Anchor bolt stress gauge monitoring data processing method and system based on data analysis

Abstract

The present invention relates to the technical field of data processing, and discloses a method and system for processing anchor stress gauge monitoring data based on data analysis. First, the initial anchor stress gauge monitoring data set is collected, and a hybrid anomaly detection method is obtained based on a neural network and a dynamic threshold interval; secondly, the hybrid anomaly detection method is used to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set to obtain a processed anchor stress gauge monitoring data set; then, the missing data in the processed anchor stress gauge monitoring data set is encoded, and an error objective function is established; finally, the black kite optimization algorithm is used to find the global optimal solution, a data filling sequence is established, the missing data is filled, and the final anchor stress gauge monitoring data set is generated. The present invention achieves the purpose of anchor stress gauge monitoring data processing by processing and analyzing the anchor stress gauge monitoring data, and the method is accurate and objective.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for processing anchor stress gauge monitoring data based on data analysis. Background Art

[0002] Chinese patent CN116776258B discloses a method and system for processing monitoring data of electric power equipment. The method specifically includes: obtaining monitoring time series data of data attributes of electric power equipment, grouping the monitoring time series data according to equipment control parameters to obtain monitoring time series data grouping results; establishing a similarity matrix based on the monitoring time series data in the monitoring time series data grouping results, and then performing clustering processing to obtain several clustering results; performing outlier division on the several clustering results to obtain abnormal data division results; obtaining missing values ​​of monitoring time series data, combining the abnormal data division results and the missing values ​​of monitoring time series data, using the cluster centroid mean of the clustering result to replace them, obtaining monitoring time series data optimization results, and transmitting them to the optimization terminal to complete the processing of electric power equipment monitoring data. However, this invention uses a clustering algorithm to remove outliers, and the processing method is traditional and has poor accuracy.

[0003] Traditional data processing methods contain a lot of noise in the raw data, and it is usually difficult to effectively separate the noise using traditional data processing methods. At the same time, manual experience is used to judge abnormal values ​​in the anchor stress gauge monitoring data, which is inefficient and prone to missed detection. It cannot meet the engineering requirements of high precision and low latency, and does not use technologies such as artificial intelligence. The automation level and reliability of the monitoring system are low, making it unsuitable for long-term health monitoring in complex environments. Summary of the Invention

[0004] In response to the problems in the related art, the present invention provides a method and system for processing anchor stress gauge monitoring data based on data analysis to overcome the above-mentioned technical problems existing in the existing related art.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a method for processing anchor stress gauge monitoring data based on data analysis, comprising the following steps:

[0007] S1. Collect anchor stress gauge monitoring data to obtain an initial anchor stress gauge monitoring data set, upload it to the smart operation and maintenance platform, establish a neural network model, and combine it with the dynamic threshold interval to obtain a hybrid anomaly detection method;

[0008] S2. Performing hybrid anomaly detection on the initial anchor stress gauge monitoring data set based on the abnormal residual data and abnormal interval data output by the hybrid anomaly detection method to obtain a processed anchor stress gauge monitoring data set;

[0009] S3, encoding the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set, and establishing an error objective function based on the data error;

[0010] S4. According to the error objective function, use the Black Kite optimization algorithm to find the global optimal solution, establish a data filling sequence, fill the missing data in the complete anchor stress gauge monitoring data set, generate the final anchor stress gauge monitoring data set, and complete the anchor stress gauge monitoring data processing.

[0011] This invention obtains the initial anchor stress gauge monitoring data set, obtains a hybrid anomaly detection method based on the LSTM neural network and dynamic threshold interval, uses the neural network to predict the detection data to ensure accurate detection capabilities, and then combines the dynamic threshold interval to mark the candidate anomaly data, adapts to the non-stationary nature of the data distribution, has good applicability, and facilitates subsequent detection of abnormal data; secondly, the hybrid anomaly detection method is used to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set to obtain the processed anchor stress gauge monitoring data set; combining the two methods increases the accuracy of abnormal data detection while ensuring that key data is avoided from being accidentally deleted, providing high-quality data for subsequent Data foundation; then encode the missing data in the processed anchor stress gauge monitoring data set, establish a mapping between individual and filled data, and establish an objective function based on data error, use the black kite optimization algorithm to find the global optimal solution, establish a data filling sequence, fill the missing data in the complete anchor stress gauge monitoring data set, and generate the final anchor stress gauge monitoring data set; the black kite optimization algorithm imitates the survival strategy of the black kite to iteratively optimize the monitoring data set. It has the characteristics of strong evolutionary ability, fast search speed and strong optimization ability, which can quickly complete the filling of missing data, improve the performance of the data mining algorithm, and improve the level of automated monitoring.

[0012] Preferably, the S1 comprises the following steps:

[0013] S11. During the construction process, anchor stress gauges are embedded in the anchor concrete to monitor the stress state of the anchor, including tensile stress and compressive stress. Anchor stress gauge monitoring data is collected to form an initial anchor stress gauge monitoring data set, and the initial anchor stress gauge monitoring data set is uploaded to the smart operation and maintenance platform.

[0014] S12. Re-acquire the anchor stress gauge historical monitoring data set, obtain the time points of collecting the anchor stress gauge historical monitoring data, form a historical monitoring data time series, train the LSTM neural network, and obtain the LSTM neural network prediction model. The specific steps are as follows:

[0015] S121, the LSTM neural network includes a forget gate, an input gate, and an output gate, and sets an activation function and a learning rate; the anchor stress gauge historical monitoring data set is divided according to the time series of the historical monitoring data to obtain a sample training set and a sample test set, respectively, a sliding window is set, the sliding window is placed on the sample training set, the anchor stress gauge historical monitoring data in the sliding window is sequentially input into the LSTM neural network, the LSTM neural network is started and trained until the LSTM neural network converges, and a trained LSTM neural network is obtained;

[0016] S122, then placing the sliding window on the sample test set, sequentially inputting the historical monitoring data of the anchor stress gauge in the sliding window into the trained LSTM neural network, and outputting the prediction result; setting an accuracy threshold, when the accuracy of the prediction result is less than the accuracy threshold, stopping the iteration to obtain the LSTM neural network prediction model, otherwise adjusting the weight until the accuracy of the prediction result is less than the accuracy threshold;

[0017] S13, obtaining the time points of collecting anchor stress gauge monitoring data, forming a monitoring data time series, setting a time series window in the monitoring data time series, calculating the mean and standard deviation of the anchor stress gauge monitoring data corresponding to the time series window; setting a dynamic threshold interval, taking the difference between the mean of the anchor stress gauge monitoring data and three times the standard deviation of the anchor stress gauge monitoring data as the lower bound of the dynamic threshold interval, taking the sum of the mean of the anchor stress gauge monitoring data and three times the standard deviation of the anchor stress gauge monitoring data as the upper bound of the dynamic threshold interval, recording the anchor stress gauge monitoring data corresponding to the time series window that exceeds the dynamic threshold interval as abnormal interval data, otherwise it is recorded as normal data; traversing the monitoring data time series in sequence until all anchor stress gauge monitoring data are detected, and obtaining an abnormal interval data set;

[0018] S14. Combining the LSTM neural network prediction model and the dynamic threshold interval, a hybrid anomaly detection method is obtained.

[0019] This invention obtains the initial anchor stress gauge monitoring data set, obtains a hybrid anomaly detection method based on the LSTM neural network and dynamic threshold interval, uses the neural network to predict the detection data to ensure accurate detection capabilities, and then combines the dynamic threshold interval to mark candidate abnormal data, adapting to the non-stationary nature of the data distribution. It has good applicability and facilitates subsequent detection of abnormal data.

[0020] Preferably, said S2 comprises the following steps:

[0021] S21. For the LSTM neural network prediction model in the hybrid anomaly detection method, the initial anchor stress gauge monitoring data set is sequentially input into the LSTM neural network prediction model according to the monitoring data time series, the residuals between the time points of collecting anchor stress gauge monitoring data in the monitoring data time series are calculated, a residual threshold is set, and the anchor stress gauge monitoring data corresponding to the residual greater than the residual threshold is recorded as a candidate abnormal residual data set; and the abnormal residual data in the candidate abnormal residual data set is found using a box plot. The specific steps are as follows:

[0022] S211, sort the candidate abnormal residual data set in ascending order to obtain the minimum value, maximum value and median of the candidate abnormal residual data set, find the lower quartile between the minimum value and the median, and find the upper quartile between the maximum value and the median; calculate the abnormal data interval based on the lower quartile and the upper quartile, and the upper bound of the abnormal data interval is , the lower bound of the abnormal data interval is , where A represents the upper quartile and B represents the lower quartile;

[0023] S212, comparing the abnormal data interval with the candidate abnormal residual data set, searching for candidate abnormal residual data between the upper bound of the abnormal data interval and the lower bound of the abnormal data interval in the candidate abnormal residual data set, and obtaining abnormal residual data;

[0024] S22. Perform mixed anomaly detection on the initial anchor stress gauge monitoring data set, compare the abnormal residual data and the abnormal interval data, select the overlapping data of the abnormal residual data and the abnormal interval data, record them as anchor stress gauge monitoring abnormal data, and delete the anchor stress gauge monitoring abnormal data to obtain the processed anchor stress gauge monitoring data set.

[0025] This invention uses a hybrid anomaly detection method to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set. Combining the two methods, while ensuring that key data is avoided from being accidentally deleted, it increases the accuracy of abnormal data detection and provides a high-quality data foundation for the subsequent development.

[0026] Preferably, the step S3 includes the following steps:

[0027] S31, setting the processed anchor stress gauge monitoring data set ,in Represents the i-th anchor stress gauge monitoring data with dimension j, sets the number of missing data in the processed anchor stress gauge monitoring data set to h, and adds a flag bit to the processed anchor stress gauge monitoring data set , when the flag bit is equal to 0, it means that there is no missing data in the k dimension, and when the flag bit is equal to 1, it means that there is missing data in the k dimension, and the encoding is performed, and the flag bit is inserted into the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set;

[0028] S32. Fill the missing data in the complete anchor stress gauge monitoring data set, randomly select missing data, record them as missing data to be filled, calculate the local mean square error between the missing data to be filled and the adjacent data; use linear regression to fit the slope of the first m anchor stress gauge monitoring data of the missing data to be filled and the slope of the missing data to be filled, and calculate the slope error; assign weights to the local mean square error and the slope error, and establish an error objective function based on the data error as follows:

[0029] ;

[0030] Where F represents the error objective function, and represents the weight, represents the local mean square error, Represents the slope error.

[0031] Preferably, said S4 comprises the following steps:

[0032] S41, using the error objective function as a fitness function, inputting a complete anchor stress gauge monitoring data set, treating the missing data filling process as a search space, using the Black Kite optimization algorithm to find the global optimal solution according to the error objective function, and establishing a data filling sequence, the specific steps are as follows:

[0033] S411. Assume that there is a black kite population in the search space, the dimension of the black kite population is j, the number of black kite population is n, and the black kite individuals in the black kite population represent candidate solutions, wherein the candidate solutions include local mean square error and slope error; initialize the black kite population, set the current best fitness function value selected from the black kite population to correspond to the best position of the black kite individual; the black kite population enters the attack phase, selects the mth black kite individual from the black kite population, sets the current number of iterations to t, the maximum number of iterations to T, and the position of the mth black kite individual in the j dimension at the tth iteration is , Indicates that the interval Random number, attack coefficient ; At this time, the position of the black kite is updated. At the t+1th iteration, the position of the mth black-winged kite individual in the jth dimension is , otherwise you get ;

[0034] S412, the black kite population enters the migration phase, and finds the black kite individual corresponding to the best fitness function in the tth iteration, which is recorded as the black kite leader. The position of the black kite leader in the jth dimension is ,set up represents the Cauchy variation, Indicates that the interval Random number, migration coefficient , select a random black-winged kite individual from the black-winged kite population, calculate the fitness function values ​​of the random black-winged kite individual position and the mth black-winged kite individual position, respectively recorded as and , update the black kite position again, when When ,when When ; At this time, all stages are completed, and the initial data filling sequence is generated by the t-th iteration of the black kite individual. Iteration continues until the current number of iterations reaches the maximum number of iterations, then the iteration is stopped, the global optimal solution is obtained, and the data filling sequence is generated;

[0035] S42: Use the data to fill in the missing data in the sequence synchronized complete anchor stress gauge monitoring data set, generate a final anchor stress gauge monitoring data set, and complete the anchor stress gauge monitoring data processing.

[0036] This invention uses the Black Kite optimization algorithm to find the global optimal solution, establish a data filling sequence, and fill in the missing data in the complete anchor stress gauge monitoring data set. The Black Kite optimization algorithm imitates the survival strategy of the black kite to iteratively optimize the monitoring data set. It has the characteristics of strong evolutionary ability, fast search speed and strong optimization ability. It can quickly complete the filling of missing data, improve the performance of the data mining algorithm, improve the level of automated monitoring, and realize the processing of anchor stress gauge monitoring data.

[0037] The present invention also discloses a system for processing monitoring data of an anchor stress gauge based on data analysis, which specifically includes: a monitoring data processing preparation module, a monitoring data anomaly detection module, a missing data encoding module and a missing data filling module;

[0038] The monitoring data processing preparation module is used to combine the neural network model and the state threshold interval to obtain a hybrid anomaly detection method;

[0039] The monitoring data anomaly detection module is used to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set using a hybrid anomaly detection method;

[0040] The missing data encoding module is used to encode missing data and establish an error objective function based on data error;

[0041] The missing data filling module is used to use the Black Kite optimization algorithm to find a data filling sequence and then fill in the missing data.

[0042] The present invention has the following beneficial effects:

[0043] 1. This invention obtains the initial anchor stress gauge monitoring data set, obtains a hybrid anomaly detection method based on the LSTM neural network and dynamic threshold interval, uses the neural network to predict the detection data to ensure accurate detection capabilities, and then combines the dynamic threshold interval to mark candidate abnormal data, adapting to the non-stationary nature of data distribution. It has good applicability and facilitates subsequent detection of abnormal data.

[0044] 2. This invention uses a hybrid anomaly detection method to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set. Combining the two methods, while ensuring that key data is avoided from being accidentally deleted, it increases the accuracy of abnormal data detection and provides a high-quality data foundation for the subsequent development.

[0045] 3. The invention uses the black kite optimization algorithm to find the global optimal solution, establish a data filling sequence, and fill in the missing data in the complete anchor stress gauge monitoring data set. The black kite optimization algorithm imitates the survival strategy of the black kite to iteratively optimize the monitoring data set. It has the characteristics of strong evolutionary ability, fast search speed and strong optimization ability. It can quickly complete the filling of missing data, improve the performance of the data mining algorithm, improve the level of automated monitoring, and realize the processing of anchor stress gauge monitoring data.

[0046] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying any creative work.

[0048] Figure 1 A schematic diagram of the process of anchor stress gauge monitoring data processing in an anchor stress gauge monitoring data processing system based on data analysis provided by the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0051] Example 1

[0052] Please refer to Figure 1 The present invention is a method for processing anchor stress gauge monitoring data based on data analysis, comprising the following steps:

[0053] S1. Collect anchor stress gauge monitoring data to obtain an initial anchor stress gauge monitoring data set, upload it to the smart operation and maintenance platform, establish a neural network model, and combine it with the dynamic threshold interval to obtain a hybrid anomaly detection method;

[0054] Said S1 comprises the following steps:

[0055] S11. During the construction process, anchor stress gauges are embedded in the anchor concrete to monitor the stress state of the anchor, including tensile stress and compressive stress. Anchor stress gauge monitoring data is collected to form an initial anchor stress gauge monitoring data set, and the initial anchor stress gauge monitoring data set is uploaded to the smart operation and maintenance platform.

[0056] S12. Re-acquire the anchor stress gauge historical monitoring data set, obtain the time points of collecting the anchor stress gauge historical monitoring data, form a historical monitoring data time series, train the LSTM neural network, and obtain the LSTM neural network prediction model. The specific steps are as follows:

[0057] S121, the LSTM neural network includes a forget gate, an input gate, and an output gate, and sets an activation function and a learning rate; the anchor stress gauge historical monitoring data set is divided according to the time series of the historical monitoring data to obtain a sample training set and a sample test set, respectively, a sliding window is set, the sliding window is placed on the sample training set, the anchor stress gauge historical monitoring data in the sliding window is sequentially input into the LSTM neural network, the LSTM neural network is started and trained until the LSTM neural network converges, and a trained LSTM neural network is obtained;

[0058] S122, then placing the sliding window on the sample test set, sequentially inputting the historical monitoring data of the anchor stress gauge in the sliding window into the trained LSTM neural network, and outputting the prediction result; setting an accuracy threshold, when the accuracy of the prediction result is less than the accuracy threshold, stopping the iteration to obtain the LSTM neural network prediction model, otherwise adjusting the weight until the accuracy of the prediction result is less than the accuracy threshold;

[0059] S13, obtaining the time points of collecting anchor stress gauge monitoring data, forming a monitoring data time series, setting a time series window in the monitoring data time series, calculating the mean and standard deviation of the anchor stress gauge monitoring data corresponding to the time series window; setting a dynamic threshold interval, taking the difference between the mean of the anchor stress gauge monitoring data and three times the standard deviation of the anchor stress gauge monitoring data as the lower bound of the dynamic threshold interval, taking the sum of the mean of the anchor stress gauge monitoring data and three times the standard deviation of the anchor stress gauge monitoring data as the upper bound of the dynamic threshold interval, recording the anchor stress gauge monitoring data corresponding to the time series window that exceeds the dynamic threshold interval as abnormal interval data, otherwise it is recorded as normal data; traversing the monitoring data time series in sequence until all anchor stress gauge monitoring data are detected, and obtaining an abnormal interval data set;

[0060] S14. Combining the LSTM neural network prediction model and the dynamic threshold interval to obtain a hybrid anomaly detection method;

[0061] S2. Performing hybrid anomaly detection on the initial anchor stress gauge monitoring data set based on the abnormal residual data and abnormal interval data output by the hybrid anomaly detection method to obtain a processed anchor stress gauge monitoring data set;

[0062] The S2 comprises the following steps:

[0063] S21. For the LSTM neural network prediction model in the hybrid anomaly detection method, the initial anchor stress gauge monitoring data set is sequentially input into the LSTM neural network prediction model according to the monitoring data time series, the residuals between the time points of collecting anchor stress gauge monitoring data in the monitoring data time series are calculated, a residual threshold is set, and the anchor stress gauge monitoring data corresponding to the residual greater than the residual threshold is recorded as a candidate abnormal residual data set; and the abnormal residual data in the candidate abnormal residual data set is found using a box plot. The specific steps are as follows:

[0064] S211, sort the candidate abnormal residual data set in ascending order to obtain the minimum value, maximum value and median of the candidate abnormal residual data set, find the lower quartile between the minimum value and the median, and find the upper quartile between the maximum value and the median; calculate the abnormal data interval based on the lower quartile and the upper quartile, and the upper bound of the abnormal data interval is , the lower bound of the abnormal data interval is , where A represents the upper quartile and B represents the lower quartile;

[0065] S212, comparing the abnormal data interval with the candidate abnormal residual data set, searching for candidate abnormal residual data between the upper bound of the abnormal data interval and the lower bound of the abnormal data interval in the candidate abnormal residual data set, and obtaining abnormal residual data;

[0066] S22, performing mixed anomaly detection on the initial anchor stress gauge monitoring data set, comparing the abnormal residual data and the abnormal interval data, selecting the overlapping data of the abnormal residual data and the abnormal interval data, recording them as anchor stress gauge monitoring abnormal data, and deleting the anchor stress gauge monitoring abnormal data to obtain a processed anchor stress gauge monitoring data set;

[0067] S3, encoding the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set, and establishing an error objective function based on the data error;

[0068] The S3 includes the following steps:

[0069] S31, setting the processed anchor stress gauge monitoring data set ,in Represents the i-th anchor stress gauge monitoring data with dimension j, sets the number of missing data in the processed anchor stress gauge monitoring data set to h, and adds a flag bit to the processed anchor stress gauge monitoring data set , when the flag bit is equal to 0, it means that there is no missing data in the k dimension, and when the flag bit is equal to 1, it means that there is missing data in the k dimension, and the encoding is performed, and the flag bit is inserted into the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set;

[0070] S32. Fill the missing data in the complete anchor stress gauge monitoring data set, randomly select missing data, record them as missing data to be filled, calculate the local mean square error between the missing data to be filled and the adjacent data; use linear regression to fit the slope of the first m anchor stress gauge monitoring data of the missing data to be filled and the slope of the missing data to be filled, and calculate the slope error; assign weights to the local mean square error and the slope error, and establish an error objective function based on the data error as follows:

[0071] ;

[0072] Where F represents the error objective function, and represents the weight, represents the local mean square error, represents the slope error;

[0073] S4. According to the error objective function, use the Black Kite optimization algorithm to find the global optimal solution, establish a data filling sequence, fill the missing data in the complete anchor stress gauge monitoring data set, generate the final anchor stress gauge monitoring data set, and complete the anchor stress gauge monitoring data processing;

[0074] The S4 comprises the following steps:

[0075] S41, using the error objective function as a fitness function, inputting a complete anchor stress gauge monitoring data set, treating the missing data filling process as a search space, using the Black Kite optimization algorithm to find the global optimal solution according to the error objective function, and establishing a data filling sequence, the specific steps are as follows:

[0076] S411. Assume that there is a black kite population in the search space, the dimension of the black kite population is j, the number of black kite population is n, and the black kite individuals in the black kite population represent candidate solutions, wherein the candidate solutions include local mean square error and slope error; initialize the black kite population, set the current best fitness function value selected from the black kite population to correspond to the best position of the black kite individual; the black kite population enters the attack phase, selects the mth black kite individual from the black kite population, sets the current number of iterations to t, the maximum number of iterations to T, and the position of the mth black kite individual in the j dimension at the tth iteration is , Indicates that the interval Random number, attack coefficient ; At this time, the position of the black kite is updated. At the t+1th iteration, the position of the mth black-winged kite individual in the jth dimension is , otherwise you get ;

[0077] S412, the black kite population enters the migration phase, and finds the black kite individual corresponding to the best fitness function in the tth iteration, which is recorded as the black kite leader. The position of the black kite leader in the jth dimension is ,set up represents the Cauchy variation, Indicates that the interval Random number, migration coefficient , select a random black-winged kite individual from the black-winged kite population, calculate the fitness function values ​​of the random black-winged kite individual position and the mth black-winged kite individual position, respectively recorded as and , update the black kite position again, when When ,when When ; At this time, all stages are completed, and the initial data filling sequence is generated by the t-th iteration of the black kite individual. Iteration continues until the current number of iterations reaches the maximum number of iterations, then the iteration is stopped, the global optimal solution is obtained, and the data filling sequence is generated;

[0078] S42: Use the data to fill in the missing data in the sequence synchronized complete anchor stress gauge monitoring data set, generate a final anchor stress gauge monitoring data set, and complete the anchor stress gauge monitoring data processing.

[0079] Example 2

[0080] The present invention also discloses a system for processing monitoring data of an anchor stress gauge based on data analysis, which specifically includes: a monitoring data processing preparation module, a monitoring data anomaly detection module, a missing data encoding module and a missing data filling module;

[0081] The monitoring data processing preparation module is used to combine the neural network model and the state threshold interval to obtain a hybrid anomaly detection method;

[0082] The monitoring data anomaly detection module is used to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set using a hybrid anomaly detection method;

[0083] The missing data encoding module is used to encode missing data and establish an error objective function based on data error;

[0084] The missing data filling module is used to use the Black Kite optimization algorithm to find a data filling sequence and then fill in the missing data.

[0085] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0086] The preferred embodiments of the invention disclosed above are intended only to help illustrate the invention. These preferred embodiments do not exhaust all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for processing anchor stress gauge monitoring data based on data analysis, characterized in that: The steps include: S1. Collect anchor stress gauge monitoring data to obtain an initial anchor stress gauge monitoring data set, upload it to the smart operation and maintenance platform, establish a neural network model, and combine it with the dynamic threshold interval to obtain a hybrid anomaly detection method; S2. Performing hybrid anomaly detection on the initial anchor stress gauge monitoring data set based on the abnormal residual data and abnormal interval data output by the hybrid anomaly detection method to obtain a processed anchor stress gauge monitoring data set; S3, encoding the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set, and establishing an error objective function based on the data error; S4. Finding a global optimal solution according to the error objective function, establishing a data filling sequence, filling in the missing data in the complete anchor stress gauge monitoring data set, generating a final anchor stress gauge monitoring data set, and completing anchor stress gauge monitoring data processing; S2 specifically includes: inputting the initial anchor stress gauge monitoring data set into the LSTM neural network prediction model, outputting the residual, setting the residual threshold, and recording the anchor stress gauge monitoring data corresponding to the residual greater than the residual threshold as a candidate abnormal residual data set; using a box plot to find the abnormal residual data in the candidate abnormal residual data set; selecting the overlapping data of the abnormal residual data and the abnormal interval data, recording it as the anchor stress gauge monitoring abnormal data, and deleting the anchor stress gauge monitoring abnormal data to obtain the processed anchor stress gauge monitoring data set; S3 specifically includes: setting the number of missing data in the processed anchor stress gauge monitoring data set to h, adding a flag bit h to the processed anchor stress gauge monitoring data set k , when the flag bit is equal to 0, it indicates that there is no missing data in the k dimension, and when the flag bit is equal to 1, it indicates that there is missing data in the k dimension, and the encoding is performed, and the flag bit is inserted into the missing data in the processed anchor stress gauge monitoring data set to obtain a complete anchor stress gauge monitoring data set; the missing data in the complete anchor stress gauge monitoring data set are filled, the missing data to be filled are selected, the local mean square error and slope error of the missing data to be filled are calculated, the local mean square error and slope error are assigned weights, and the error objective function is established based on the data error as follows: Among them, F represents the error objective function, σ1 and σ2 represent weights, represents the local mean square error, represents the slope error; S4 specifically includes: S41, using the error objective function as a fitness function, inputting a complete anchor stress gauge monitoring data set, treating the missing data filling process as a search space, using the Black Kite optimization algorithm to find a global optimal solution based on the error objective function, and establishing a data filling sequence; S42: Use the data to fill in the missing data in the sequence synchronized complete anchor stress gauge monitoring data set, generate a final anchor stress gauge monitoring data set, and complete the anchor stress gauge monitoring data processing.

2. The anchor stress gauge monitoring data processing method based on data analysis according to claim 1 is characterized in that: The S1 comprises the following steps: S11. During the construction process, anchor stress gauge monitoring data is collected to form an initial anchor stress gauge monitoring data set, and the initial anchor stress gauge monitoring data set is uploaded to the smart operation and maintenance platform; S12, re-acquire the anchor stress gauge historical monitoring data set, obtain the time points of collecting the anchor stress gauge historical monitoring data, form a historical monitoring data time series, train the LSTM neural network, and obtain the LSTM neural network prediction model; S13, obtaining the time points of collecting anchor stress gauge monitoring data, forming a monitoring data time series, and then setting a dynamic threshold interval, recording the anchor stress gauge monitoring data corresponding to the dynamic threshold interval as abnormal interval data, otherwise recording it as normal data, and obtaining an abnormal interval data set; S14. Combining the LSTM neural network prediction model and the dynamic threshold interval, a hybrid anomaly detection method is obtained.

3. The anchor stress gauge monitoring data processing method based on data analysis according to claim 2 is characterized in that: The S12 includes the following steps: S121, the LSTM neural network includes a forget gate, an input gate, and an output gate, and sets an activation function and a learning rate; the anchor stress gauge historical monitoring data set is divided according to the time series of the historical monitoring data to obtain a sample training set and a sample test set, respectively, a sliding window is set, the sliding window is placed on the sample training set, the anchor stress gauge historical monitoring data in the sliding window is sequentially input into the LSTM neural network, the LSTM neural network is started and trained until the LSTM neural network converges, and a trained LSTM neural network is obtained; S122, then placing the sliding window on the sample test set, inputting the historical monitoring data of the anchor stress gauge in the sliding window into the trained LSTM neural network in sequence, and outputting the prediction results; Set an accuracy threshold. When the accuracy of the prediction result is less than the accuracy threshold, stop the iteration and obtain the LSTM neural network prediction model. Otherwise, adjust the weight until the accuracy of the prediction result is less than the accuracy threshold.

4. The anchor stress gauge monitoring data processing method based on data analysis according to claim 1 is characterized in that: The method of using a box plot to find abnormal residual data in the candidate abnormal residual data set comprises the following steps: Sort the candidate abnormal residual data set in ascending order to obtain the lower quartile and upper quartile of the candidate abnormal residual data set, and calculate the abnormal data interval according to the lower quartile and upper quartile; The candidate abnormal residual data between the upper bound of the abnormal data interval and the lower bound of the abnormal data interval are found in the candidate abnormal residual data set to obtain the abnormal residual data.

5. The anchor stress gauge monitoring data processing method based on data analysis according to claim 1 is characterized in that: The S41 includes the following steps: S411. Assume that there is a black kite population in the search space, the dimension of the black kite population is j, the number of black kite population is n, and the black kite individuals in the black kite population represent candidate solutions, wherein the candidate solutions include local mean square error and slope error; initialize the black kite population, set the current best fitness function value selected from the black kite population to correspond to the best position of the black kite individual; the black kite population enters the attack phase, selects the mth black kite individual from the black kite population, sets the current number of iterations to t, the maximum number of iterations to T, and the position of the mth black kite individual in the j dimension at the tth iteration is d1 represents a random number between the interval [0,1], the attack coefficient At this time, the position of the black-winged kite is updated. When d1>0.9, the position of the m-th black-winged kite in the j-th dimension in the t+1th iteration is Otherwise you get S412, the black kite population enters the migration phase, and finds the black kite individual corresponding to the best fitness function in the tth iteration, which is recorded as the black kite leader. The position of the black kite leader in the jth dimension is Set β(0,1) to represent Cauchy mutation, d2 to represent a random number between the interval [0,1], and the migration coefficient Select a random black-winged kite individual from the black-winged kite population, calculate the fitness function values ​​of the random black-winged kite individual position and the mth black-winged kite individual position, record them as F′ and F″ respectively, update the black-winged kite position again, and when F′>F″, we get When F′≤F″, we get At this point, all stages are completed, and the tth iteration of the black kite individual generates the initial data filling sequence. Iterate continuously until the current number of iterations reaches the maximum number of iterations, then stop iterating, obtain the global optimal solution, and generate the data filling sequence.

6. A system for implementing the anchor stress gauge monitoring data processing method based on data analysis as described in any one of claims 1 to 5, characterized in that: Specifically include: Monitoring data processing preparation module, monitoring data anomaly detection module, missing data encoding module and missing data filling module; The monitoring data processing preparation module is used to combine the neural network model and the state threshold interval to obtain a hybrid anomaly detection method; The monitoring data anomaly detection module is used to perform hybrid anomaly detection on the initial anchor stress gauge monitoring data set using a hybrid anomaly detection method; The missing data encoding module is used to encode missing data and establish an error objective function based on data error; The missing data filling module is used to use the Black Kite optimization algorithm to find a data filling sequence and then fill in the missing data.

Images