neighborhood time period, the greater the impact of the data mask, and the more interval divisions are required.

[0073] And because the humidity of the pipe roof affects the soil strength, if the change in stress data is more consistent with the change in humidity data, it means that the correlation between the stress change and the humidity data change in the specific scenario is stronger, that is, the stress change of the corresponding local neighborhood data segment is more affected by the objective factors caused by the environment, so the necessity for more accurate anomaly detection is greater, and the corresponding number of interval divisions is larger. Therefore, the embodiment of the present invention obtains the number of adjustment interval divisions corresponding to each high stress mask moment according to the mask distribution of stress data in the local neighborhood time period at each high stress mask moment, and the time series correlation between the pipe roof stress data and the pipe roof humidity data. It should be noted that for other moments other than the high stress mask moment, the implementer can perform anomaly detection through the traditional HBOS algorithm, and the HBOS algorithm is a prior art well known to those skilled in the art, and is not further defined or elaborated here.

[0074] Preferably, the method for obtaining the number of adjustment interval divisions includes:

[0075] See also Figure 3 , which shows a flowchart of obtaining the number of adjustment interval divisions provided by an embodiment of the present invention, the method comprising:

[0076] Step S301: taking the normalized value of the negative correlation mapping value of the time interval length between each neighborhood moment and the target moment as the time interval weight of each neighborhood moment; taking the product of the stress data masking property of each neighborhood moment and the time interval weight as the weighted data masking property of each neighborhood moment; taking the average of the weighted data masking properties of all neighborhood moments corresponding to the target moment as the local mask feature value of the target moment.

[0077] Each stress high mask moment is taken as the target moment in turn; in the time sequence, all sampling moments within the preset neighborhood range of the target moment are taken as the neighborhood moments of the target moment. Although the overall stress data masking of all sampling moments within the local neighborhood time period of the target moment can be directly characterized by the influence of the data mask, the time intervals between different sampling moments and the target moment are different in the local neighborhood time period corresponding to the target moment. The larger the corresponding time interval, the smaller the degree of correlation; therefore, when the data masking influence of the local neighborhood time period is calculated with the time interval size as the weight, the number of adjustment interval divisions obtained later can be made more accurate. Therefore, in the embodiment of the present invention, the normalized value of the negative correlation mapping value of the time interval length between each neighborhood moment and the target moment is taken as the time interval weight of each neighborhood moment. In the embodiment of the present invention, all the time ranges corresponding to the 50 sampling moments before the target moment and the 50 sampling moments after the target moment are taken as the preset neighborhood range, and the implementer can adjust it according to the specific implementation environment. It should be noted that all normalization methods in the embodiment of the present invention adopt linear normalization, and the implementer can adjust the normalization method according to the specific implementation environment, which will not be further described later.

[0078] The product of the stress data masking property at each neighborhood moment and the time interval weight is used as the weighted data masking property at each neighborhood moment; the mean value of the weighted data masking property at all neighborhood moments corresponding to the target moment is used as the local masking feature value at the target moment; the larger the local masking feature value, the greater the overall stress data masking property in the local neighborhood time period of the target moment, and the greater the number of interval divisions that need to be adjusted. It should be noted that the implementer can also directly use the mean value of the stress data masking property at all sampling moments corresponding to the target moment as the local masking feature value according to the specific implementation environment.

[0079] In this embodiment of the present invention, the target time The method for obtaining the local mask feature value is expressed in the formula as follows:

[0080]

[0081] in, Target time The local mask feature value of ; Target time the number of corresponding neighborhood moments; Target time The corresponding The time interval between neighboring moments; Target time The corresponding Masking of stress data at each neighborhood moment; is the normalization function; Target time The corresponding Weighted data masking at each neighborhood moment.

[0082] Step S302: Based on the distribution similarity between the pipe-roof stress data and the pipe-roof humidity data at all neighboring moments corresponding to the target moment, the stress-humidity change correlation at the target moment is obtained.

[0083] Preferably, the calculation formula of stress-humidity change correlation includes:

[0084]

[0085] in, Target time The correlation between stress and humidity changes; Target time the number of corresponding neighborhood moments; Target time The corresponding Pipe shed humidity data at each neighborhood moment; Target time The corresponding Pipe shed stress data at each neighborhood moment; is the absolute value symbol; is the normalization function; is an exponential function with a natural constant as the base. Since the more consistent the changes in stress data are with the changes in humidity data, the larger the number of corresponding interval divisions will be; and the stronger the correlation between pipe shed stress data and pipe shed humidity data, the greater the possibility that the pipe shed stress data will change to the same degree when the pipe shed humidity data changes. Considering the different dimensions between the pipe shed humidity data and the pipe shed stress data, the difference is obtained after normalization. , and The closer the target time is, the more likely the pipe-roof stress data at the corresponding neighboring time is to be affected by the pipe-roof humidity data. Therefore, the values ​​of all neighboring times corresponding to the target time are and The closer the overall value is, the stronger the correlation between the pipe shed stress data change and the pipe shed humidity data change in the local neighborhood time period corresponding to the target time is; The more hours, and The closer the whole is, the Negatively correlated with the stress-humidity change correlation. In another embodiment of the present invention, after arranging the pipe-roof stress data of all neighboring moments corresponding to the target moment in chronological order, a neighborhood pipe-roof stress data sequence is obtained; after arranging the pipe-roof humidity data of all neighboring moments corresponding to the target moment in chronological order, a neighborhood pipe-roof humidity data sequence is obtained; the normalized value of the Pearson correlation coefficient between the neighborhood pipe-roof stress data sequence and the neighborhood pipe-roof humidity data sequence is used as the stress-humidity change correlation at the target moment; in other embodiments of the present invention, the normalized value of the structural correlation coefficient between the neighborhood pipe-roof stress data sequence and the neighborhood pipe-roof humidity data sequence is used as the stress-humidity change correlation at the target moment. It should be noted that the calculation method of the Pearson correlation coefficient and the structural correlation coefficient is a prior art well known to those skilled in the art, and will not be further limited or elaborated herein.

[0086] Step S303: taking the product of the local mask feature value and the stress-humidity change correlation as the interval quantity adjustment weight at the target time.

[0087] Since the larger the local mask eigenvalue is, the greater the stress-humidity change correlation is, the greater the corresponding interval quantity demand is, and the corresponding interval quantity adjustment weight is greater, so the embodiment of the present invention uses the product of the local mask eigenvalue and the stress-humidity change correlation as the interval quantity adjustment weight at the target moment. In another embodiment of the present invention, the normalized value of the product of the local mask eigenvalue and the stress-humidity change correlation is used as the interval quantity adjustment weight; the implementer can also calculate the interval quantity adjustment weight by other methods according to the specific implementation environment.

[0088] Step S304: obtaining a preset initial division number in the HBOS anomaly detection method; weighting the preset initial division number by the interval number adjustment weight to obtain the adjusted interval division number at the target moment.

[0089] In the embodiment of the present invention, the preset initial number of divisions is set to 8, and the implementer can adjust it according to the specific implementation environment, which will not be further described here. Preferably, the method of weighting the preset initial number of divisions by adjusting the number of intervals to obtain the adjusted number of interval divisions at the target time includes:

[0090] The product between the positive correlation mapping value of the interval quantity adjustment weight and the preset initial division number is rounded up to obtain the adjustment interval division number at the target moment. Since the larger the interval quantity adjustment weight, the larger the corresponding adjustment interval division number, the weighting is performed by multiplication.

[0091] In this embodiment of the present invention, the target time The method for obtaining the number of adjustment interval divisions is expressed in the formula as follows:

[0092]

[0093] in, Target time The number of adjustment intervals divided into; The preset initial division quantity; Target time The local mask feature value of ; Target time The correlation between stress and humidity changes; Target time The number of intervals adjusts the weight; Target time The number of intervals adjusts the weights of the positively correlated mapping values.

[0094] Step S4: according to the number of adjustment interval divisions combined with the HBOS anomaly detection method, the stress anomaly degree at each stress high mask moment is obtained; and the pipe roof deformation analysis is performed according to the stress anomaly degree.

[0095] At this point, the number of adjustment interval divisions corresponding to each stress high mask moment in the local neighborhood time period is obtained, so that HBOS anomaly detection with different precisions can be performed according to the size of the number of adjustment interval divisions, which improves the calculation efficiency and makes the calculated stress anomaly degree more accurate. Further, the stress anomaly degree at each stress high mask moment is obtained according to the number of adjustment interval divisions combined with the HBOS anomaly detection method.

[0096] Preferably, the method for obtaining the degree of stress anomaly includes:

[0097] The embodiment of the present invention uses the pipe-roof stress data at the target time and the corresponding pipe-roof stress data at all neighboring times as samples for HBOS anomaly detection, and calculates the degree of anomaly through the HBOS anomaly detection method according to the number of adjustment interval divisions at the target time, to obtain the degree of stress anomaly at the target time. Since the number of interval divisions is a technical term in the HBOS anomaly detection algorithm, the embodiment of the present invention does not further elaborate on the significance, implementation method, and calculation of the degree of stress anomaly by the number of interval divisions.

[0098] Finally, the embodiment of the present invention performs pipe roof deformation analysis according to the degree of stress anomaly. Preferably, the method for obtaining the pipe roof deformation analysis according to the degree of stress anomaly includes:

[0099] Since the greater the stress anomaly, the more abnormal the pipe roof is, the pipe roof deformation warning is performed when the normalized stress anomaly is greater than the preset abnormal threshold, and the stress high mask moment corresponding to the normalized stress anomaly greater than the preset abnormal threshold is recorded, that is, the stress high mask moment with a larger stress anomaly is used as the abnormal pipe roof deformation moment. When receiving an early warning, the pipe roof is reinforced in time to avoid a larger safety accident. In the embodiment of the present invention, the preset abnormal threshold is set to 0.8, and the implementer can adjust it according to the specific implementation environment.

[0100] In summary, the present invention is a pipe roof deformation calculation method and system based on data analysis. The present invention selects high stress mask moments according to the dramatic time series changes and fluctuations of the pipe roof stress data; obtains the number of adjustment interval divisions used in the HBOS algorithm according to the distribution of the masking properties of the stress data in the local neighborhood time period of the high stress mask moment, and the time series correlation between the pipe roof stress data and the pipe roof humidity data; makes the stress anomaly degree obtained after further combining with the HBOS algorithm more accurate, improves the accuracy of data anomaly detection of small dynamic stress changes, and thus makes the accuracy of pipe roof deformation analysis based on the stress anomaly degree higher.

[0101] The present invention also proposes a pipe roof deformation calculation system based on data analysis, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any step of a pipe roof deformation calculation method based on data analysis.

[0102] It should be noted that the sequence of the above embodiments of the present invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0103] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A pipe-roof deformation calculation method based on data analysis, characterized in that: The method comprises: Obtain the pipe shed stress data and pipe shed humidity data at each sampling moment; According to the dramatic time series changes and fluctuations of the pipe-roof stress data, the masking property of the stress data at each sampling moment is obtained; according to the masking property of the stress data at all sampling moments, the masked moments with high stress are screened out; According to the mask distribution of stress data in the local neighborhood time period of each high stress mask moment, and the temporal correlation between the pipe shed stress data and the pipe shed humidity data, the number of adjustment interval divisions corresponding to each high stress mask moment is obtained; According to the number of adjustment interval divisions combined with the HBOS anomaly detection method, the stress anomaly degree at each stress high mask moment is obtained; and the pipe roof deformation analysis is performed according to the stress anomaly degree; The method for obtaining the masked stress data includes: The difference between the pipe roof stress data at each sampling moment and the pipe roof stress data at the previous sampling moment is taken as the pipe roof stress change value at each sampling moment; In terms of time sequence, all sampling moments within the preset neighborhood time window of each sampling moment are taken as the reference moment of each sampling moment; the maximum pipe-roof stress change value among all reference moments corresponding to each sampling moment is taken as the reference stress change amplitude of each sampling moment; The difference between the mean of the pipe-roof stress change values ​​of all reference moments corresponding to each sampling moment and the pipe-roof stress change value at each sampling moment is taken as the reference stress change volatility at each sampling moment; According to the reference stress change amplitude and the reference stress change volatility, the stress data masking property at each sampling moment is obtained; the reference stress change amplitude and the reference stress change volatility are both positively correlated with the stress data masking property; The method for obtaining the masking property of stress data at each sampling moment according to the reference stress change amplitude and the reference stress change volatility comprises: The normalized value of the product of the reference stress change amplitude and the reference stress change volatility is used as the stress data mask at each sampling moment.

2. The pipe-roof deformation calculation method based on data analysis according to claim 1 is characterized in that: The method for obtaining the stress high mask moment comprises: The sampling moment corresponding to the masked stress data greater than the preset mask threshold is taken as the high stress mask moment.

3. The pipe-roof deformation calculation method based on data analysis according to claim 1 is characterized in that: The method for obtaining the number of adjustment interval divisions includes: Taking each high stress mask moment as the target moment in turn; taking all sampling moments within a preset neighborhood range of the target moment as the neighborhood moments of the target moment in time sequence; The normalized value of the negative correlation mapping value of the time interval length between each neighborhood moment and the target moment is used as the time interval weight of each neighborhood moment; the product of the stress data masking property of each neighborhood moment and the time interval weight is used as the weighted data masking property of each neighborhood moment; the mean value of the weighted data masking properties of all neighborhood moments corresponding to the target moment is used as the local mask feature value of the target moment; According to the distribution similarity between the pipe shed stress data and the pipe shed humidity data at all neighboring moments corresponding to the target moment, the stress-humidity change correlation at the target moment is obtained; The product of the local mask feature value and the stress-humidity change correlation is used as the interval quantity adjustment weight at the target time; The preset initial division number in the HBOS anomaly detection method is obtained; the preset initial division number is weighted by the interval number adjustment weight to obtain the adjusted interval division number at the target moment.

4. The pipe-roof deformation calculation method based on data analysis according to claim 3 is characterized in that: The calculation formula of the stress-humidity change correlation includes: Among them, X k is the stress-humidity change correlation at the target time k; N k is the number of neighboring moments corresponding to the target moment k; s k,i is the pipe-roof humidity data of the i-th neighboring time corresponding to the target time k; f k,i is the pipe-roof stress data of the i-th neighborhood moment corresponding to the target moment k; || is the absolute value symbol; Norm() is the normalization function; exp() is the exponential function with the natural constant as the base.

5. The pipe-roof deformation calculation method based on data analysis according to claim 3 is characterized in that: The method of weighting the preset initial division quantity by the interval quantity adjustment weight to obtain the adjusted interval division quantity at the target moment includes: The product of the positive correlation mapping value of the interval quantity adjustment weight and the preset initial division quantity is rounded up to obtain the adjustment interval division quantity at the target moment.

6. The pipe-roof deformation calculation method based on data analysis according to claim 3 is characterized in that: The method for obtaining the stress anomaly degree includes: The pipe-roof stress data at the target moment and the corresponding pipe-roof stress data at all neighboring moments are used as samples for HBOS anomaly detection. The degree of anomaly is calculated by the HBOS anomaly detection method according to the number of adjustment interval divisions at the target moment, and the degree of stress anomaly at the target moment is obtained.

7. The pipe-roof deformation calculation method based on data analysis according to claim 1 is characterized in that: The method for obtaining the pipe roof deformation analysis according to the stress anomaly degree includes: When there is a stress abnormality greater than the preset abnormality threshold, a pipe roof deformation warning is issued, and the stress high mask moment corresponding to the stress abnormality greater than the preset abnormality threshold is recorded.

8. A pipe-roof deformation calculation system based on data analysis, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the pipe-roof deformation calculation method based on data analysis as claimed in any one of claims 1 to 7 are implemented.

Images