A method and system for online monitoring of mine filling deformation

Abstract

The present invention discloses an online monitoring method and system for the deformation of mine filling bodies, which specifically relates to the field of filling body deformation monitoring, and is used to solve the problem of insufficient comprehensive impact assessment of sensor loss. It is achieved by constructing a correlation model of multi-sensor data and using statistical inference technology to perform implicit pattern recognition on the observed multi-sensor data to identify sensors in abnormal states. A data interpolation and impact analysis model is adopted, based on the mutual interaction of multi-sensor data, to evaluate the comprehensive impact of sensor loss on the monitoring system from a global perspective, and provide a more comprehensive fault analysis. Finally, by arranging the sensor data in chronological order, marking the location of missing data, and using highly correlated neighbor sensors to repair data, the degree of fault diffusion and data repair effect are comprehensively considered, and timely warning information is provided to the monitoring system, thereby effectively improving the stability and reliability of the monitoring system and ensuring the safe operation of mine filling bodies.

Description

Technical Field

[0001] The present invention relates to the field of filling body deformation monitoring, and more specifically, to an online monitoring method and system for mine filling body deformation. Background Art

[0002] Mine filling refers to an artificial filling formed by filling the mine cavity with waste or ore generated during the mining process after processing. The constituent materials of the filling usually include ore crushed stone, slag, tailings, ore powder, cement, etc. The importance of monitoring mine filling is mainly reflected in the following aspects: First, the stability of the filling is directly related to the surrounding environment and social security of the mine. Therefore, it is necessary to regularly monitor the deformation of the filling, detect abnormal conditions in time and take corresponding measures. Secondly, monitoring the filling can help evaluate its engineering effect and durability, and provide a scientific basis for subsequent filling management and maintenance. In addition, mine filling is usually a comprehensive utilization of mine resources. Monitoring can ensure the stable operation of the filling and maximize its resource utilization benefits. Therefore, monitoring mine filling is one of the important means to ensure mine safety, environmental protection and sustainable development.

[0003] There are several problems with the current online monitoring of mine filling body deformation. First, the failure of sensors to be identified in a timely manner and the lack of effective identification and processing mechanisms for abnormal state sensors may affect the accuracy of monitoring data. In addition, the comprehensive impact assessment of sensor loss is insufficient, and there is a lack of comprehensive analysis from a global perspective. It is difficult to accurately assess the impact of sensor loss and active data repair on the monitoring system, which can easily lead to safety hazards caused by untimely maintenance and waste of resources caused by too frequent maintenance. These problems affect the stability and reliability of the monitoring system.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an online monitoring method and system for the deformation of mine filling bodies. By constructing a correlation model of multi-sensor data and using statistical inference technology to perform implicit pattern recognition on the observed multi-sensor data, it is possible to effectively identify sensors in abnormal states. Next, a data interpolation and impact analysis model is used, based on the mutual interaction of multi-sensor data, to evaluate the comprehensive impact of sensor loss on the monitoring system from a global perspective, providing a more comprehensive fault analysis. Finally, by arranging the sensor data in chronological order, marking the location of missing data, and using highly correlated neighboring sensors to repair data, the degree of fault diffusion and the effect of data repair are comprehensively considered, providing timely warning information for the monitoring system, thereby effectively improving the stability and reliability of the monitoring system, and ensuring the safe operation of the mine filling body, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: S1. construct a mine filling structure model, arrange installation base points on the sensor installation surface according to the rated space distance, and deploy actual sensors according to the installation base points;

[0007] S2. By constructing a correlation model of multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify sensors in abnormal states;

[0008] S3. Using data interpolation and impact analysis models, based on multi-sensor data interaction, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective;

[0009] S4. Arrange the sensor data in chronological order, mark the location of missing data, find neighboring sensors with high correlation with the missing sensor, and use the constructed completion model to repair the data. Consider the degree of fault diffusion and data repair effect comprehensively, and provide corresponding warning information to the monitoring system.

[0010] In a preferred embodiment, by constructing a correlation model of multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify abnormal state sensors, including the following contents:

[0011] Step 1: Use the Bayesian network to build a correlation model between sensors, where each node represents a sensor variable and the edge represents the dependency relationship between variables;

[0012] Step 2: estimate the conditional rerouting distribution of each node, which includes the conditional probability of the node given its parent node;

[0013] Based on historical data, the number of occurrences of parent nodes and sensor variables is counted, and then the conditional probability distribution is estimated using the kernel density estimation method: ;

[0014] in: Indicates that at a given parent node In the case of The conditional probability distribution of is the target that needs to be estimated;

[0015] Indicates the parent node in the historical data The number of occurrences of

[0016] Represents sensor variables in historical data and its parent node The number of joint occurrences of is the sample size, which indicates the number of samples in the historical data;

[0017] is the bandwidth parameter, which is used to control the width of the kernel function;

[0018] is the kernel function, used to weight the samples;

[0019] Represents the sensor variable In the observations in samples or time points;

[0020] It is used for normalization to ensure that the estimated probability density function integrates to 1.

[0021] In a preferred embodiment, in step three, the posterior probability of the target sensor is calculated using Bayesian theorem, that is, the state probability of the target sensor given other sensor data: ;

[0022] In the formula Indicates the target sensor The posterior probability is the probability of the sensor state to be determined after observing the actual data;

[0023] Indicates that the sensor The status is The probability of observing the actual data under the condition of , that is, the possibility of the sensor giving the observed data;

[0024] The sensor The prior probability of is the initial estimate of the sensor state before any data is observed;

[0025] Step 4: compare the posterior probability of the target sensor with the set threshold; if the posterior probability of the target sensor is greater than or equal to the set threshold, it is considered that the target sensor is in a normal state and a normal state signal is generated;

[0026] If the posterior probability of the target sensor is less than the set threshold, an abnormal situation exists and a signal prompt of the abnormal state is generated.

[0027] In a preferred embodiment, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective based on multi-sensor data interaction using data interpolation and impact analysis models, including the following:

[0028] For sensors that generate abnormal status signals, they are marked as missing sensors;

[0029] When the number of missing sensors is greater than a proportional threshold, urgent maintenance action is required.

[0030] In a preferred embodiment, when the number is less than or equal to the quantity threshold, the fault diffusion index is calculated using the following calculation method:

[0031] Represent the sensor matrix as a graph ,in is a set of sensor nodes, is a set of edges, representing the connection relationship between sensors;

[0032] Define the fault propagation path matrix ,in Represents a slave node To Node The propagation path length;

[0033] Use the shortest path algorithm in graph theory to calculate the shortest path length matrix between sensor nodes ,in Represents a slave node To Node The shortest path length;

[0034] For missing sensor nodes , calculate its influence range matrix ,in Representation Node For Node The degree of influence, Defined as: ;

[0035] The formula represents the missing sensor node For Node The degree of influence and node To Node The shortest path length and nodes To Node The shortest path length is inversely proportional to the reciprocal of the product of and is the weight parameter;

[0036] When the number of missing sensors is greater than or equal to the threshold, the sum of the rows and columns of each impact range matrix represents the degree of impact on each node and the degree of impact caused, respectively. These two vectors are defined as and ;

[0037] The fault diffusion index can be defined as , represents the degree of sensor data anomaly caused by node failure, and is calculated by the following formula: .

[0038] In a preferred embodiment, the sensor data is arranged in time order, the missing data position is marked, the neighboring sensors with high correlation with the missing sensor are found, and the data is repaired using the constructed completion model, including the following:

[0039] Step 1: Arrange the data of each sensor in the monitoring system in chronological order to form a sensor data matrix D, where the rows represent time steps and the columns represent sensor numbers;

[0040] Step 2: Mark the location of missing sensor data in the sensor data matrix, that is, determine the row and column indexes of the missing data;

[0041] Step 3: For each missing sensor, find a sensor with a high correlation based on the data of the surrounding sensors;

[0042] The operation process of finding sensors with high correlation with the missing sensors is described as follows:

[0043] First, for each missing sensor, determine the range of its surrounding neighbor sensors by fixing the neighborhood radius or selecting a fixed number of nearest neighbor sensors;

[0044] Then, for each neighbor sensor, the correlation between it and the missing sensor is calculated;

[0045] Next, the calculated correlations are sorted, and the sensor with the highest correlation with the missing sensor is selected as its candidate completion sensor;

[0046] Finally, according to the selected candidate completion sensors, the data is completed using their data, for example, the missing sensor data is predicted using the completion model;

[0047] Step 4: Use the constructed completion model to predict and complete the missing sensor data;

[0048] Merge the completed data with the original data matrix to form a complete sensor data matrix;

[0049] The root mean square error between the repaired data and the true data of each missing sensor is averaged to obtain the overall data repair index.

[0050] In a preferred embodiment, the fault diffusion degree and data repair effect are comprehensively considered, and corresponding warning information is presented to the monitoring system, including the following contents:

[0051] Based on the fault diffusion degree index and the overall data repair index, a comprehensive calculation is performed to obtain the intervention coefficient, and the intervention coefficient is compared with the intervention threshold interval; if several pre-coefficients are not in the intervention threshold interval, and the intervention coefficient is less than the minimum value of the intervention threshold interval, a good signal is generated; if several pre-coefficients are in the intervention threshold interval, an advanced intervention signal is generated; if several pre-coefficients are not in the intervention threshold interval, and the intervention coefficient is greater than the maximum value of the intervention threshold interval, an alarm signal is generated.

[0052] An online monitoring method for mine filling body deformation, comprising: a structure modeling module, an anomaly identification module, an impact analysis module, a data repair module, and a comprehensive evaluation module;

[0053] The structural modeling module constructs the mine filling structure model, arranges the sensor installation base points, and deploys sensors according to the base points, and sends the sensor deployment location information to the anomaly recognition module, impact analysis module and data repair module;

[0054] The anomaly recognition module establishes a correlation model of multi-sensor data, uses statistical inference technology to perform implicit pattern recognition, identifies abnormal state sensors, and sends relevant information of abnormal state sensors to the impact analysis module and data repair module;

[0055] The impact analysis module evaluates the comprehensive impact of sensor loss on the monitoring system from a global perspective based on multi-sensor data interaction through data interpolation and impact analysis modules, and sends the results to the comprehensive evaluation module;

[0056] The data repair module arranges the sensor data in chronological order, marks the location of missing data, searches for neighboring sensors for data repair, builds a completion model for data repair, and sends the results to the comprehensive evaluation module;

[0057] The comprehensive assessment module performs comprehensive analysis based on the data from the impact analysis module and the data repair module to generate corresponding warning information.

[0058] The technical effects and advantages of the online monitoring method and system for mine filling deformation of the present invention are as follows:

[0059] 1. The present invention establishes an association model between multiple sensors by using a Bayesian network, with each node representing a specific sensor variable, and the edge representing the correlation and influence relationship between sensor variables. It further adopts the statistical occurrence number of parent nodes and sensor variables, and combines the kernel density estimation method to estimate the conditional probability distribution, and more flexibly handles the uncertainty and complexity in the parameter estimation process. By counting the occurrence number of parent nodes and sensor variables, the conditional probability distribution can be estimated more accurately, and the kernel density estimation method can better handle the nonlinear and high-dimensional characteristics of the data, improving the fitting ability and accuracy of the model. When calculating the posterior probability of the target sensor, the influence of other sensor data is comprehensively considered. A larger posterior probability indicates that the sensor is in a normal state, while a smaller posterior probability indicates that an abnormal situation may exist. Comparing the posterior probability with the set threshold, generating a corresponding signal prompt, thereby improving the accuracy and fault detection capability of the sensor monitoring system, which is beneficial to discover and solve potential problems in advance and ensure the safe operation of the mine filling body.

[0060] 2. The present invention first determines the degree of influence of sensor failure on the monitoring system by setting a quantitative threshold, thereby improving the timeliness and accuracy of fault identification. Subsequently, the fault propagation path is analyzed using methods such as graph theory and the shortest path algorithm to further evaluate the degree of fault diffusion. The fault diffusion index comprehensively considers the degree of influence and the scope of influence of the sensor, and reflects the degree of influence of the faulty sensor on the entire monitoring system. By analyzing the impact of abnormal sensors on the entire monitoring system, the stability and reliability of the monitoring system can be improved, providing protection for the safe operation of the mine filling body.

[0061] 3. The present invention first arranges the sensor data in the monitoring system in chronological order, marks and determines the location of the missing data, and then repairs the data by finding surrounding sensors with high correlation with the missing sensors, and uses the constructed completion model to predict and complete the missing sensor data. The repaired data is compared with the real data, and the overall data repair index is obtained by calculating the root mean square error, so as to evaluate the repair effect of all missing sensors. Then, according to the fault diffusion index and the overall data repair index, the intervention coefficient is calculated to comprehensively evaluate the fault diffusion degree and data repair effect in the monitoring system, and provide an important reference for subsequent operation and maintenance decisions. Finally, according to the comparison of the intervention coefficient and the intervention threshold interval, the corresponding warning information is generated to determine whether active intervention measures are needed. This improves the timeliness and accuracy of fault identification, effectively ensures the stable operation of the monitoring system and the reliability of the data, and provides strong support for the safe operation of mine filling bodies. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic flow chart of an online monitoring method for mine filling deformation of the present invention;

[0063] Figure 2 The present invention is a schematic structural diagram of an online monitoring system for mine filling body deformation. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0065] Embodiment 1: Figure 1 The present invention provides an online monitoring method for mine filling deformation, comprising:

[0066] S1. Construct a mine filling structure model, arrange installation base points on the sensor installation surface according to the rated space distance, and deploy actual sensors according to the installation base points;

[0067] S2. By constructing a correlation model of multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify sensors in abnormal states;

[0068] S3. Using data interpolation and impact analysis models, based on multi-sensor data interaction, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective;

[0069] S4. Arrange the sensor data in chronological order, mark the location of missing data, find neighboring sensors with high correlation with the missing sensor, and use the constructed completion model to repair the data. Consider the degree of fault diffusion and data repair effect comprehensively, and provide corresponding warning information to the monitoring system.

[0070] Construct a mine filling structure model, lay out the installation base points on the sensor installation surface according to the rated space distance, and deploy the actual sensors according to the installation base points, including the following contents:

[0071] Step 1: Analyze the composition of the filling body, such as ore, slag, cement, etc., and its distribution underground;

[0072] Draw the geological profile and three-dimensional model of the filling body based on geological exploration data and on-site measurements;

[0073] Use computer-aided design (CAD) software or professional geological modeling software to build a detailed filling structure model, including volume, shape, layer and other information;

[0074] Step 2: Based on the 3D model, determine the location, density, and layout of sensors to ensure coverage of key areas of the fill volume and minimize blind spots.

[0075] Step 3: Determine the rated space interval for sensor quality inspection based on sensor specifications and usage requirements to ensure monitoring coverage and monitoring accuracy;

[0076] Select installation base points on the surface or inside of the filling body, use a measuring tool (such as a laser distance meter) to measure the distance between the reference points, and mark the installation position on the surface or inside of the filling body according to the rated space distance;

[0077] Step 4: Install the sensor on the filling body according to the marked installation position, connect the sensor to the monitoring system, and perform line inspection and wiring confirmation to ensure that the sensor is working properly;

[0078] Step 5: Connect the sensor to the monitoring system, establish a data communication link, calibrate and test the sensor, and set the predetermined sampling frequency, range and sensitivity parameters according to the sensor's technical specifications and monitoring requirements;

[0079] Step six, then perform functional testing of the sensor to verify the accuracy and stability of the sensor and ensure its normal operation.

[0080] Through the above detailed steps, ensure that the installation base points are arranged on the sensor installation surface according to the rated spatial distance, and deploy the actual sensors according to the base points to build a reliable online monitoring system.

[0081] By building a correlation model for multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify sensors in abnormal states, including the following:

[0082] In the context of mine filling monitoring, it is of great significance to analyze the sensors with fault tendency hidden in the sensor matrix. First of all, the mine filling body is a complex engineering structure, and its stability and safety are crucial to mine operation. As the core component of the monitoring system, the sensor matrix is ​​responsible for real-time monitoring of the deformation and state changes of the filling body, which plays a vital role in the safe operation of the filling body. If the sensor fails or fails, the monitoring system will not be able to accurately capture the changes in the filling body, which may lead to misjudgment of the filling body state or missed warnings. Therefore, analyzing sensors with fault tendency can timely detect and eliminate faults, ensure the reliability and accuracy of the monitoring system, effectively prevent possible safety problems of the filling body, and improve the stability and safety of the filling body. In addition, by timely repairing or replacing faulty sensors, human and material resources can be saved, and the operating efficiency and cost-effectiveness of the mine filling body monitoring system can be improved. Therefore, analyzing sensors with fault tendency is of great significance in mine filling body monitoring, which helps to ensure the safe operation and production efficiency of mine filling bodies.

[0083] Step 1: Use the Bayesian network to build a correlation model between sensors, where each node represents a sensor variable and the edge represents the dependency relationship between variables;

[0084] Each node represents a specific sensor variable, such as a displacement sensor, strain gauge, pressure sensor, etc.;

[0085] Edges represent the correlation and influence relationship between sensor variables, that is, how the value of a sensor variable is affected by other sensor variables;

[0086] The dependencies between sensor variables can be unidirectional or bidirectional, representing causal relationships or joint effects between sensors;

[0087] The direction and weight of the edge reflect the strength and directionality of the causal relationship between variables, that is, the degree of influence of one sensor variable on another sensor variable;

[0088] Step 2: Estimate the conditional rerouting distribution of each node, which includes the conditional probability of the node given its parent node.

[0089] Based on historical data, the number of occurrences of parent nodes and sensor variables is counted, and then the conditional probability distribution is estimated using the kernel density estimation method: ;

[0090] in: Indicates that at a given parent node In the case of The conditional probability distribution of is the target that needs to be estimated;

[0091] Indicates the parent node in the historical data The number of occurrences of represents the frequency information of the parent node and is used to estimate the conditional probability distribution;

[0092] Represents sensor variables in historical data and its parent node The number of joint occurrences of is the sample size, which indicates the number of samples in the historical data;

[0093] is the bandwidth parameter, which is used to control the width of the kernel function. The smaller the bandwidth, the smaller the range of the kernel function and the more precise the estimation of the probability density;

[0094] It is a kernel function used to weight samples. For example, Gaussian kernel function and Epanechnikov kernel function;

[0095] Represents the sensor variable In the observations in samples or time points;

[0096] It is used for normalization to ensure that the estimated probability density function integrates to 1 and that it is a legal probability distribution.

[0097] Step 3: Use Bayes’ theorem to calculate the posterior probability of the target sensor, that is, the state probability of the target sensor given the data of other sensors: ;

[0098] In the formula Indicates the target sensor The posterior probability is the probability of the sensor state to be determined after observing the actual data;

[0099] Indicates that the sensor The status is The probability of observing the actual data under the condition of , that is, the possibility of the sensor giving the observed data;

[0100] The sensor The prior probability of , that is, the initial estimate of the sensor state before any data is observed.

[0101] The posterior probability of the target sensor given the observed data is calculated by multiplying the sensor's prior probability with the conditional probability of the observed actual data and taking a weighted sum of all possible sensor states. This posterior probability can help determine whether the state of the target sensor is abnormal or has changed.

[0102] The posterior probability of the target sensor is used to represent the state probability of the target sensor when other sensor data have been observed. When the posterior probability of the target sensor is large, it indicates that after considering the influence of other sensor data, the target sensor is more likely to be in a normal state. Conversely, when the posterior probability of the target sensor is small, it means that after considering the influence of other sensor data, the target sensor is more likely to be in an abnormal state. Therefore, the larger the posterior probability of the target sensor, the more likely its state is normal, and the smaller the posterior probability, the more likely its state is abnormal. When monitoring and identifying abnormal states, the posterior probabilities of the target sensors can be compared according to the set threshold to determine whether they are abnormal, and further take corresponding measures to deal with them.

[0103] Step 4: compare the posterior probability of the target sensor with the set threshold; if the posterior probability of the target sensor is greater than or equal to the set threshold, it means that the monitoring system has sufficient confidence in the state of the target sensor and believes that the target sensor is in a normal state, and generates a normal state signal;

[0104] If the posterior probability of the target sensor is less than the set threshold, it means that the monitoring system is highly suspicious of the state of the target sensor and there is an abnormality. A signal prompt of the abnormal state is generated to remind the operator or system monitoring personnel to pay attention to possible problems.

[0105] The present invention establishes an association model between multiple sensors by using a Bayesian network, with each node representing a specific sensor variable, and the edge representing the correlation and influence relationship between sensor variables. It further adopts the statistical occurrence number of parent nodes and sensor variables, and combines the kernel density estimation method to estimate the conditional probability distribution, so as to more flexibly handle the uncertainty and complexity in the parameter estimation process. By counting the occurrence number of parent nodes and sensor variables, the conditional probability distribution can be estimated more accurately, and the kernel density estimation method can better handle the nonlinear and high-dimensional characteristics of the data, thereby improving the fitting ability and accuracy of the model. When calculating the posterior probability of the target sensor, the influence of other sensor data is comprehensively considered. A larger posterior probability indicates that the sensor is in a normal state, while a smaller posterior probability indicates that an abnormal situation may exist. The posterior probability is compared with the set threshold, and a corresponding signal prompt is generated, thereby improving the accuracy and fault detection capability of the sensor monitoring system, which is beneficial to discover and solve potential problems in advance and ensure the safe operation of the mine filling body.

[0106] Using data interpolation and impact analysis models, based on multi-sensor data interaction, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective, including the following:

[0107] In the online monitoring scenario of mine filling deformation, it is very important to analyze the impact of the sensors that have been identified as abnormal on the entire monitoring system. The purpose of doing so is to comprehensively evaluate the impact of abnormal sensors on the overall performance of the monitoring system so that appropriate maintenance measures can be taken in a timely manner. Although the monitoring system usually has redundant monitoring resources, even the loss of one or two sensors may have unexpected or unpredictable effects on the entire monitoring system. First, the data of abnormal sensors may mislead the analysis and decision-making process of the monitoring system, resulting in inaccurate assessment of the deformation state of the filling body. Second, abnormal sensors may affect other sensors or monitoring nodes associated with them, thereby expanding the scope of the fault and causing the overall performance of the monitoring system to deteriorate. Therefore, by analyzing the impact of abnormal sensors on the entire monitoring system, the severity of the fault can be more comprehensively evaluated, which helps to determine whether immediate maintenance measures are needed. Although immediate maintenance may increase the maintenance burden when it is not necessary, it is very necessary to maintain abnormal sensors in a timely manner under the premise of ensuring the normal operation of the monitoring system and data reliability. Therefore, by analyzing the impact of abnormal sensors on the entire monitoring system, the stability and reliability of the monitoring system can be improved, providing protection for the safe operation of mine filling bodies.

[0108] For sensors that generate abnormal status signals, they are marked as missing sensors.

[0109] When the number of missing sensors is greater than the ratio threshold, urgent maintenance measures are required. When it is less than or equal to the number threshold, the fault diffusion index is calculated using the following calculation method;

[0110] Assume that there are 100 sensors and the set number threshold is 5%. If the number of missing sensors is greater than 5, an alarm is issued. If it is less than 5, the deformation monitoring impact caused by the loss of these 5 sensors is further calculated. Based on the degree of impact, further analysis is performed to determine whether attention or intervention is needed.

[0111] The main reason for setting the quantity threshold is to timely identify and respond to the impact of sensor failures on the monitoring system. Multiple missing sensors will complicate and inaccurate the analysis of the abnormal degree of fault expansion, resulting in difficulty in calculation and distortion of calculation results. Therefore, it is necessary to control the accuracy of the calculation results within a reasonable and controllable proportion of calculations. In the monitoring system, the role of sensors is crucial. Once a failure occurs, it may cause inaccurate monitoring data or loss of important information, which in turn affects the monitoring and evaluation of the deformation of the mine filling body. Therefore, setting a quantity threshold can help determine when urgent maintenance measures are needed to ensure the normal operation of the monitoring system and the reliability of the data. For the deformation monitoring of mine filling bodies, the setting of the quantity threshold of sensors is particularly important. Changes in the filling body structure may cause serious safety hazards, so it is crucial to detect and deal with sensor failures in a timely manner. When the number of missing sensors exceeds the threshold, the issuance of an alarm can remind the monitoring personnel to take timely action to ensure the stability and safety of the filling body structure; when the number of missing sensors is less than or equal to the threshold, further calculating the degree of fault diffusion can help analyze the deformation monitoring impact caused by the loss of these sensors, so as to timely determine whether attention or intervention is needed, which helps to ensure the safe operation of the mine filling body and the accuracy of monitoring data. Therefore, setting a quantitative threshold has important practical significance for monitoring the deformation of mine filling bodies and helps to ensure the safety and stability of the filling body structure.

[0112] Represent the sensor matrix as a graph ,in is a set of sensor nodes, is a set of edges, representing the connection relationship between sensors;

[0113] Define the fault propagation path matrix ,in Represents a slave node To Node The propagation path length;

[0114] Use the shortest path algorithm in graph theory to calculate the shortest path length matrix between sensor nodes ,in Represents a slave node To Node The shortest path length.

[0115] For missing sensor nodes , calculate its influence range matrix ,in Representation Node For Node degree of impact.

[0116] It can be defined as: ;

[0117] The formula represents the missing sensor node For Node The degree of influence and node To Node The shortest path length and nodes To Node The shortest path length is inversely proportional to the reciprocal of the product of and is a weight parameter used to adjust the contribution of path length to the degree of influence.

[0118] When the number of missing sensors is greater than or equal to the threshold, the sum of the rows and columns of each impact range matrix represents the degree of impact and the degree of impact caused by each node. Define these two vectors as and .

[0119] The fault diffusion index can be defined as , which represents the degree of sensor data anomaly caused by node failure. It can be calculated by the following formula: ;

[0120] The formula indicates that the abnormality index caused by the faulty sensor node is half of the average value of the affected degree of all nodes. It is mainly to comprehensively consider the affected degree and impact degree of the sensor and to more comprehensively evaluate the impact degree of the missing sensor on the overall monitoring system.

[0121] The fault diffusion index reflects the degree of influence of the faulty sensor on the entire monitoring system and is an important indicator for evaluating the operating status and data reliability of the monitoring system. The size of the index reflects the spread and severity of the abnormal situation caused by the faulty sensor in the monitoring system. When the fault diffusion index is large, it means that the abnormal impact of the faulty sensor is more extensive and serious, which may cause distortion of the monitoring data or loss of the effectiveness of the deformation monitoring of the mine filling body. It is necessary to take timely measures to repair the fault or adjust the monitoring system. On the contrary, when the fault diffusion index is small, it means that the abnormal impact of the faulty sensor is relatively small, and the monitoring system is less affected, but it is still necessary to further monitor and analyze the abnormal situation of the faulty sensor to ensure the accuracy and reliability of the monitoring data. Therefore, the larger the fault diffusion index, the more serious the impact of the faulty sensor on the monitoring system, and the smaller it is, the lighter the impact.

[0122] The present invention firstly determines the degree of influence of sensor faults on the monitoring system by setting a quantitative threshold, thereby improving the timeliness and accuracy of fault identification. Subsequently, the fault propagation path is analyzed using methods such as graph theory and the shortest path algorithm to further evaluate the degree of fault diffusion. The fault diffusion index comprehensively considers the degree of influence and the scope of influence of the sensor, and reflects the degree of influence of the faulty sensor on the entire monitoring system. By analyzing the impact of abnormal sensors on the entire monitoring system, the stability and reliability of the monitoring system can be improved, providing protection for the safe operation of the mine filling body.

[0123] By arranging the sensor data in chronological order, marking the missing data locations, finding neighboring sensors with high correlation with the missing sensors, and using the constructed completion model to repair the data, the following is included:

[0124] The reason for analyzing the data repair of abnormal sensors is to ensure the accuracy and reliability of the monitoring system. Abnormal sensors may cause distortion of the monitoring system data or loss of important information, which in turn affects the monitoring and evaluation of the deformation of the mine filling body. Therefore, it is crucial to repair the data of abnormal sensors. Comprehensive analysis of the fault diffusion degree and data repair effect of abnormal sensors can comprehensively evaluate their impact on the monitoring system. The fault diffusion index reflects the impact of abnormal sensors on the entire monitoring system, while the data repair effect indicates the accuracy and credibility of the abnormal sensor data repair. Comprehensive analysis of these two factors can help determine whether further maintenance measures are needed to ensure the normal operation of the monitoring system and provide important reference for subsequent data analysis and decision-making.

[0125] Step 1: Arrange the data of each sensor in the monitoring system in chronological order to form a sensor data matrix D. The rows represent time steps and the columns represent sensor numbers.

[0126] Step 2: Mark the location of missing sensor data in the sensor data matrix, that is, determine the row and column indexes of the missing data;

[0127] Step 3: For each missing sensor, find a sensor with a high correlation based on the data of the surrounding sensors;

[0128] The operation process of finding sensors with high correlation with the missing sensors can be described as follows:

[0129] First, for each missing sensor, determine the range of its surrounding neighbor sensors by fixing the neighborhood radius or selecting a fixed number of nearest neighbor sensors;

[0130] Then, for each neighbor sensor, the correlation between it and the missing sensor is calculated. Commonly used methods include Pearson correlation coefficient, Spearman correlation coefficient or cross correlation;

[0131] Next, the calculated correlations are sorted, and the sensor with the highest correlation with the missing sensor is selected as its candidate completion sensor;

[0132] Finally, according to the selected candidate completion sensors, data completion operations are performed using their data conditions, such as using the completion model to predict the missing sensor data.

[0133] This process can effectively utilize the information of surrounding sensors and find sensors that are highly correlated with the missing sensors, thereby improving the accuracy and efficiency of the completion process.

[0134] Step 4: Use the constructed completion model to predict and complete the missing sensor data. For example, the following formula can be used to complete the data: ;

[0135] in, Indicates repairing sensor data. and are the left singular matrix and right singular matrix after SVD decomposition, Sigma is a diagonal matrix, and K is the number of selected singular values.

[0136] The completed data is merged with the original data matrix to form a complete sensor data matrix.

[0137] The root mean square error between the repaired data and the true data of each missing sensor is averaged to obtain the overall data repair index.

[0138] The real data here refers to the data at the time point closest to the current time before the missing sensor obtains the signal of the abnormal state, and this data is regarded as the real data.

[0139] The overall data repair index is used to comprehensively evaluate the repair effect of all missing sensors. It reflects the overall difference between the repaired data and the real data of the missing sensors in the entire monitoring system. When the overall data repair index is smaller, it means that the root mean square error between the repaired data and the real data is smaller, that is, the repair effect is better, and the accuracy and reliability of the sensor data are higher; conversely, when the overall data repair index is larger, it means that the root mean square error between the repaired data and the real data is larger, that is, the repair effect is poor, and the accuracy and reliability of the sensor data are lower. Therefore, the size of the overall data repair index can intuitively reflect the repair status of the missing sensor data in the entire monitoring system, which helps to evaluate the data quality and accuracy of the monitoring system and provide a reference for subsequent data analysis and decision-making.

[0140] Taking into account the degree of fault propagation and the effect of data repair, the monitoring system is provided with corresponding warning information, including the following:

[0141] According to the fault diffusion index and the overall data repair index, a comprehensive calculation is performed to obtain the intervention coefficient. For example, it can be calculated by the following calculation formula: ;

[0142] in, represents the intervention coefficient, and They are the fault diffusion index and the overall data repair index, and are the preset proportional coefficients of the fault diffusion degree index and the overall data repair index, respectively, and both are greater than zero.

[0143] The intervention coefficient is used to comprehensively evaluate the degree of fault diffusion and data repair effect in the monitoring system, so as to help determine whether active intervention measures are needed to maintain the normal operation of the system. When the intervention coefficient is large, it means that the degree of fault diffusion in the monitoring system is high and the data repair effect is poor, and timely intervention measures are needed to avoid further problems; conversely, when the intervention coefficient is small, it means that the degree of fault diffusion in the monitoring system is low and the data repair effect is good, the system is in a relatively stable state, and intervention measures may not be the primary consideration. Therefore, the size of the intervention coefficient can intuitively understand the overall operation of the monitoring system, providing an important reference for subsequent operation and maintenance decisions.

[0144] Step 6: Display the corresponding warning information including the following:

[0145] Compare the intervention coefficient and intervention threshold interval;

[0146] If the intervention coefficient is not in the intervention threshold interval, and the intervention coefficient is less than the minimum value of the intervention threshold interval, it means that the current operating state of the monitoring system is stable, the degree of fault diffusion is low, and the data repair effect is good. There is no need to take active intervention measures to generate a good signal.

[0147] If the intervention coefficient is in the intervention threshold range, it means that there are certain problems in the current operating status of the monitoring system, the degree of fault diffusion is high or the data repair effect is poor, and active intervention measures need to be taken to prevent the problem from further deteriorating and generate advanced intervention signals.

[0148] If the intervention coefficient is not in the intervention threshold interval, and the intervention coefficient is greater than the maximum value of the intervention threshold interval, it means that there is a serious problem with the current operating status of the monitoring system, the degree of fault diffusion is high or the data repair effect is poor, and urgent active intervention measures need to be taken immediately to maintain the normal operation of the system to avoid possible serious consequences. An alarm signal is generated and the faulty sensor needs to be replaced immediately.

[0149] The present invention first arranges the sensor data in the monitoring system in chronological order, marks and determines the location of the missing data, and then repairs the data by finding surrounding sensors with high correlation with the missing sensors, and uses the constructed completion model to predict and complete the missing sensor data. The repaired data is compared with the real data, and the overall data repair index is obtained by calculating the root mean square error, so as to evaluate the repair effect of all missing sensors. Then, according to the fault diffusion index and the overall data repair index, the intervention coefficient is calculated to comprehensively evaluate the fault diffusion degree and data repair effect in the monitoring system, and provide an important reference for subsequent operation and maintenance decisions. Finally, according to the comparison of the intervention coefficient and the intervention threshold interval, the corresponding warning information is generated to determine whether active intervention measures are needed. This improves the timeliness and accuracy of fault identification, effectively ensures the stable operation of the monitoring system and the reliability of the data, and provides strong support for the safe operation of the mine filling body.

[0150] Embodiment 2: Figure 2 The invention provides an online monitoring system for the deformation of mine filling bodies, comprising: a structure modeling module, an abnormality identification module, an impact analysis module, a data repair module, and a comprehensive evaluation module;

[0151] The structural modeling module constructs the mine filling structure model, arranges the sensor installation base points, and deploys sensors according to the base points, and sends the sensor deployment location information to the anomaly recognition module, impact analysis module and data repair module;

[0152] The anomaly recognition module establishes a correlation model of multi-sensor data, uses statistical inference technology to perform implicit pattern recognition, identifies abnormal state sensors, and sends relevant information of abnormal state sensors to the impact analysis module and data repair module;

[0153] The impact analysis module evaluates the comprehensive impact of sensor loss on the monitoring system from a global perspective based on multi-sensor data interaction through data interpolation and impact analysis modules, and sends the results to the comprehensive evaluation module;

[0154] The data repair module arranges the sensor data in chronological order, marks the location of missing data, searches for neighboring sensors for data repair, builds a completion model for data repair, and sends the results to the comprehensive evaluation module;

[0155] The comprehensive assessment module performs comprehensive analysis based on the data from the impact analysis module and the data repair module to generate corresponding warning information.

[0156] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0157] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0158] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0159] In the several embodiments provided in the present application, it should be understood that the disclosed systems and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0160] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0161] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0162] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0163] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0164] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for online monitoring of mine filling deformation, characterized in that: S1. Construct a mine filling structure model, arrange installation base points on the sensor installation surface according to the rated space distance, and deploy actual sensors according to the installation base points; S2. By constructing a correlation model of multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify sensors in abnormal states; S3. Using data interpolation and impact analysis models, based on multi-sensor data interaction, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective; S4. Arrange the sensor data in chronological order, mark the missing data location, find neighbor sensors with high correlation with the missing sensor, and use the constructed completion model to repair the data. Consider the fault diffusion degree and data repair effect comprehensively, and provide the monitoring system with corresponding warning information; By building a correlation model for multi-sensor data and using statistical inference technology, implicit pattern recognition is performed based on the observed multi-sensor data to identify sensors in abnormal states, including the following: Step 1: Use the Bayesian network to build a correlation model between sensors, where each node represents a sensor variable and the edge represents the dependency relationship between variables; Step 2: estimate the conditional rerouting distribution of each node, which includes the conditional probability of the node given its parent node; Based on historical data, the number of occurrences of parent nodes and sensor variables is counted, and then the conditional probability distribution is estimated using the kernel density estimation method: ; in: Indicates that at a given parent node In the case of The conditional probability distribution of is the target that needs to be estimated; Indicates the parent node in the historical data The number of occurrences of Represents sensor variables in historical data and its parent node The number of joint occurrences of is the sample size, which indicates the number of samples in the historical data; is the bandwidth parameter, which is used to control the width of the kernel function; is the kernel function, used to weight the samples; Represents the sensor variable In the observations in samples or time points; It is used for normalization to ensure that the estimated probability density function integrates to 1; Step 3: Use Bayes’ theorem to calculate the posterior probability of the target sensor, that is, the state probability of the target sensor given the data of other sensors: ; In the formula Indicates the target sensor The posterior probability is the probability of the sensor state to be determined after observing the actual data; Indicates that the sensor The status is The probability of observing the actual data under the condition of , that is, the possibility of the sensor giving the observed data; The sensor The prior probability of is the initial estimate of the sensor state before any data is observed; Step 4: compare the posterior probability of the target sensor with the set threshold; if the posterior probability of the target sensor is greater than or equal to the set threshold, it is considered that the target sensor is in a normal state and a normal state signal is generated; If the posterior probability of the target sensor is less than the set threshold, there is an abnormal situation, and a signal prompt of the abnormal state is generated; Using data interpolation and impact analysis models, based on multi-sensor data interaction, the comprehensive impact of sensor loss on the monitoring system is evaluated from a global perspective, including the following: For sensors that generate abnormal status signals, they are marked as missing sensors; When the number of missing sensors is greater than the ratio threshold, urgent maintenance measures are required; When the number of missing sensors is less than or equal to the number threshold, the fault diffusion index is calculated using the following calculation method: Represent the sensor matrix as a graph ,in is a set of sensor nodes, is a set of edges, representing the connection relationship between sensors; Define the fault propagation path matrix ,in Represents a slave node To Node The propagation path length; Use the shortest path algorithm in graph theory to calculate the shortest path length matrix between sensor nodes ,in Represents a slave node To Node The shortest path length; For missing sensor nodes , calculate its influence range matrix ,in Representation Node For Node The degree of influence, Defined as: ; The formula represents the missing sensor node For Node The degree of influence and node To Node The shortest path length and nodes To Node The shortest path length is inversely proportional to the reciprocal of the product of and is the weight parameter; When the number of missing sensors is less than or equal to the threshold, the sum of the rows and columns of each impact range matrix represents the degree of impact on each node and the degree of impact caused, respectively. These two vectors are defined as and ; The fault diffusion index is defined as , represents the degree of sensor data anomaly caused by node failure, and is calculated by the following formula: ; By arranging the sensor data in chronological order, marking the missing data locations, finding neighboring sensors with high correlation with the missing sensors, and using the constructed completion model to repair the data, the following is included: Step 1: Arrange the data of each sensor in the monitoring system in chronological order to form a sensor data matrix D, where the rows represent time steps and the columns represent sensor numbers; Step 2: Mark the location of missing sensor data in the sensor data matrix, that is, determine the row and column indexes of the missing data; Step 3: For each missing sensor, find a sensor with a high correlation based on the data of the surrounding sensors; The operation process of finding sensors with high correlation with the missing sensors is described as follows: First, for each missing sensor, determine the range of its surrounding neighbor sensors by fixing the neighborhood radius or selecting a fixed number of nearest neighbor sensors; Then, for each neighbor sensor, the correlation between it and the missing sensor is calculated; Finally, the calculated correlations are sorted, and the sensor with the highest correlation with the missing sensor is selected as its candidate completion sensor; Step 4: Use the constructed completion model to predict and complete the missing sensor data; Merge the completed data with the original data matrix to form a complete sensor data matrix; The root mean square error between the repaired data and the true data of each missing sensor is averaged to obtain the overall data repair index.

2. The method for online monitoring of mine filling deformation according to claim 1, characterized in that: Taking into account the degree of fault propagation and the effect of data repair, the monitoring system is provided with corresponding warning information, including the following: Based on the fault diffusion degree index and the overall data repair index, a comprehensive calculation is performed to obtain the intervention coefficient, and the intervention coefficient is compared with the intervention threshold interval; if several pre-coefficients are not in the intervention threshold interval, and the intervention coefficient is less than the minimum value of the intervention threshold interval, a good signal is generated; if several pre-coefficients are in the intervention threshold interval, an advanced intervention signal is generated; if several pre-coefficients are not in the intervention threshold interval, and the intervention coefficient is greater than the maximum value of the intervention threshold interval, an alarm signal is generated.

3. A method for online monitoring of mine filling deformation, used to implement the method for online monitoring of mine filling deformation as claimed in any one of claims 1 to 2, comprising: Structural modeling module, anomaly identification module, impact analysis module, data repair module, and comprehensive evaluation module; The structural modeling module constructs the mine filling structure model, arranges the sensor installation base points, and deploys sensors according to the base points, and sends the sensor deployment location information to the anomaly recognition module, impact analysis module and data repair module; The anomaly recognition module establishes a correlation model of multi-sensor data, uses statistical inference technology to perform implicit pattern recognition, identifies abnormal state sensors, and sends relevant information of abnormal state sensors to the impact analysis module and data repair module; The impact analysis module evaluates the comprehensive impact of sensor loss on the monitoring system from a global perspective based on multi-sensor data interaction through data interpolation and impact analysis models, and sends the results to the comprehensive evaluation module; The data repair module arranges the sensor data in chronological order, marks the location of missing data, searches for neighboring sensors for data repair, builds a completion model for data repair, and sends the results to the comprehensive evaluation module; The comprehensive assessment module performs comprehensive analysis based on the data from the impact analysis module and the data repair module to generate corresponding warning information.

Images