Multi-source sensor abnormal edge identification method and device based on prior constraint and medium

Abstract

The invention discloses a priori constraint-based multi-source sensor abnormal edge identification method and device and a medium, and relates to the technical field of slope engineering health monitoring. The method comprises the following steps: acquiring an increment sequence and an accumulation sequence of multi-source sensor monitoring data in a safe operation stage; acquiring real-time monitoring data of the multi-source sensor, and performing abnormal point identification on the data at each moment; for moments when abnormal points appear in the cumulative sequence and the incremental sequence at the same time, analyzing the correlation between the sequences by adopting a prior constraint method, and judging an abnormal state; when the abnormal state is judged, the acquisition frequency of the sensor is accelerated, and the working state of the sensor is judged based on a rejection comparison method. The method provided by the invention solves the problem of misjudgment caused by neglecting the relevance of the related sensing device by depending on the data of the sensor per se when the working state of the sensor is judged in the past.

Description

Technical Field

[0001] The present invention relates to the technical field of slope engineering health monitoring, and more specifically, to a method, device and medium for identifying abnormal edges of multi-source sensors based on prior constraints. Background Art

[0002] Slope health monitoring is an important research direction in the fields of geological disaster prevention and control and engineering safety guarantee. With the acceleration of infrastructure construction and urbanization, the stability problem of slopes has become a major hidden danger affecting the operation safety of key projects such as roads, railways, and tunnels. Slope instability is usually accompanied by complex dynamic processes such as soil mass sliding, rock mass cracking, and stress redistribution, which are sudden and destructive, and the need for real-time monitoring and early warning is particularly urgent.

[0003] In recent years, the progress of sensing technology, artificial intelligence, and big data technology has provided new solutions for slope health monitoring. Monitoring systems based on sensor technologies such as distributed optical fiber sensing, GNSS (Global Navigation Satellite System), and inclinometers can accurately obtain multi-dimensional data such as slope displacement, crack changes, and rainfall. At the same time, the combination of machine learning algorithms and numerical simulation technologies can analyze the spatial distribution characteristics and time evolution laws of monitoring data, and realize the prediction and evaluation of potential slope instability risks.

[0004] However, due to its special geographical location and complex geological structure, the slope has very strict requirements for the accuracy of monitoring technology, but the sensors are easily affected by external interference, and the current monitoring technology still has quite deficiencies. In actual operation, due to equipment limitations, environmental interference (such as electromagnetic noise, changes in atmospheric conditions, etc.) or improper selection of installation locations, the monitoring data may be affected, resulting in a decrease in accuracy. Moreover, affected by environmental factors (such as extreme weather, long-term vibration, etc.), the sensors may also be damaged or their performance may decline, affecting the working stability of the monitoring system. In addition, slope health monitoring often involves multiple sensors and algorithm technologies, and each sensor provides different information about the slope state. The inconsistency, redundancy, and complementarity of these data all require further processing and analysis to improve the accuracy of the monitoring system's determination. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a method, device and medium for identifying abnormal edges of multi-source sensors based on prior constraints, so as to solve the problem that the previous determination of the working state of sensors relies on the data of the sensors themselves and ignores the relevance of related sensing devices, resulting in inaccurate determination.

[0006] In a first aspect, the present invention provides a method for identifying abnormal edges of multi-source sensors based on prior constraints, the method comprising:

[0007] Obtain the incremental sequence and cumulative sequence of the multi-source sensor monitoring data in the safe operation stage;

[0008] Obtain the real-time monitoring data of the multi-source sensor, and identify the abnormal points for the data at each moment;

[0009] For the moments when abnormal points appear in both the cumulative sequence and the incremental sequence, use the prior constraint method to analyze the correlation between the sequences and determine the abnormal state;

[0010] When it is determined that an abnormal state occurs, increase the acquisition frequency of the sensor, and determine the working state of the sensor based on the rejection comparison method.

[0011] Further, obtaining the incremental sequence and cumulative sequence of the multi-source sensor monitoring data in the safe operation stage includes:

[0012] Taking the sequence composed of the change values of the current moment relative to the previous moment at each acquisition moment as the incremental sequence, denoted as X′ = [x′1, x′2,..., x′ n ; where X′ is the incremental sequence, and x′1, x′2,..., x′ n represent the increments at different moments, and the initial moment increment x′1 is set to 0;

[0013] Taking the sequence composed of the sensor values collected at each acquisition moment as the cumulative sequence, denoted as X = [x1, x2,..., x n ; where X is the cumulative sequence, and x1, x2,..., x n represent the sensor values at different moments, and the initial moment sensor value x1 is set to 0.

[0014] Further, obtaining the real-time monitoring data of the multi-source sensor and identifying the abnormal points for the data at each moment includes:

[0015] Based on the real-time monitoring data of the multi-source sensor, use the fast Fourier transform to identify the first main frequency of the data;

[0016] Taking the real-time monitoring data of the multi-source sensor as the main sequence, determining the division parameter according to the period of the first main frequency, and using the division parameter to divide the main sequence. If the final data length does not meet the division parameter, borrow forward until a complete sequence is formed to obtain at least two subsequences; where the subsequence is expressed as:

[0017] X k = [[x1,x2,…,x 2T ,[x 2T+1 ,x 2T+2 ,…,x 4T ,…,[x 2*kT-a+1 ,x2*kT-a+2 ,...,x n

[0018] In the formula, X k represents a subsequence, T represents the first main frequency period, a represents the carry borrowed forward, x1, x2, x 2T respectively represent the sensor data at the initial moment, the second moment, and the 2T moment, x 2T+1 , x 2T+2 , x 4T respectively represent the sensor data at the 2T + 1, 2T + 12, and 4T moments, x 2*kT-a+1 , x 2*kT-a+2 , x n respectively represent the sensor data at the 2*kT - a + 1, 2*kT - a + 2, and n moments;

[0019] Set a judgment formula, and determine the data points that meet the judgment formula in each subsequence as outliers.

[0020] Furthermore, the division parameter is twice the period of the first main frequency.

[0021] Furthermore, the judgment formula is expressed as:

[0022]

[0023] In the formula, ABS is the absolute value function, x i is the i-th value in the subsequence, is the average value of the subsequence, σ is the standard deviation in the subsequence, is the absolute value of the average slope in this subsequence.

[0024] Furthermore, for the moments when abnormal points appear simultaneously in the cumulative sequence and the incremental sequence, use the prior constraint method to analyze the correlation between sequences and determine the abnormal state, including:

[0025] Select the moment when abnormal points first appear simultaneously among all sensors, and divide the data sequences of all sensors into prior cumulative stable sequences

[0026] X f1 = [X f1 = [x f11 , x f12 , …, x f1y , X f2 = [x f21 , x f22 , …, x f2y , …, X fm = [x fm1 , x fm2 ,..., x fmy ​​

[0027] Where m is the number of multi-source sensors, X f1 is the prior cumulative stable sequence of the first sensor, X f2 is the prior cumulative stable sequence of the second sensor, X fm is the prior cumulative stable sequence of the m-th sensor, x f11 , x f12 , x f1y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the first sensor, x f21 , x f22 , x f2y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the second sensor, x fm1 , x fm2 , x fmy are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the m-th sensor.

[0028] The Pearson correlation analysis method is used to analyze the correlation between the cumulative sequence and the incremental sequence data respectively, and the prior cumulative sequence correlation matrix and the incremental sequence correlation matrix are obtained;

[0029] The Pearson correlation analysis method is used to analyze the original cumulative sequence and incremental sequence containing outliers respectively, and the cumulative sequence correlation matrix and the incremental sequence correlation matrix are obtained;

[0030] An anomaly discriminant is established. For the anomaly data of the i-th sensor, if the anomaly discriminant is satisfied, it is determined that the monitoring system is in an abnormal state; if the anomaly discriminant is not satisfied, it is determined that the monitoring system is in a normal state.

[0031] Furthermore, the anomaly discriminant is expressed as:

[0032]

[0033] Where A ij and A ij ' represent the elements in the correlation matrix of the cumulative sequence and the correlation matrix of the incremental sequence respectively; i and j are different sensor numbers. When checking the i-th sensor, the value of i is fixed; A f,ij ' is the element in the correlation matrix of the incremental sequence of the prior stable sequence, and A f,ij is the element in the correlation matrix of the cumulative sequence of the prior stable sequence.

[0034] Furthermore, when an abnormal state is determined, the acquisition frequency of the sensor is increased, and the working state of the sensor is determined based on the elimination comparison method, including:

[0035] Delete the data values at the abnormal point moment, and record the cumulative sequence of the subsequent sampling data of each sensor and the incremental sequence After deleting the values at the abnormal moment, reconstruct the cumulative sequence and the incremental sequence from the start of monitoring. The reconstructed cumulative sequence and incremental sequence are respectively expressed as:

[0036] X a =[x1, x2,..., x n , x n+2 ,..., x n+l ;

[0037] X′ a =[x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′];

[0038] Based on the prior constraint method, perform posterior analysis to analyze the cumulative sequence of the subsequent sampling data between different sensors and the incremental sequence to determine whether the state is normal; use the prior constraint method to perform prior analysis to analyze the cumulative sequence X a =[x1, x2,..., x n , x n+2 ,..., x n+l and the incremental sequence X′ a =[x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′] to determine whether the state is normal;

[0039] If the posterior analysis determines normal and the prior analysis determines abnormal, it indicates that there is a disturbance in the sensor, then re-zeroing is performed; if the posterior analysis determines abnormal and the prior analysis determines abnormal, it indicates that the working state of the sensor is abnormal, then further manual confirmation is performed; if the posterior analysis determines normal and the prior analysis determines normal, it indicates that the acquisition process is disturbed and the cause of the abnormal point comes from external interference.

[0040] In a second aspect, the present invention provides a multi-source sensor abnormal edge recognition device based on prior constraints. The device includes:

[0041] A data acquisition module configured to acquire the incremental sequence and the cumulative sequence of the multi-source sensor monitoring data in the safe operation stage;

[0042] An abnormal point recognition module configured to acquire the real-time monitoring data of the multi-source sensor and perform abnormal point recognition on the data at each moment;

[0043] The abnormal state identification module is configured to use a priori constraint method to analyze the correlation between the sequences and determine the abnormal state when the abnormal point appears simultaneously in the cumulative sequence and the incremental sequence;

[0044] The working state determination module is configured to increase the sensor acquisition frequency when determining that an abnormal state occurs, and determine the sensor working state based on a elimination comparison method.

[0045] In a third aspect, the present invention provides a readable storage medium, wherein the readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method as described above.

[0046] The present invention has at least the following beneficial effects:

[0047] The present invention is easy to deploy at the edge of monitoring and can analyze the monitoring status in real time, solving the problem of bandwidth waste caused by data transmission and analysis and uploading to the site, so that the status parameters can be transmitted back to the data center in time to report the effectiveness of the monitoring system; in addition, in the process of analyzing data, the correlation between different data is considered, and the monitoring system status is determined by the historical prior correlation and posterior correlation between the data collected by different sensors. The deformation development relationship of the structure can be analyzed comprehensively, avoiding the autoregression problem of predicting its own state by its own data. In addition, an abnormality recognition algorithm considering the pane and absolute slope is constructed, which improves the adaptability of abnormal point discrimination of data series with certain trends or growth slopes while considering the abnormality test of short-term stability of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flowchart of a method for identifying abnormal edges of multi-source sensors based on prior constraints according to an embodiment of the present invention is shown.

[0049] Figure 2 A Fourier fast changing main frequency diagram according to an embodiment of the present invention is shown.

[0050] Figure 3 A data partitioning and error identification diagram according to an embodiment of the present invention is shown.

[0051] Figure 4 A correlation matrix diagram according to an embodiment of the present invention is shown.

[0052] Figure 5 A structural diagram of a multi-source sensor abnormal edge recognition device based on a priori constraints according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0053] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings and specific examples, but it is not a limitation to the present invention. For the various steps described herein, if there is no necessity for a sequential relationship between them, the order in which they are described as examples herein should not be regarded as a limitation. Those skilled in the art should know that they can be adjusted in order as long as the logic between them is not destroyed and the entire process cannot be realized.

[0054] An embodiment of the present invention provides a multi-source sensor abnormal edge recognition method based on prior constraints.

[0055] Please refer to Figure 1 . The multi-source sensor abnormal edge recognition method based on prior constraints specifically includes the following steps S1 to S4, which are introduced in detail as follows.

[0056] S1. Obtain the incremental sequence and cumulative sequence of the monitoring data of the multi-source sensor in the safe operation stage.

[0057] In some embodiments, in step S1, the incremental sequence is a sequence X′ = [x′1, x′2,..., x′ n composed of the change values at each acquisition moment compared with the previous moment. For the initial moment x′1, the increment is set to 0; the cumulative sequence is a sequence X = [x1, x2,..., x n composed of the sensor values collected at each acquisition moment. For the initial moment x1, the sensor value is adjusted to 0;

[0058] S2. Obtain the real-time monitoring data of the multi-source sensor and identify abnormal points for the data at each moment.

[0059] In some embodiments, in step S2, the abnormal points of the data at each moment are specifically identified through the following steps S21 to S23:

[0060] S21. For the newly collected data, perform a fast Fourier transform to obtain a new transformed sequence:

[0061]

[0062] And extract the first main frequency f max of the transformed sequence from the maximum value of the absolute value |X[k]| of X[k], that is

[0063]

[0064] k max = argmax∣X[k]∣

[0065]

[0066] Wherein, X[k] is the value of the Fourier transform of the signal x[n] at the k-th frequency point. j is the imaginary unit. is the kernel function of the Fourier transform. Due to the symmetry of the FFT result, we usually only need to consider the part where k ranges from 0 to N / 2 - 1 (or an integer multiple of N / 2, depending on whether N is odd or even), because the higher frequency part is a mirror image of the first half.

[0067] Such as Figure 2 shown, it is the main frequency diagram of the fast Fourier transform.

[0068] S22, divide the sequence according to twice the first main frequency period T, i.e., 2T, as a parameter. If the length of the last data is less than 2T, consider borrowing forward until a complete sequence is formed to obtain several subsequences X k =[[x1, x2,..., x 2T , [x 2T+1 , x 2T+2 ,..., x 4T ,..., [x 2*kT-a+1 , x 2*kT-a+2 , …, x n , where a represents the number of borrows forward. The division result is as shown in Figure 3 shown.

[0069] S23, in each subsequence, if there is a point that satisfies the following formula, then this point is considered an outlier.

[0070]

[0071] Wherein, ABS() is the absolute value function, x i is each value in this subseries, is the average value of this subsequence, σ is the standard deviation in this subsequence, is the absolute value of the average slope in this subsequence, and T is the period.

[0072] S3. For the moments when abnormal points appear simultaneously in the cumulative sequence and the incremental sequence, use the prior constraint method to analyze the correlation between the sequences and determine the abnormal state.

[0073] In some embodiments, step S30 is implemented by the following steps S31 to S34:

[0074] S31, analyze the monitoring data of different sensors, and select the moment when abnormal points first appear simultaneously among all sensors to divide the data sequences of all sensors into prior cumulative stable sequences X f =[X f1 =[xf11 , x f12 ,..., x f1y , X f2 = [x f21 , x f22 ,..., x f2y ,..., X fm = [x fm1 , x fm2 ..., x fmy ,

[0075] In the formula, m is the number of multi-source sensors, X f1 is the prior cumulative stable sequence of the first sensor, X f2 is the prior cumulative stable sequence of the second sensor, X fm is the prior cumulative stable sequence of the m-th sensor, x f11 , x f12 , x f1y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the first sensor, x f21 , x f22 , x f2y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the second sensor, x fm1 , x fm2 , x fmy are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the m-th sensor.

[0076] S32. Use the Pearson correlation analysis method to analyze the correlation between the data of the prior cumulative sequence X f [k] and the increment sequence X' f [k]. There is the following formula

[0077]

[0078]

[0079] In the formula, ρ f，ij is the Pearson correlation coefficient, X fik and X fjk are the observed values of two variables, and are the means of the two variables.

[0080] Thus, the correlation matrix of the prior cumulative sequence and the correlation matrix of the increment sequence are obtained. As Figure 4 shown, it is the obtained correlation matrix diagram. The correlation matrix of the prior cumulative sequence and the correlation matrix of the increment sequence are expressed as:

[0081]

[0082] S33. Use the Pearson correlation analysis method to analyze the original cumulative sequence and the incremental sequence containing outliers respectively. Similarly, obtain the cumulative sequence correlation matrix and the incremental sequence correlation matrix, which are expressed as:

[0083]

[0084] S34. For the abnormal data of the i-th sensor, compare the correlation relationship of the correlation matrix according to the following formula. If the following formula is satisfied, it is considered that the monitoring system is in an abnormal state; otherwise, it is considered to be in a normal state.

[0085]

[0086] In the formula, A ij and A ij ′ respectively represent the elements in the correlation matrix of the cumulative sequence and the correlation matrix of the incremental sequence. i and j are different sensor numbers respectively. When performing the test of the i-th sensor, the value of i is fixed.

[0087] S4. When it is determined that an abnormal state occurs, increase the acquisition frequency of the sensor, and determine the working state of the sensor based on the elimination comparison method.

[0088] In some embodiments, step S40 is implemented by the following steps S41 to S43:

[0089] S41. Delete the data value at the outlier moment, and record the cumulative sequence and the incremental sequence of the subsequent sampling data of each sensor. In addition, delete the value at the abnormal moment and reconstruct the cumulative sequence X a = [x1, x2,..., x n , x n+2 ,..., x n+l and the incremental sequence X′ a = [x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′].

[0090] S42. Use the prior constraint method described in S3 to perform posterior analysis and analyze the cumulative sequence and the incremental sequence of the subsequent sampling data between different sensors to determine whether the state is normal; use the prior constraint method described in S3 to perform prior analysis and analyze the cumulative sequence X a = [x1, x2,..., x n , x n+2 ,..., x n+l and the incremental sequence X′a = [x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′], determine whether the status is normal.

[0091] S43, as shown in Table 1, is the sensor status determination matrix. If the posterior analysis determines normal and the prior analysis determines abnormal, it indicates that there is a perturbation in the sensor and re-zeroing is required; if the posterior analysis determines abnormal and the prior analysis determines abnormal, it indicates that the working state of the sensor is abnormal and further manual confirmation is needed; if the posterior analysis determines normal and the prior analysis determines normal, it indicates that the acquisition process is perturbed and the cause of the abnormal point comes from external interference.

[0092] Table 1 Sensor Status Determination Matrix

[0093] Status Zero adjustment required for perturbation External interference Abnormal sensor operation Prior analysis Abnormal Normal Abnormal Posterior analysis Normal Normal Abnormal

[0094] The embodiment of the present invention also provides a multi-source sensor abnormal edge recognition device based on prior constraints, as Figure 5 shown. The device includes:

[0095] A data acquisition module 501, configured to acquire the incremental sequence and cumulative sequence of multi-source sensor monitoring data during the safe operation stage;

[0096] An abnormal point recognition module 502, configured to acquire the real-time monitoring data of multi-source sensors and identify abnormal points for the data at each moment;

[0097] An abnormal status recognition module 503, configured to, for the moments when abnormal points appear simultaneously in the cumulative sequence and the incremental sequence, analyze the correlation between the sequences using the prior constraint method and determine the abnormal status;

[0098] A working status determination module 504, configured to, when an abnormal status is determined, increase the acquisition frequency of the sensor and determine the working status of the sensor based on the elimination comparison method.

[0099] In some embodiments, the data acquisition module is further configured to:

[0100] Take the sequence composed of the change values occurring at the current moment relative to the previous moment at each acquisition moment as the incremental sequence, denoted as X′ = [x′1, x′2,..., x′ n ; where X′ is the incremental sequence, and x′1, x′2,..., x′ n represent the increments at different moments, and the initial moment increment x′1 is set to 0;

[0101] The sequence composed of the sensor values collected at each acquisition moment is used as the cumulative sequence, denoted as X = [x1, x2,..., x n ; where X is the cumulative sequence, and x1, x2,..., x n represent the sensor values at different moments, and the sensor value x1 at the initial moment is set to 0.

[0102] In some embodiments, the outlier recognition module is further configured to:

[0103] Based on the real-time monitoring data of multi-source sensors, use the fast Fourier transform to identify the first main frequency of the data;

[0104] Taking the real-time monitoring data of the multi-source sensors as the main sequence, determining the division parameter according to the period of the first main frequency, and using the division parameter to divide the main sequence. If the final data length does not meet the division parameter, borrow forward until a complete sequence is formed, obtaining at least two subsequences; where the subsequences are expressed as:

[0105] X k =[[x1, x2,…, x 2T ,[x 2T+1 , x 2T+2 ,…, x 4T ,…,[x 2*kT-a+1 , x 2*kT-a+2 ,..., x n

[0106] In the formula, X k represents the subsequence, T represents the period of the first main frequency, a represents the number of borrows forward, x1, x2, x 2T represent the sensor data at the initial moment, the second moment, and the 2T moment respectively, and x 2T+1 , x 2T+2 , x 4T represent the sensor data at 2T + 1, 2T + 2, and 4T moments respectively, and x 2*kT-a+1 , x 2*kT-a+2 , x n represent the sensor data at 2*kT - a + 1, 2*kT - a + 2, and n moments respectively;

[0107] Set a judgment formula, and determine the data points that meet the judgment formula in each subsequence as outliers.

[0108] In some embodiments, the division parameter is twice the period of the first main frequency.

[0109] In some embodiments, the judgment formula is expressed as:

[0110] ​

[0111] In the formula, ABS is the absolute value function, and x i is the i-th value in the subsequence, is the average value of the subsequence, σ is the standard deviation in the subsequence, is the absolute value of the average slope in this subsequence.

[0112] In some embodiments, the abnormal state recognition module is further configured to:

[0113] Select the moment when abnormal points first appear simultaneously in all sensors to divide the data sequences of all sensors into prior cumulative stable sequences

[0114] X f1 =[X f1 =[x f11 , x f12 , …, x f1y , X f2 =[x f21 , x f22 , …, x f2y , …, X fm =[x fm1 , x fm2 ,..., x fmy

[0115] In the formula, m is the number of multi-source sensors, X f1 is the prior cumulative stable sequence of the first sensor, X f2 is the prior cumulative stable sequence of the second sensor, X fm is the prior cumulative stable sequence of the m-th sensor, x f11 , x f12 , x f1y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the first sensor, x f21 , x f22 , x f2y are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the second sensor, x fm1 , x fm2 , x fmy are the 1st, 2nd, and y-th values in the prior cumulative stable sequence of the m-th sensor.

[0116] Adopt the Pearson correlation analysis method to analyze the correlation between the cumulative sequence and the incremental sequence data respectively, and obtain the prior cumulative sequence correlation matrix and the incremental sequence correlation matrix;

[0117] Adopt the Pearson correlation analysis method to analyze the original cumulative sequence and incremental sequence containing abnormal points respectively, and obtain the cumulative sequence correlation matrix and the incremental sequence correlation matrix; ​

[0118] An anomaly discriminant is established. For the anomaly data of the i-th sensor, if the anomaly discriminant is satisfied, it is determined that the monitoring system is in an abnormal state; if the anomaly discriminant is not satisfied, it is determined that the monitoring system is in a normal state.

[0119] In some embodiments, the anomaly discriminant is expressed as:

[0120]

[0121] In the formula, A ij and A ij ' respectively represent elements in the correlation matrix of the cumulative sequence and the correlation matrix of the incremental sequence; i and j are different sensor numbers respectively. When performing the test on the i-th sensor, the value of i is fixed; A f,ij ' is an element in the correlation matrix of the incremental sequence of the prior stable sequence, and A f,ij is an element in the correlation matrix of the cumulative sequence of the prior stable sequence.

[0122] In some embodiments, the working state determination module is further configured to:

[0123] Delete the data values at the anomaly point moment, and record the cumulative sequence and the incremental sequence of the subsequent sampling data of each sensor

[0124] After deleting the values at the anomaly moment, reconstruct the cumulative sequence and the incremental sequence starting from the monitoring. The reconstructed cumulative sequence and incremental sequence are respectively expressed as: a X n = [x1, x2,..., x n+2 , x n+l ,..., x a X' n = [x1', x2',..., x n+2 ', x n+l ',..., x a '];

[0126] Based on the prior constraint method, perform posterior analysis to analyze the cumulative sequence and the incremental sequence of the subsequent sampling data between different sensors to determine whether the state is normal; adopt the prior constraint method to perform prior analysis to analyze the cumulative sequence X n = [x1, x2,..., x n+2 , x n+l ,..., x a and the incremental sequence X' n= [x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′], determine whether the determination status is normal;

[0127] If the posterior analysis determines normal and the prior analysis determines abnormal, it indicates that there is a disturbance in the sensor, and then re-zeroing is performed; if the posterior analysis determines abnormal and the prior analysis determines abnormal, it indicates that the working state of the sensor is abnormal, and then further manual confirmation is performed; if the posterior analysis determines normal and the prior analysis determines normal, it indicates that the acquisition process is disturbed and the cause of the abnormal point comes from external interference.

[0128] It should be noted that the structures of the various multi-source sensor abnormal edge recognition devices based on prior constraints described in this embodiment belong to the same inventive concept as the previously described multi-source sensor abnormal edge recognition method based on prior constraints, and achieve the same beneficial effects through the same principle, which will not be elaborated here.

[0129] The embodiment of the present invention also provides a readable storage medium, and the readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in any of the above embodiments.

[0130] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present invention having equivalent elements, modifications, omissions, combinations (e.g., schemes of cross-combination of various embodiments), adaptations or changes. The elements in the claims will be broadly interpreted based on the language used in the claims and are not limited to the examples described in this specification or during the implementation of this application, and the examples will be interpreted as non-exclusive. Therefore, this specification and the examples are only intended to be considered as examples, and the true scope and spirit are indicated by the full scope of the following claims and their equivalents.

[0131] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of their schemes) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description. Additionally, in the above specific embodiments, various features can be grouped together to simplify the present invention. This should not be construed as an intention that the features of an invention not claimed are necessary for any claim. On the contrary, the subject matter of the present invention can be less than all the features of a particular embodiment of the invention. Thus, the following claims are incorporated herein as examples or embodiments into the specific embodiments, where each claim independently serves as a separate embodiment, and considering these embodiments, they can be combined with each other in various combinations or arrangements. The scope of the present invention should be determined with reference to the appended claims and the full scope of the equivalent forms empowered by these claims.

Claims

1. A multi-source sensor abnormal edge recognition method based on prior constraints, characterized in that, The method includes: Obtaining the incremental sequence and cumulative sequence of multi-source sensor monitoring data in the safe operation stage; Obtaining the real-time monitoring data of multi-source sensors and identifying abnormal points for the data at each moment; For the moments when abnormal points appear in both the cumulative sequence and the incremental sequence, using the prior constraint method to analyze the correlation between the sequences and determining the abnormal state; When it is determined that an abnormal state occurs, increasing the acquisition frequency of the sensors and determining the working state of the sensors based on the elimination and comparison method.

2. The multi-source sensor abnormal edge recognition method based on prior constraints according to claim 1, characterized in that Obtaining the incremental sequence and cumulative sequence of multi-source sensor monitoring data in the safe operation stage includes: At each acquisition moment, the sequence composed of the change values that occur at the current moment relative to the previous moment is used as the incremental sequence, denoted as X′ = [x′1, x′2,..., x′ n ; where X′ is the incremental sequence, and x′1, x′2,..., x′ n represent the increments at different moments, and the increment x′1 at the initial moment is set to 0; The sequence composed of the sensor values collected at each acquisition moment is used as the cumulative sequence, denoted as X = [x1, x2,..., x n ; where X is the cumulative sequence, and x1, x2,..., x n represent the sensor values at different moments, and the sensor value x1 at the initial moment is set to 0.

3. The method for abnormal edge recognition of multi-source sensors based on prior constraints according to claim 1, wherein Obtaining the real-time monitoring data of multi-source sensors and identifying abnormal points for the data at each moment, including: Based on the real-time monitoring data of multi-source sensors, using the fast Fourier transform to identify the first main frequency of the data; Taking the real-time monitoring data of the multi-source sensors as the main sequence, determining the division parameter according to the period of the first main frequency, using the division parameter to divide the main sequence, and if the final data length does not meet the division parameter, borrowing forward until a complete sequence is formed to obtain at least two subsequences; where the subsequence is expressed as: X k = [[x1,x2,…,x 2T ,[x 2T+1 ,x 2T+2 ,…,x 4T ,…,[x 2*kT-a+1 ,x 2*kT-a+2 ,...,x n ​ Wherein, X k represents a subsequence, T represents the first main frequency period, a represents the carry borrowed forward, x1, x2, x 2T respectively represent the sensor data at the initial moment, the second moment, and the 2T moment, x 2T+1 , x 2T+2 , x 4T respectively represent the sensor data at the 2T+1, 2T+2, and 4T moments, x 2*kT-a+1 , x 2*kT-a+2 , x n respectively represent the sensor data at the 2*kT-a+1, 2*kT-a+2, and n moments; Setting a judgment formula and determining the data points that meet the judgment formula in each subsequence as abnormal values.

4. The method for abnormal edge recognition of multi-source sensors based on prior constraints according to claim 3, characterized in that The division parameter is twice the period of the first main frequency.

5. The method for abnormal edge recognition of multi-source sensors based on prior constraints according to claim 3, characterized in that The judgment formula is expressed as: where ABS is the absolute value function, x i is the i-th value in the subsequence, is the average value of the subsequence, σ is the standard deviation in the subsequence, is the absolute value of the average slope in this subsequence.

6. The multi-source sensor abnormal edge recognition method based on prior constraints according to claim 1, characterized in that For the moments when abnormal points appear in both the cumulative sequence and the incremental sequence, using the prior constraint method to analyze the correlation between the sequences and determining the abnormal state, including: Selecting the moment when abnormal points first appear in all sensors to divide the data sequences of all sensors into prior cumulative stable sequences; X f1 = [X f1 = [x f11 , x f12 , …, x f1y , X f2 = [x f21 , x f22 , …, x f2y , …, x fm = [x fm1 , x fm2 , ..., x fmy ​ Where m is the number of multi-source sensors, X f1 is the prior cumulative stability sequence of the first sensor, X f2 is the prior cumulative stability sequence of the second sensor, X fm is the prior cumulative stability sequence of the m-th sensor, x f11 , x f12 , x f1y are the 1st, 2nd, and y-th values in the prior cumulative stability sequence of the first sensor, x 21 , x 22 , x 2y are the 1st, 2nd, and y-th values in the prior cumulative stability sequence of the second sensor, x fm1 , x fm2 , x fmy are the 1st, 2nd, and y-th values in the prior cumulative stability sequence of the m-th sensor. Using the Pearson correlation analysis method to analyze the correlation between the cumulative sequence and the incremental sequence data respectively to obtain the prior cumulative sequence correlation matrix and the incremental sequence correlation matrix; Using the Pearson correlation analysis method to analyze the original cumulative sequence and incremental sequence containing abnormal points respectively to obtain the cumulative sequence correlation matrix and the incremental sequence correlation matrix; Establishing an abnormal discriminant formula. For the abnormal data of the i-th sensor, if it meets the abnormal discriminant formula, it is determined that the monitoring system is in an abnormal state; if it does not meet the abnormal discriminant formula, it is determined that the monitoring system is in a normal state.

7. The method for abnormal edge recognition of multi-source sensors based on prior constraints according to claim 6, wherein The abnormal discriminant formula is expressed as: Where, A ij and A ij ' respectively represent the elements in the correlation matrix of the cumulative sequence and the correlation matrix of the incremental sequence; i and j are respectively different sensor numbers. When conducting the inspection of the i-th sensor, the value of i is fixed; A f,ij ' is an element in the correlation matrix of the incremental sequence of the prior stable sequence, A f,ij is an element in the correlation matrix of the cumulative sequence of the prior stable sequence.

8. The method for abnormal edge recognition of multi-source sensors based on prior constraints according to claim 6, wherein When it is determined that an abnormal state occurs, increasing the acquisition frequency of the sensors and determining the working state of the sensors based on the elimination and comparison method, including: Delete the data value at the abnormal point moment, and record the cumulative sequence of subsequent sampling data for each sensor and the incremental sequence After deleting the value at the abnormal moment, reconstruct the cumulative sequence and the incremental sequence starting from the monitoring. The reconstructed cumulative sequence and incremental sequence are respectively expressed as: X a = [x1, x2,..., x n , x n+2 ,..., x n+l ; X′ a = [x1′, x2′,..., x n ′, x n+2 ′,..., x n+l ′]; Based on the prior constraint method, posterior analysis is carried out to analyze the cumulative sequences of subsequent sampling data between different sensors and the incremental sequences to determine whether the state is normal; using the prior constraint method, prior analysis is carried out to analyze the cumulative sequence X a = [x1, x2,..., x n , x n+2 ,..., x n+1 and the incremental sequence X' a = [x1', x2',..., x n ', x n+2 ',..., x n+1 '], and determine whether the state is normal; If the posterior analysis determines normal and the prior analysis determines abnormal, indicating that the sensor has been disturbed, then re-zeroing is performed; if the posterior analysis determines abnormal and the prior analysis determines abnormal, indicating that the working state of the sensor is abnormal, then further manual confirmation is performed; if the posterior analysis determines normal and the prior analysis determines normal, it indicates that the acquisition process has been disturbed and the cause of the abnormal point is external interference.

9. An abnormal edge recognition device for multi-source sensors based on prior constraints, characterized in that, The device includes: A data acquisition module configured to obtain the incremental sequence and cumulative sequence of multi-source sensor monitoring data in the safe operation stage; An abnormal point identification module configured to obtain the real-time monitoring data of multi-source sensors and identify abnormal points for the data at each moment; An abnormal state recognition module, configured to, at the moment when abnormal points simultaneously appear in both the cumulative sequence and the incremental sequence, analyze the correlation between the sequences by using a priori constraint method and determine the abnormal state; A working state determination module, configured to, when it is determined that an abnormal state occurs, increase the acquisition frequency of the sensor and determine the working state of the sensor based on the elimination comparison method.

10. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, perform the method according to any one of claims 1 to 8.

Images