An intelligent monitoring system and method for underground filling quality in metal mines based on the Internet of Things

Abstract

The present invention relates to the field of mines, and in particular to an intelligent monitoring system and method for the quality of underground filling bodies in metal mines based on the Internet of Things. The system comprises: collecting original measurement values, automatically calibrating the received original measurement values ​​using a sensor cluster mutual verification technology based on Kalman filtering, and only performing mutual verification and repair on original measurement values ​​that have been detected to be drifting or failing, to obtain calibrated measurement values; fusing the calibrated measurement values ​​of sensors of the same type to obtain fused data and generate a low-dimensional feature vector; predicting the risk probability of the quality of underground filling bodies in metal mines by using a feature selection weighted random forest prediction algorithm, comparing it with the risk threshold, performing graded warning, obtaining a warning level, and generating a corresponding reinforcement plan. The system solves the technical problems that underground sensors are affected by environmental interference or equipment aging, measurement values ​​are prone to drift and failure, and traditional methods lack an automatic calibration mechanism, require manual intervention, and are inefficient.

Description

Technical Field

[0001] The present invention relates to the field of mines, and in particular to an intelligent monitoring system and method for underground filling quality of metal mines based on the Internet of Things. Background Art

[0002] With the continued growth in global demand for mineral resources, the mining depth and scale of metal mines are constantly expanding. Deep mine mining faces more complex geological conditions and high-stress environments. Underground backfill mining, as an efficient and environmentally friendly mining technology, has been widely used in metal mines. Underground backfill mining forms a backfill by filling the goaf with tailings, waste rock, or other materials. This method not only effectively controls ground pressure and reduces surface subsidence, but also improves resource recovery and reduces environmental pollution. However, the quality of the backfill directly affects the safe production and economic benefits of the mine. Unstable backfill quality may lead to serious accidents such as goaf instability, surrounding rock collapse, and even ground collapse. Therefore, real-time and accurate monitoring of the quality of underground backfill has become one of the key technical requirements for safe production in metal mines.

[0003] Traditional backfill quality monitoring methods rely primarily on manual inspections, periodic sampling, and simple physical testing. These methods suffer from the following problems: First, monitoring efficiency is low; manual inspections and sampling analysis are time-consuming and preclude real-time monitoring. Second, monitoring accuracy is limited; traditional methods struggle to fully reflect the mechanical properties, density, and long-term stability of the backfill. Third, data acquisition and analysis methods are outdated, lacking systematic and automated data processing capabilities, making it difficult to meet the dynamic monitoring needs of backfill quality in complex deep environments. Furthermore, traditional monitoring methods are often unable to fully perceive the internal structure and stress state of the backfill, especially in deep, high-stress, high-temperature, and high-humidity underground environments. Monitoring equipment is susceptible to environmental interference, and data acquisition reliability and stability are insufficient.

[0004] In recent years, the rapid development of the Internet of Things (IoT) has provided a new technological path for intelligent mine monitoring. By integrating sensors, communication networks, and data processing platforms, the IoT enables real-time perception, data transmission, and intelligent analysis of the physical world. The IoT-based intelligent monitoring system for underground backfill quality in metal mines deploys multiple types of sensors, combines wireless communication technology with a big data analysis platform, and enables real-time, dynamic, and accurate monitoring of backfill quality. This not only improves monitoring efficiency but also provides a scientific basis for mine safety management and reduces accident risks.

[0005] However, the above-mentioned existing intelligent monitoring system for underground filling quality of metal mines still has technical problems such as insufficient data transmission reliability in complex environments, insufficient accuracy and reliability of sensor data, difficulty in fusing multi-source heterogeneous data, and delayed risk prediction and early warning. Summary of the Invention

[0006] The present invention provides an intelligent monitoring system and method for the quality of underground filling bodies in metal mines based on the Internet of Things, so as to solve the technical problems that the traditional single network suffers from severe signal attenuation, limited bandwidth, high energy consumption, and difficulty in ensuring the stability of data transmission in the underground metal interference and tunnel bending environment; underground sensors are affected by environmental interference or equipment aging, and the measurement values ​​are prone to drift and failure. The traditional method lacks an automatic calibration mechanism and requires manual intervention, which is inefficient; the data formats and characteristics of different types of sensors (such as fiber gratings, ultrasonic flaw detectors, microseismic sensors, etc.) vary greatly, and traditional fusion methods cannot be effectively integrated, resulting in inaccurate filling body status assessment; the traditional random forest algorithm treats features equally, is easily interfered by noise characteristics, has limited prediction accuracy, and the early warning system has a slow response speed, which is difficult to meet the needs of real-time underground monitoring.

[0007] The present invention provides an intelligent monitoring system and method for underground filling material quality in metal mines based on the Internet of Things, which specifically includes the following technical solutions:

[0008] An intelligent monitoring method for underground filling quality in metal mines based on the Internet of Things comprises the following steps:

[0009] S1. Build an adaptive hybrid transmission network, dynamically select the optimal transmission path by quantifying path performance, and deploy different types of sensors to collect raw measurement values ​​in real time. Leveraging sensor cluster inter-calibration technology based on Kalman filtering, the received raw measurement values ​​are automatically calibrated. During the automatic calibration process, only raw measurement values ​​detected to be drifted or invalid are inter-calibrated and repaired to obtain calibrated measurement values.

[0010] S2. Based on the Demster-Schafer evidence theory, the calibrated measurements of sensors of the same type are fused to obtain fused data from sensors of different types. This fused data is used as input to construct a time series using a time window. Key features of filling quality degradation are extracted to generate a low-dimensional feature vector.

[0011] S3. Based on low-dimensional feature vectors, the risk probability of the quality of underground filling materials in metal mines is predicted through feature selection and weighted random forest prediction algorithm. The risk probability of the quality of underground filling materials in metal mines is compared with the risk threshold, and a graded warning is performed to obtain the warning level. The corresponding reinforcement plan is generated and pushed to the user terminal in real time.

[0012] Preferably, the S1 specifically includes:

[0013] In the implementation process of the sensor cluster mutual verification technology based on Kalman filtering, the difference between the original measurement value of each sensor and the state estimation value obtained by iteratively calculating the original measurement value of the sensor is calculated. When the difference between the two exceeds the error threshold, the original measurement value is judged as a drifted or failed original measurement value. For sensors that are detected to be drifted or failed, the state estimation values ​​of sensors of the same type excluding themselves are used to calculate the average value as the calibrated measurement value.

[0014] Preferably, the S1 specifically includes:

[0015] In the implementation process of the sensor cluster mutual verification technology based on Kalman filtering, when the difference between the two is less than or equal to the error threshold, the original measurement value is judged to be normal and used as the calibrated measurement value.

[0016] Preferably, the S2 specifically includes:

[0017] In the process of fusing calibrated measurement values ​​of sensors of the same type, the calibrated measurement values ​​of the sensors are used as evidence, and the support degree of the evidence for various evidence states is quantified through basic probability assignment.

[0018] Preferably, the S2 specifically includes:

[0019] In the process of fusing calibrated measurements of sensors of the same type, a basic probability assignment is assigned to each calibrated measurement of each type of sensor. For each sensor, the basic probability assignment is calculated based on the absolute deviation of the calibrated measurement value of the sensor and its mean, and divided by the standard deviation of the calibrated measurement value of the sensor to obtain the standardized deviation. The basic probability assignment of the reliable state is generated by adding one and taking the inverse transformation. The basic probability assignment of the unreliable state is one minus the basic probability assignment of the reliable state.

[0020] Preferably, the S2 specifically includes:

[0021] The basic probability assignments of sensors of the same type are fused. By accumulating all combinations of evidence supporting the "reliable" state and eliminating the influence of conflicting evidence, the initial trust weight of each sensor is calculated. The normalized trust weight is then used to perform a weighted average of the calibrated measurements of sensors of the same type to generate fused data of different types of sensors.

[0022] Preferably, the S3 specifically includes:

[0023] In the implementation process of the feature selection weighted random forest prediction algorithm, a feature importance vector is introduced. The feature importance vector assigns an importance weight to each selected feature, reflecting the relative importance of the feature to the quality risk prediction of underground filling in metal mines.

[0024] Preferably, the S3 specifically includes:

[0025] In the implementation process of the feature selection weighted random forest prediction algorithm, the prediction output of each decision tree is mapped through the Sigmoid function, and the prediction outputs of all decision trees are summed and combined with the average factor to obtain the risk probability of the quality of underground filling in metal mines.

[0026] An intelligent monitoring system for underground filling quality in metal mines based on the Internet of Things includes the following parts:

[0027] Sensor data acquisition module, data calibration module, data fusion module, feature extraction module, risk prediction module, and early warning push module;

[0028] Sensor data acquisition module: deploys different types of sensors, each of which collects raw measurement values ​​in real time at a fixed frequency and outputs the raw measurement values ​​to the data calibration module;

[0029] Data calibration module: This module automatically calibrates the raw measurement values ​​from the sensor data acquisition module using sensor cluster mutual verification technology based on Kalman filtering. During the automatic calibration process, only the raw measurement values ​​that have been detected to be drifting or invalid are inter-calibrated and repaired to obtain the calibrated measurement values, which are then output to the data fusion module.

[0030] Data fusion module: Based on the Demster-Schafer evidence theory, the calibrated measurement values ​​of the same type of sensors from the data calibration module are fused to obtain different types of sensor fusion data, and the different types of sensor fusion data are output to the feature extraction module;

[0031] Feature extraction module: This module takes the sensor fusion data from the data fusion module as input, constructs a time series through time windows, and uses a lightweight long short-term memory network to extract key features of filling quality degradation, generating low-dimensional feature vectors. This low-dimensional feature vector is then output to the risk prediction module.

[0032] Risk prediction module: Based on the low-dimensional feature vector of the feature extraction module, the feature selection weighted random forest prediction algorithm is used to predict the risk probability of the quality of the underground filling body of the metal mine, and the risk probability of the quality of the underground filling body of the metal mine is output to the early warning push module;

[0033] Early warning push module: Compares the risk probability of the quality of the underground filling body of the metal mine from the risk prediction module with the preset risk threshold, divides the risk of the quality of the underground filling body of the metal mine into three risk levels: yellow, orange, and red, obtains the early warning level, generates the corresponding reinforcement plan, and pushes the early warning results and reinforcement plan to the user terminal in real time to support on-site personnel in making quick decisions.

[0034] The beneficial effects of the technical solution of the present invention are:

[0035] 1. By integrating narrowband IoT and wireless mesh networks, an adaptive hybrid transmission network was constructed, achieving efficient transmission in the complex underground environment of metal mines. Based on the existing game theory model, the optimal transmission path was dynamically selected, and factors such as signal strength, energy consumption, and bandwidth were comprehensively considered to ensure communication stability in harsh environments such as metal interference and curved tunnels. This significantly improved the reliability of data transmission and reduced the risk of communication interruption. It provided a stable data channel for real-time monitoring of the quality of underground filling in metal mines and ensured the continuity of mine safety monitoring.

[0036] 2. The original measurement values ​​of the sensors are automatically calibrated through the sensor cluster mutual verification technology based on Kalman filtering. Only the original measurement values ​​of the sensors that are detected to be drifted or failed are inter-calibrated and repaired, obtaining high-precision calibrated measurement values, significantly improving the reliability and accuracy of the data, and reducing the risk of misjudgment due to sensor failure or drift. At the same time, automatic calibration reduces the need for manual maintenance, reduces operation and maintenance costs, enhances the fault tolerance and real-time performance of the intelligent monitoring system for underground filling quality of metal mines, and provides efficient protection for the continuous monitoring of mine filling quality.

[0037] 3. Based on the Demster-Schafer evidence theory, the calibrated measurement values ​​of sensors of the same type are fused to generate fused data of sensors of different types, ensuring that the fused data of the sensors accurately reflects the true physical state of the filling body, improving the credibility of the monitoring data, providing high-quality input for subsequent feature extraction and risk prediction, and enhancing the ability of the intelligent monitoring system for underground filling body quality in metal mines to accurately judge the filling body state.

[0038] 4. By introducing the feature selection weighted random forest prediction algorithm, the risk probability of the quality of underground filling in metal mines is predicted based on low-dimensional feature vectors, and graded warning is achieved by comparing with the preset risk threshold. Through feature importance screening, the contribution of key physical quantities to the prediction is highlighted, which reduces noise interference, improves prediction accuracy, realizes timely warning and rapid response to risks, and significantly improves the efficiency and decision-making speed of mine safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1This is a structural diagram of an intelligent monitoring system for underground filling quality in metal mines based on the Internet of Things according to the present invention;

[0040] Figure 2 This is a flow chart of the method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things described in the present invention. DETAILED DESCRIPTION

[0041] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 making creative efforts shall fall within the scope of protection of the present invention.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific solution for intelligent monitoring of underground filling material quality in metal mines based on the Internet of Things provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] Refer to the attached Figure 1 , which shows a structural diagram of an intelligent monitoring system for underground filling quality in metal mines based on the Internet of Things provided by one embodiment of the present invention. The system includes the following parts:

[0045] Sensor data acquisition module, data calibration module, data fusion module, feature extraction module, risk prediction module, and early warning push module;

[0046] Sensor data acquisition module: responsible for deploying various types of sensors in key monitoring areas of underground filling bodies in metal mines. Each sensor collects raw measurement values ​​in real time at a fixed frequency and outputs the raw measurement values ​​to the data calibration module;

[0047] Data calibration module: This module automatically calibrates the raw measurement values ​​from the sensor data acquisition module using sensor cluster mutual verification technology based on Kalman filtering. During the automatic calibration process, only the raw measurement values ​​that have been detected to be drifting or invalid are inter-calibrated and repaired to obtain the calibrated measurement values, which are then output to the data fusion module.

[0048] Data fusion module: Based on the Dempster-Shafer evidence theory, it fuses the calibrated measurement values ​​of the same type of sensors from the data calibration module to obtain different types of sensor fusion data, and outputs the different types of sensor fusion data to the feature extraction module;

[0049] Feature extraction module: This module takes the sensor fusion data from the data fusion module as input, constructs a time series through time windows, and uses a lightweight long short-term memory network to extract key features of filling quality degradation, generating low-dimensional feature vectors. This low-dimensional feature vector is then output to the risk prediction module.

[0050] Risk prediction module: Based on the low-dimensional feature vector of the feature extraction module, the feature selection weighted random forest prediction algorithm is used to predict the risk probability of the quality of the underground filling body of the metal mine, and the risk probability of the quality of the underground filling body of the metal mine is output to the early warning push module;

[0051] Early warning push module: Compares the risk probability of the quality of underground filling materials in metal mines from the risk prediction module with the preset risk threshold, divides the risk of the quality of underground filling materials in metal mines into three risk levels: yellow (low risk), orange (medium risk), and red (high risk), obtains the early warning level, and generates the corresponding reinforcement plan. The early warning level and reinforcement plan are pushed to the user terminal in real time through the WebSocket protocol, such as the VR interface or mobile device, to support on-site personnel in making quick decisions.

[0052] Refer to the attached Figure 2 , which shows a flow chart of an intelligent monitoring method for underground filling material quality in a metal mine based on the Internet of Things provided by one embodiment of the present invention. The method includes the following steps:

[0053] S1. Build an adaptive hybrid transmission network, dynamically select the optimal transmission path by quantifying path performance, and deploy different types of sensors to collect raw measurement values ​​in real time. Utilize sensor cluster mutual verification technology based on Kalman filtering to automatically calibrate the received raw measurement values. During the automatic calibration process, only the raw measurement values ​​detected to be drifted or invalid are inter-verified and repaired to obtain the calibrated measurement values.

[0054] Build an adaptive hybrid transmission network consisting of narrowband Internet of Things (NB-IoT) and wireless mesh network (Mesh network) to support data transmission in the intelligent monitoring system of underground filling quality in metal mines;

[0055] Based on an existing game theory model, the optimal transmission path is dynamically selected by quantifying path performance. This path performance is quantified by integrating signal strength, energy consumption, and bandwidth to ensure communication reliability in complex underground metal mine environments (such as metal interference and curved tunnels).

[0056] Deployment of key monitoring areas of underground filling bodies in metal mines, such as stress concentration areas and crack prone areas Types of sensors, each type of sensor contains The sensors collect raw measurements in real time at a fixed sampling frequency (e.g., once per second). The sensor types include fiber grating sensors for measuring strain and temperature, ultrasonic flaw detectors for detecting internal cracks, microseismic sensors for capturing stress waves, and humidity sensors;

[0057] Furthermore, the sensor cluster mutual verification technology based on Kalman filtering is used to automatically calibrate the received raw measurement values. During the automatic calibration process, only the raw measurement values ​​that are detected to be drifted or invalid are inter-calibrated and repaired to obtain the calibrated measurement values.

[0058] Specifically, the difference between the original measurement value of each sensor and the state estimation value obtained by iteratively calculating the original measurement value of the sensor through the existing Kalman filter algorithm is calculated. If the difference between the two exceeds the error threshold set based on the sensor accuracy experiment, the original measurement value is determined to be a drifted or failed original measurement value. For the sensor detected to be drifted or failed, the state estimation value of the same type of sensor excluding itself is used to calculate the average value as the calibrated measurement value; conversely, if the difference between the two is less than or equal to the error threshold set based on the sensor accuracy experiment, the original measurement value is determined to be normal or used as the calibrated measurement value. The formula is expressed as follows:

[0059]

[0060] in, Indicates time No. Type sensor The measured value of each sensor after calibration; Indicates the Number of sensors of type; Indicates the Type sensor except All sensors except the first sensor are summed; Indicates time No. Type sensor The state estimation value of each sensor is obtained by iteratively calculating the original measurement value of the sensor through the existing Kalman filter algorithm. , the state estimate is initialized to the first raw measurement value of the sensor; Indicates time No. Type sensor The raw measurement values ​​of each sensor; Indicates time No. Type sensor The state estimation value of each sensor is obtained by iteratively calculating the original measurement value of the sensor through the existing Kalman filter algorithm. , the state estimate is initialized to the first raw measurement value of the sensor; Indicates time No. Type sensor The raw measurement values ​​of the sensors are compared with the No. Type sensor The absolute difference of the state estimates of the sensors; represents the error threshold, which is set based on sensor accuracy experiments;

[0061] The sensor cluster mutual verification technology based on Kalman filtering calculates the calibrated sensor measurement values, significantly improving data accuracy and reliability, reducing operation and maintenance costs, and enhancing the fault tolerance and real-time performance of the intelligent monitoring system for underground filling quality in metal mines.

[0062] S2. Based on the Demster-Schafer evidence theory, the calibrated measurement values ​​of sensors of the same type are fused to obtain fused data of sensors of different types. The fused data of different types of sensors are used as input to construct a time series through a time window, and the key features of filling quality degradation are extracted to generate a low-dimensional feature vector.

[0063] Based on the Dempster-Shafer evidence theory, the calibrated measurement values ​​of sensors of the same type are fused to obtain fusion data of sensors of different types.

[0064] The measured value after calibration of the sensor is used as evidence, and the degree of support of the evidence for various possible evidence states is quantified by basic probability assignment, reflecting the reliability and uncertainty of the evidence;

[0065] Specifically, a basic probability assignment is assigned to each calibrated measurement value of each type of sensor to evaluate the reliability of each calibrated measurement value. The identification framework is defined as including two states: "reliable" and "unreliable". The uncertain state is not considered, that is, the probability assignment of uncertainty is zero. For each sensor, the basic probability assignment is calculated based on the absolute deviation of the sensor's calibrated measurement value from its mean. The smaller the absolute deviation, the higher the reliability. The standard deviation is obtained by dividing it by the standard deviation of the sensor's calibrated measurement value. The basic probability assignment of the reliable state is then generated by adding one and taking the inverse transformation. The formula is expressed as follows:

[0066]

[0067] in, Indicates time No. Type sensor The measured values ​​of the sensors after calibration are judged to be reliable ( ) basic probability assignment; Indicates time No. The mean of the measured values ​​after calibration of the sensor type; Indicates time No. The standard deviation of the measured values ​​after calibration of the type sensor; Indicates time No. Type sensor The measured values ​​of the sensors after calibration are compared with the values ​​at time No. Type Absolute deviation from the mean of the measured values ​​after calibration of the sensor;

[0068] The basic probability assignment of the unreliable state is 1 minus the basic probability assignment of the reliable state, which is expressed as follows:

[0069]

[0070] in, Indicates time No. Type sensor The measured values ​​of the sensors after calibration are judged to be unreliable ( ) basic probability assignment;

[0071] Based on the Dempster combination rule of the Dempster-Shafer evidence theory, the basic probability assignments of sensors of the same type are fused to calculate the initial trust weight of each sensor.

[0072] The Dempster combination rule generates an initial trust weight by accumulating all evidence combinations that support the "reliable" status and eliminating the influence of conflicting evidence. The formula is as follows:

[0073]

[0074] in, Indicates time No. Type sensor The initial trust weight of each sensor; It means summing up all the combinations of evidence states whose intersection is a reliable state; and Indicates the status of evidence; Indicates time No. Type sensor Sensor evidence status The basic probability assignment of ; Indicates that at time No. Type sensor Sensor evidence status The basic probability assignment of ; It means summing up all the evidence state combinations whose intersection is empty;

[0075] Using the normalized trust weight, the calibrated measurement values ​​of the same type of sensors are weighted averaged to generate fusion data of different types of sensors. The formula is as follows:

[0076]

[0077] in, Indicates time No. Type of sensor fusion data; Indicates the All sensors of the same type are summed to aggregate their weighted contributions; represents the normalized trust weight, which is used to adjust the contribution of each sensor's calibrated measurement value;

[0078] By evaluating the reliability of each sensor data, dynamically generating trust weights, and fusing data from similar sensors, errors caused by a single sensor are eliminated, ensuring that the sensor fusion data accurately reflects the true physical state of the filling.

[0079] Taking the fused data of different types of sensors as input, a time series is constructed through time windows. A lightweight long short-term memory network is used to extract key features of filling quality degradation, such as strain trends and crack growth rates, to generate low-dimensional feature vectors.

[0080] S3. Predict the risk probability of underground filling quality in metal mines using a feature selection weighted random forest prediction algorithm based on low-dimensional feature vectors. Compare the risk probability of underground filling quality in metal mines with the risk threshold, perform graded warnings, obtain warning levels, and generate corresponding reinforcement plans. The warning levels and reinforcement plans are pushed to the user terminal in real time.

[0081] Based on low-dimensional feature vectors, the risk probability of underground filling quality in metal mines is predicted by feature selection weighted random forest prediction algorithm.

[0082] The feature selection weighted random forest prediction algorithm is based on the ensemble learning idea of ​​the classic random forest algorithm, combining feature selection and weighting mechanism to optimize the prediction method of the traditional decision tree;

[0083] Traditional random forests treat all features equally, which may cause noise features to interfere with the risk probability of underground filling quality in metal mines. The feature selection weighted random forest prediction algorithm introduces a feature importance vector and uses the Gini index to evaluate the contribution of features in low-dimensional feature vectors to the risk prediction of underground filling quality in metal mines. It then selects highly important features and, based on feature selection theory, reduces the impact of irrelevant features, thereby improving sensitivity to changes in key physical quantities.

[0084] The feature importance vector assigns an importance weight to each selected feature, reflecting the relative importance of the feature to the risk prediction of underground filling mass in metal mines. This can highlight the role of key features and reduce the interference of secondary features, thereby improving prediction accuracy.

[0085] Based on the theory of logistic regression, the prediction output of each decision tree is passed through The function is mapped to the range of 0 to 1 and the formula is as follows:

[0086]

[0087] in, Indicates time Risk probability of filling quality in underground metal mines; represents the averaging factor, which is used to average the predictions of all decision trees; represents the number of decision trees in the random forest; Indicates the sum of the prediction outputs of all decision trees; Indicates the The prediction output of a decision tree; Represents the activation function, which is used to map the linear combination to ; Represents a set of selected feature indexes Perform summation; Indicates the The feature index set selected by the decision tree is a feature index set selected based on the importance weight sorting, and the number of features is set to 5; Indicates the The first decision tree The importance weight of each feature is calculated by the Gini index; Indicates time The low-dimensional feature vector Features Indicates the Bias term of a decision tree;

[0088] Compare the risk probability of the underground filling quality of metal mines with the risk threshold preset based on expert experience, conduct graded warning, obtain the warning level, and generate the corresponding reinforcement plan;

[0089] The graded warning formula is expressed as follows:

[0090]

[0091] in, Indicates the warning level. Yellow, orange, and red reflect the different degrees of severity of the risk, corresponding to low risk (no immediate intervention required), medium risk (needs attention and preparation for reinforcement), and high risk (needs immediate reinforcement measures). Furthermore, the warning level and reinforcement plan are pushed to the user terminal in real time via the WebSocket protocol. The push results can be displayed through a VR interface or mobile device, facilitating quick decision-making by on-site personnel.

[0092] In summary, an intelligent monitoring system and method for underground filling quality in metal mines based on the Internet of Things has been completed.

[0093] The order in which the embodiments of the invention are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0094] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. An intelligent monitoring method for underground filling quality of metal mines based on the Internet of Things, characterized in that: The following steps are involved: S1. Build an adaptive hybrid transmission network, dynamically select the optimal transmission path by quantifying path performance, and deploy different types of sensors to collect raw measurement values ​​in real time. Leveraging sensor cluster inter-calibration technology based on Kalman filtering, the received raw measurement values ​​are automatically calibrated. During the automatic calibration process, only raw measurement values ​​detected to be drifted or invalid are inter-calibrated and repaired to obtain calibrated measurement values. S2. Based on the Demster-Schafer evidence theory, the calibrated measurements of sensors of the same type are fused to obtain fused data from sensors of different types. This fused data is used as input to construct a time series using a time window. Key features of filling quality degradation are extracted to generate a low-dimensional feature vector. S3. Based on low-dimensional feature vectors, the risk probability of the quality of underground filling materials in metal mines is predicted through feature selection and weighted random forest prediction algorithm. The risk probability of the quality of underground filling materials in metal mines is compared with the risk threshold, and a graded warning is performed to obtain the warning level. The corresponding reinforcement plan is generated and pushed to the user terminal in real time.

2. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 1 is characterized in that: Said S1 specifically includes: In the implementation process of the sensor cluster mutual verification technology based on Kalman filtering, the difference between the original measurement value of each sensor and the state estimation value obtained by iteratively calculating the original measurement value of the sensor is calculated. When the difference between the two exceeds the error threshold, the original measurement value is judged as a drifted or failed original measurement value. For sensors that are detected to be drifted or failed, the state estimation values ​​of sensors of the same type excluding themselves are used to calculate the average value as the calibrated measurement value.

3. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 2 is characterized in that: Said S1 specifically includes: In the implementation process of the sensor cluster mutual verification technology based on Kalman filtering, when the difference between the two is less than or equal to the error threshold, the original measurement value is judged to be normal and used as the calibrated measurement value.

4. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 1 is characterized in that: Said S2 specifically includes: In the process of fusing calibrated measurement values ​​of sensors of the same type, the calibrated measurement values ​​of the sensors are used as evidence, and the support degree of the evidence for various evidence states is quantified through basic probability assignment.

5. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 4 is characterized in that: Said S2 specifically includes: In the process of fusing calibrated measurements of sensors of the same type, a basic probability assignment is assigned to each calibrated measurement of each type of sensor. For each sensor, the basic probability assignment is calculated based on the absolute deviation of the calibrated measurement value of the sensor and its mean, and divided by the standard deviation of the calibrated measurement value of the sensor to obtain the standardized deviation. The basic probability assignment of the reliable state is generated by adding one and taking the inverse transformation. The basic probability assignment of the unreliable state is one minus the basic probability assignment of the reliable state.

6. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 5 is characterized in that: Said S2 specifically includes: The basic probability assignments of sensors of the same type are fused. By accumulating all combinations of evidence supporting the "reliable" state and eliminating the influence of conflicting evidence, the initial trust weight of each sensor is calculated. The normalized trust weight is then used to perform a weighted average of the calibrated measurements of sensors of the same type to generate fused data from different sensor types.

7. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 1 is characterized in that: Said S3 specifically includes: In the implementation process of the feature selection weighted random forest prediction algorithm, a feature importance vector is introduced. The feature importance vector assigns an importance weight to each selected feature, reflecting the relative importance of the feature to the quality risk prediction of underground filling in metal mines.

8. The method for intelligent monitoring of underground filling quality in metal mines based on the Internet of Things according to claim 7 is characterized in that: Said S3 specifically includes: In the implementation process of the feature selection weighted random forest prediction algorithm, the prediction output of each decision tree is mapped through the Sigmoid function, and the prediction outputs of all decision trees are summed and combined with the average factor to obtain the risk probability of the quality of underground filling in metal mines.

9. An intelligent monitoring system for underground filling material quality of metal mines based on the Internet of Things, applied to the intelligent monitoring method for underground filling material quality of metal mines based on the Internet of Things according to claim 1, characterized in that: Includes the following sections: Sensor data acquisition module, data calibration module, data fusion module, feature extraction module, risk prediction module, and early warning push module; Sensor data acquisition module: deploys different types of sensors, each of which collects raw measurement values ​​in real time at a fixed frequency and outputs the raw measurement values ​​to the data calibration module; Data calibration module: This module automatically calibrates the raw measurement values ​​from the sensor data acquisition module using sensor cluster mutual verification technology based on Kalman filtering. During the automatic calibration process, only the raw measurement values ​​that have been detected to be drifting or invalid are inter-calibrated and repaired to obtain the calibrated measurement values, which are then output to the data fusion module. Data fusion module: Based on the Demster-Schafer evidence theory, the calibrated measurement values ​​of the same type of sensors from the data calibration module are fused to obtain different types of sensor fusion data, and the different types of sensor fusion data are output to the feature extraction module; Feature extraction module: This module takes the sensor fusion data from the data fusion module as input, constructs a time series through time windows, and uses a lightweight long short-term memory network to extract key features of filling quality degradation, generating low-dimensional feature vectors. This low-dimensional feature vector is then output to the risk prediction module. Risk prediction module: Based on the low-dimensional feature vector of the feature extraction module, the feature selection weighted random forest prediction algorithm is used to predict the risk probability of the quality of the underground filling body of the metal mine, and the risk probability of the quality of the underground filling body of the metal mine is output to the early warning push module; Early warning push module: Compares the risk probability of the quality of the underground filling body of the metal mine from the risk prediction module with the preset risk threshold, divides the risk of the quality of the underground filling body of the metal mine into three risk levels: yellow, orange, and red, obtains the early warning level, generates the corresponding reinforcement plan, and pushes the early warning results and reinforcement plan to the user terminal in real time to support on-site personnel in making quick decisions.

Images