A mine multi-sensor data acquisition, cleaning and fault identification system

By designing a mine multi-sensor data acquisition, cleaning and fault determination system, the problem of multi-sensor data management and perceived data interruption in the mine is solved, data continuity and accuracy are achieved, and sensor reliability and fault detection are ensured.

CN115567562BActive Publication Date: 2024-11-12NORTHEASTERN UNIV CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211338736.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-10-28
Publication Date
2024-11-12
Estimated Expiration
2042-10-28

AI Technical Summary

Technical Problem

The differences in communication protocols of multiple sensors in mines lead to the inability to achieve unified management. The harsh environment affects data accuracy, which may lead to interruption of perceived data, affecting the continuity and accuracy of the perceived environment of the mine.

Method used

Design a mine multi-variable sensor data acquisition, cleaning and fault determination system, configure the data acquisition subsystem in the cloud to collect rock mechanics response information in real time, build a dynamic queue for data classification, and combine the principle of binary classification and dynamic time window to realize data cleaning and fault detection, using cloud database for data storage and transmission.

Benefits of technology

One-click access and unified management of multi-sensor data is realized, the continuity and accuracy of perceived data is ensured, and the reliability of sensor data is guaranteed to the greatest extent through automated detection and alarm mechanisms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115567562B_ABST
    Figure CN115567562B_ABST
Patent Text Reader

Abstract

The present invention provides a mine multi-sensor data collection, cleaning and fault discrimination system, which belongs to the technical field of real-time intelligent perception of mine engineering environment; it comprises: a data collection subsystem for cloud configuration of rock mass mechanical response monitoring equipment information and real-time collection and storage of mechanical response information in the mine production process; a data collection subsystem based on the monitoring equipment information configuration, real-time collection of rock mass mechanical response information data in the mine production process, building a dynamic queue, realizing queue data classification according to the binary classification principle, and judging the abnormal state of the latest collected monitoring data according to the classification result of the tail element of the queue; a sensor fault detection subsystem which adopts the form of a dynamic time window and uses a data threshold or a data acquisition time interval to realize the judgment of the fault state in the operation of the sensor.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of real-time intelligent perception of mining engineering environment, and in particular to a system for data collection, cleaning and fault identification of multiple sensors in a mine. Background Art

[0002] In recent years, the vigorous development of disruptive technologies such as 5G, Internet+, cloud computing, artificial intelligence, and big data has ushered in the fourth industrial revolution, accelerated the transformation of traditional industries such as mining engineering, and enabled the transformation and upgrading of the mining industry with intelligent technologies. The digitalization and intelligent transformation of mines has become an inevitable trend in the development of mines.

[0003] Installing diversified sensors in mines to achieve real-time perception of the environment is the basis for the digital and intelligent transformation of mines. At present, a large number of mines in China have installed diversified monitoring equipment, which has initially realized real-time perception of the mine environment, but there are still some problems. First, due to the differences in the communication protocols of various devices, it is impossible to achieve unified management of multi-heterogeneous data; secondly, the harsh environment of the mine will affect the data collected by the sensor, making it impossible for the sensor to accurately obtain real environmental data, and it is more likely to have the problem of perception data interruption. The existence of these problems will have a significant impact on the continuity and accuracy of mine environmental perception, and affect the process of digitalization and intelligent transformation of mines. It can be seen that it is very necessary to develop a method and system for data collection, cleaning and fault identification of multi-sensor sensors in mines, realize one-click access and unified management of multi-sensor data, intelligent cleaning of monitoring data and automatic detection of sensor anomalies. Summary of the invention

[0004] In order to solve the above problems, a mine multi-sensor data collection, cleaning and fault identification system is proposed. The specific technical solution of the present invention is as follows:

[0005] A mine multi-sensor data collection, cleaning and identification system, comprising:

[0006] A data acquisition subsystem for cloud configuration of rock mass mechanical response monitoring equipment information and real-time acquisition and storage of mechanical response information during mining production;

[0007] The data acquisition subsystem is configured based on the monitoring equipment information. The rock mass mechanical response information data collected in real time during the mining production process is constructed to build a dynamic queue. According to the binary classification principle, the queue data is classified. According to the classification results of the tail elements of the queue, the abnormal state of the latest collected monitoring data is determined, and the multi-sensor abnormal data cleaning subsystem is completed to clean the data of the multi-sensor.

[0008] Based on the dynamic update of the operating status of the multi-sensor by the cleaning subsystem, a sensor fault detection subsystem is implemented to judge the fault status of the sensor during operation in the form of a dynamic time window and a data threshold or a data acquisition time interval.

[0009] Furthermore, the data acquisition subsystem comprises:

[0010] Multi-environment sensing equipment used for real-time perception of information such as rock mass mechanical response during mining production;

[0011] A cloud database used for persistent storage of basic sensor information and diversified monitoring data, where the basic sensor information table includes the transmission protocol, measurement point mapping code, sensor name, monitoring area, person in charge, sampling cycle, latest sampling time, number of abnormal sampling, and monitoring data storage location;

[0012] The diversified monitoring data table mainly stores the sensor number, data collection time, collection data and data status;

[0013] It is used to visualize and persistently save the sensor's data transmission protocol, measurement point mapping code, sensor sampling period, monitoring area, person in charge and other parameters to the sensor basic information table. On this basis, the perception data of multiple sensors is stored in real time to the above-mentioned diversified monitoring data table to realize the data transmission parameter configuration cloud service of real-time monitoring data to the cloud database.

[0014] Furthermore: the multi-sensor is configured in the cloud through a data transmission protocol to achieve real-time transmission of multi-sensor data.

[0015] Further: the data acquisition subsystem configured based on the monitoring equipment information collects the rock mass mechanical response information data in the mining production process in real time, constructs a dynamic queue, implements queue data classification according to the binary classification principle, and determines the abnormal state of the latest collected monitoring data based on the classification structure of the tail element of the queue. The process of completing the data cleaning of the multi-sensor is as follows:

[0016] Step 1.1, read the measurement point mapping code and real-time data storage location field in the sensor definition table to build a sensor definition matrix;

[0017] Step 1.2, traverse the sensor definition matrix, read m normal data in the specified data table as initial data, and form a dynamic queue matrix, where m is the dynamic queue length, which can be specified independently;

[0018] Step 1.3, start real-time data collection, extract the measurement point mapping code and perception data in the data for the latest transmission data collected, and use the mapping code as the search condition to find the index for this sensor in the sensor definition matrix, and push the perception data into the dynamic queue matrix index queue;

[0019] Step 1.4, find the maximum and minimum values ​​in the dynamic queue matrix index queue, generate two unequal random center points k1 and k2 between the maximum and minimum values ​​as the initial center points, calculate the Manhattan distance of each element in the queue from the center points k1 and k2, and divide the queue elements into two categories based on the minimum Manhattan distance to obtain the classification result;

[0020] Step 1.5, combine the classification results and generate new center points k1', k2' according to the midpoint calculation formula. When k1', k2' are completely equal to k1, k2, go to step 1.6, otherwise assign the values ​​of k1', k2' to k1, k2 respectively, and return to step 1.4;

[0021] Step 1.6, according to the classification results, calculate the number of elements in the category of the latest monitoring data, so as to determine the status of the latest collected monitoring data. If the number of elements in the category of the tail element is less than k, it is an abnormal state and it is removed from the queue matrix index queue, otherwise it is a normal state; the head element of the queue matrix index queue is dequeued and the monitoring data is queued to realize dynamic update of the time window; k is the evaluation threshold of the abnormal state of the monitoring data. When k=1, it means that the data is in a completely abnormal state. As k increases, the judgment standard of abnormal data becomes lower.

[0022] Furthermore: the value of the abnormal data is between [1, 4], and the optimal value of the abnormal data is 2.

[0023] Further: it also includes obtaining the current latest data collection time to update the latest sampling time data in the sensor definition table;

[0024] When it is determined that the monitoring data is in an abnormal state, the number of abnormal samples in the sensor definition table is updated to the original number plus one, and the monitoring data is stored in the data table recorded in the queue matrix index queue, and its deleted field is set to 1, indicating that it is in a logically deleted state;

[0025] When it is determined that the monitoring data is in a normal state, the data is stored in the data table recorded in the queue matrix index queue, and its deleted field is set to 0.

[0026] Further: the process of dynamically updating the operating status of the multi-sensor based on the cleaning subsystem, using a dynamic time window, and using a data threshold or a data acquisition time interval to determine the fault status of the sensor during operation is as follows:

[0027] Step 2.1, set the dynamic time window, use the timed task method to dynamically read the data of the person in charge, sampling cycle, latest sampling time, and abnormal sampling quantity in the sensor definition table;

[0028] Step 2.2, based on the latest sampling time and sampling cycle data obtained dynamically, combined with the current time, calculate the interval between the last sampling time and the current time. When this interval exceeds n times the sampling cycle, it means that the sensor is disconnected. Where n is a user-specified parameter used to characterize the sensitivity of the system's sensor disconnection detection;

[0029] Step 2.3, based on the abnormal sampling quantity data in step 2.1, determine whether it is greater than the abnormal sampling threshold k. If it is greater, it means that the sensor has data collection abnormality, where k is a user-defined parameter used to characterize the sensitivity of abnormal data detection.

[0030] Step 2.4, based on the judgment results of step 2.2 and step 2.3, when the sensor is detected to be disconnected or the data collection is abnormal, combined with the information of the person in charge obtained in step 2.1, send SMS and email alarm information to inform them to check and repair the sensor, and update the number of abnormal sampling in the sensor definition to 0.

[0031] The beneficial effects of the mine multi-sensor data collection, cleaning and fault identification system of the present invention are:

[0032] 1) The present invention uses cloud visualization to configure information related to multi-sensor data transmission, which can achieve one-click access and unified management of multi-environment perception data and provide a continuous data source for subsequent data cleaning;

[0033] 2) The present invention adopts a dynamic queue and a binary classification method to realize the abnormal state of real-time sensing data for judgment, which can realize the intelligent elimination of abnormal data and the automatic update of sensor status, ensure the reliability of sensing data, and provide a basis for sensor fault detection;

[0034] 3) The present invention uses a timed task to complete the construction of a dynamic time window, completes the detection of sensor fault status, and realizes the function of sending sensor abnormal working status alarm information to the person in charge by email or text message, thereby ensuring the continuity of sensor data collection to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0036] Figure 1 This is a system architecture diagram of data collection, cleaning and fault identification of multiple sensors in a mine according to the present invention;

[0037] Figure 2 This is a flow chart of data cleaning in an example of the present invention;

[0038] Figure 3 This is an interface diagram of a data acquisition subsystem according to an embodiment of the present invention;

[0039] Figure 4 Real-time collection of background log graphs for the example data of the present invention;

[0040] Figure 5 This is a comparison chart of the effects before and after data cleaning of the example of the present invention;

[0041] Figure 6 This is a sensor fault alarm diagram of the present invention;

[0042] Figure 7 Schematic diagram of original data in the classification method used in the present invention;

[0043] Figure 8 This is a schematic diagram of the processing results of the binary classification method used in the present invention. DETAILED DESCRIPTION

[0044] In order to more clearly illustrate the technical solution and implementation examples of the present invention, the algorithms in the embodiments or prior arts are introduced.

[0045] Figure 1 This is a system architecture diagram of data collection, cleaning and fault identification of multiple sensors in a mine according to the present invention;

[0046] A mine multi-sensor data collection, cleaning and identification system, characterized by: comprising:

[0047] A data acquisition subsystem for cloud configuration of rock mass mechanical response monitoring equipment information and real-time acquisition and storage of mechanical response information during mining production;

[0048] After the data acquisition subsystem is configured based on the monitoring equipment information, the rock mass mechanical response information data collected in real time during the mining production process is constructed to build a dynamic queue. According to the binary classification principle, the queue data is classified. According to the classification results of the tail element of the queue, the abnormal state of the latest collected monitoring data is determined;

[0049] The data acquisition subsystem shown collects only one piece of data at a time, and determines the abnormal state of this piece of data. In simple terms, it determines whether the latest data is faulty through the previous historical data, completes the data cleaning of the multi-sensor, and obtains the multi-sensor abnormal data cleaning subsystem;

[0050] After configuring the monitoring equipment information above, you can sense the rock mechanics information regularly, and then build a dynamic queue based on this real-time rock mechanics information. For example, if a piece of data is sensed at 12:00, put the data into the dynamic queue at 12:00, and then analyze the queue to determine whether the data at 12:00 is abnormal.

[0051] Based on the dynamic update of the operating status of the multi-sensor by the cleaning subsystem, a sensor fault detection subsystem is implemented to judge the fault status of the sensor during operation in the form of a dynamic time window and a data threshold or a data acquisition time interval.

[0052] Based on the dynamics of the cleaning subsystem, for example, at 12:00, we get whether the data at 12 o'clock is normal. If it is normal, we store the monitoring data in the database. If it is abnormal data, we store the data in the database with deleted as 1, and add 1 to the number of abnormalities, indicating that the data is abnormal. Regardless of whether it is abnormal or not, we update the operating status information of the monitoring equipment, mainly updating the last sampling time field, and updating it with the time of the latest collected data.

[0053] Then, the continuous operation of this cleaning system will continuously update the status of the multi-sensor, and then the sensor failure will be judged based on the last sampling time of the multi-sensor and the amount of abnormal data collected.

[0054] The data acquisition subsystem includes: a multi-environment sensing device for real-time sensing of information such as rock mass mechanical response during mining production; the multi-environment sensing device includes a multi-point displacement meter, an anchor dynamometer, a deep hole inclinometer, etc.;

[0055] A cloud database used for persistent storage of basic sensor information and diversified monitoring data, where the basic sensor information table includes the transmission protocol, measurement point mapping code, sensor name, monitoring area, person in charge, sampling cycle, latest sampling time, number of abnormal sampling, and monitoring data storage location;

[0056] The diversified monitoring data table mainly stores the sensor number, data collection time, collection data and data status;

[0057] It is used to visualize and persistently save the sensor's data transmission protocol, measurement point mapping code, sensor sampling period, monitoring area, person in charge and other parameters to the sensor basic information table. On this basis, the perception data of multiple sensors is stored in real time to the above-mentioned diversified monitoring data table to realize the data transmission parameter configuration cloud service of real-time monitoring data to the cloud database.

[0058] The multi-sensor is configured in the cloud through a data transmission protocol to achieve real-time transmission of multi-sensor data.

[0059] Further: Based on the real-time multivariate data obtained, a dynamic queue is constructed, and queue data classification is implemented according to the binary classification principle. According to the classification structure of the tail element of the queue, the data in the latest collected monitoring data is identified as abnormal state, and the process of completing the data cleaning of the multivariate sensor is as follows:

[0060] Step 1.1, when the service is started for the first time, read the measurement point mapping code and real-time data storage location field in the sensor definition table, and build the sensor definition matrix SensorList;

[0061] Step 1.2, traverse the sensor definition matrix SensorList, read m normal data in the specified data table as initial data, and form a dynamic queue matrix SensorQueue, where m is the dynamic queue length, which can be specified independently;

[0062] Step 1.3, start real-time data collection, extract the measurement point mapping code and perception data in the data for the latest transmission data collected, and use this as the search condition to find the index index for this sensor in the sensor definition matrix SensorList, and push the perception data into the dynamic queue matrix index queue SensorQueue[index];

[0063] Step 1.4, find the maximum and minimum values ​​in the dynamic queue matrix index queue SensorQueue[index], generate two unequal random center points k1 and k2 between the maximum and minimum values ​​as the initial center points, calculate the Manhattan distance of each element in the queue from the center points k1 and k2, and divide the queue elements into two categories based on the minimum Manhattan distance to obtain the classification result;

[0064] Step 1.5, combine the classification results and generate new center points k1', k2' according to the midpoint calculation formula. When k1', k2' are completely equal to k1, k2, go to step 1.6, otherwise assign the values ​​of k1', k2' to k1, k2 respectively, and return to step 1.4;

[0065] Step 1.6, based on the classification results, calculate the number of elements in the category of the latest monitoring data, so as to determine the status of the latest collected monitoring data. If the number of elements in the category of the tail element is less than k, it is an abnormal state and it is removed from the queue matrix index queue SensorList[index]. Otherwise, it is a normal state. The first element of the queue matrix index queue SensorList[index] is dequeued and the monitoring data is queued to realize dynamic update of the time window. k is the evaluation threshold of the abnormal state of the monitoring data. When k=1, it means that the data is in a completely abnormal state. As k increases, the evaluation standard of abnormal data becomes lower.

[0066] The value of the abnormal data is between [1, 4], and the optimal value of the abnormal data is 2.

[0067] The method also includes obtaining the current latest data collection time to update the latest sampling time data in the sensor definition table;

[0068] When it is determined that the monitoring data is in an abnormal state, the number of abnormal samples in the sensor definition table is updated to the original number plus one, and the monitoring data is stored in the data table recorded in the queue matrix index queue SensorList[index], and its deleted field is set to 1, indicating that it is in a logically deleted state;

[0069] When it is determined that the monitoring data is in a normal state, the data is stored in the data table recorded in the queue matrix index queue SensorList[index], and its deleted field is set to 0.

[0070] Figure 2 This is a flow chart of data cleaning in an example of the present invention;

[0071] Further: the process of dynamically updating the operating status of the multi-sensor based on the cleaning subsystem, using a dynamic time window, and using a data threshold or a data acquisition time interval to determine the fault status of the sensor during operation is as follows:

[0072] Step 2.1, set the dynamic time window, use the timed task method to dynamically read the data of the person in charge, sampling cycle, latest sampling time, and abnormal sampling quantity in the sensor definition table;

[0073] Step 2.2, based on the latest sampling time and sampling cycle data obtained dynamically, combined with the current time, calculate the interval between the last sampling time and the current time. When this interval exceeds n times the sampling cycle, it means that the sensor is disconnected. Where n is a user-specified parameter used to characterize the sensitivity of the system's sensor disconnection detection;

[0074] Step 2.3, based on the abnormal sampling quantity data in step 2.1, determine whether it is greater than the abnormal sampling threshold k. If it is greater, it means that the sensor has data collection abnormality, where k is a user-defined parameter used to characterize the sensitivity of abnormal data detection.

[0075] Step 2.4, based on the judgment results of steps 2.2 and 2.3, when the sensor is detected to be disconnected or in abnormal data collection state, combined with the information of the person in charge obtained in step 4.1, send SMS and email alarm information to inform them to check and repair the sensor, and update the number of abnormal sampling in the sensor definition to 0.

[0076] Example

[0077] This example uses JTM-V7000 series multi-point displacement meters and JTM-V1000B series vibrating-string anchor dynamometers from Jintu as environmental sensing devices, and uses cloud-based configuration of sensor parameters to achieve real-time cloud uploading of sensing data, intelligent cleaning of sensing data, and automated detection of sensor faults.

[0078] A method for cleaning data of multi-sensor in mines and a method for fault identification realizes one-key access to sensor monitoring data, intelligent cleaning of monitoring data, and automatic detection of sensor faults. The specific implementation plan is as follows:

[0079] Step 1: Build the data acquisition subsystem, including JTM-V7000 series multi-point displacement meters and JTM-V1000B series vibrating-wire anchor dynamometers as environmental sensing devices, cloud database, and data transmission parameter configuration cloud services.

[0080] Figure 3 This is an interface diagram of a data acquisition subsystem according to an embodiment of the present invention;

[0081] Specifically, open Figure 3 The data transmission parameter configuration cloud service is shown. Click the Add or Modify button, enter the sensor measurement point name, measurement point mapping code, database table name, solution formula and transmission protocol and other parameters, and then click the Save button to save the information to the sensor definition table, configure the sensor data transmission related indicators, and complete the monitoring data upload to the cloud;

[0082] Figure 4 Real-time collection of background log graphs for the example data of the present invention;

[0083] Figure 4It shows the background response log of monitoring data on the cloud after the cloud configuration is completed. The sensor measurement point name is often used to indicate the monitoring location of the sensor. The measurement point mapping code, solution formula and transmission protocol need to be determined after communication with the sensor device supplier. This example uses the MQTT protocol for data transmission; the database table name is used to determine the storage location of the real-time collected data, which is convenient for persistent storage of the data after it is uploaded to the cloud;

[0084] Step 2: Based on the dynamic queue and binary classification principle, intelligent judgment of abnormal status of real-time monitoring data is realized, such as Figure 3 Step 2 is as follows:

[0085] Step 2.1, when the service is started for the first time, read the measurement point mapping code and real-time data storage location field in the sensor definition table, and build the sensor definition matrix SensorList;

[0086] Step 2.2, traverse the SensorList matrix, read m normal data in the specified data table as initial data, and form a dynamic queue matrix SensorQueue, where m is the dynamic queue length, which can be specified independently;

[0087] Step 2.3, after the above program is finished running, with the help of the data transmission service configured in step 1, start real-time data collection, extract the measurement point mapping code and perception data in the data for the latest transmission data collected, and use this as the search condition to find the index index for this sensor in SensorList, and set queue D = SensorQueue[index], and push the perception data into queue D;

[0088] Step 2.4, according to formula 1-2, obtain two random and unequal queue center points k1', k2', and set k1, k2 to be equal to k1', k2' respectively;

[0089] k1′=random(min(D),max(D)) (1)

[0090] k2′=random(min(D),max(D)) (2)

[0091] Step 2.5, calculate the absolute value of the Manhattan distance between the elements in queue D and the center points k1 and k2 according to formula 3-4, and determine the category of each element in the queue according to formula 5;

[0092] dis1(i)=|k1-D(i)| (3)

[0093] dis2(i)=|k2-D(i)| (4)

[0094] category(i)=(dis1(i)<dis2(i))+2(dis2(i)<dis1(i)) (5)

[0095] Step 2.6, calculate the absolute value of the Manhattan distance between the elements in queue D and the center points k1 and k2 according to formula 3-4, and determine the category of each element in the queue according to formula 5;

[0096] dis1(i)=|k1-D(i)| (3)

[0097] dis2(i)=|k2-D(i)| (4)

[0098] category(i)=(dis1(i)<dis2(i))+2(dis2(i)<dis1(i)) (5)

[0099] Step 2.7, calculate the new classification center points k1', k2' after binary classification according to formula 6-7, and determine whether k1', k2' are equal to k1, k2. If they are equal, execute step 2.8, otherwise set k1, k2 to be equal to k1', k2' respectively, and execute step 2.5;

[0100]

[0101]

[0102] Step 2.8, calculate the abnormal state judgment variable noise of the latest sensed data according to formula 8. In this example, the abnormal state evaluation threshold k of the monitoring data is taken as 2. When noise≤2, it means that the latest sensed data is abnormal outlier data, and it is directly marked as deleted and directly removed from the queue; when noise>2, it means that the latest sensed data is in a normal state, and it is marked as a normal state, and the queue is allowed to perform the dequeue operation normally;

[0103]

[0104] Step 3: Based on the judgment result of the abnormal state of the real-time perception data in step 2, the data is persisted. Specifically, the current data collection time is obtained to update the latest sampling time data in the sensor definition table; when the monitoring data is judged to be in an abnormal state, the number of abnormal sampling in the sensor definition table is updated to the original number plus one, and the monitoring data is stored in the data table to which this type of data belongs, and its deleted field is set to 1, indicating that it is in a logically deleted state; when the monitoring data is judged to be in a normal state, the data is stored in the data table to which this type of data belongs, and its deleted field is set to 0. The final effect of this data cleaning method is as follows: Figure 5 As shown, Figure 5 This is a comparison chart of the effects before and after data cleaning of the example of the present invention;

[0105] Step 4: Use the dynamic time window to analyze and process the sensor parameters dynamically updated in step 3, analyze the sensor operation status, and if an abnormal sensor is detected, send an alarm message to the designated user in the form of SMS or email. Step 4 is as follows:

[0106] Step 4.1, set the dynamic time window to 1h, use the timed task method to dynamically read the data of the person in charge, sampling cycle, latest sampling time, and abnormal sampling quantity in the sensor definition table;

[0107] Step 4.2, based on the latest sampling time and sampling cycle data dynamically obtained in step 4.1, combined with the current time, calculate the interval between the last sampling time and the current time. When this interval exceeds 3 times the sampling cycle, it means that the sensor is in a disconnected state;

[0108] Step 4.3, based on the abnormal sampling quantity data in step 4.1, determine whether it is greater than the abnormal sampling threshold 10. If it is greater, it means that the sensor has data collection abnormality;

[0109] Step 4.4: Based on the calculation results of steps 4.2 and 4.3, when the sensor is detected as disconnected or the data collection is abnormal, combined with the information of the person in charge obtained in step 4.1, a text message and email alarm message are sent to inform him to check and repair the sensor. Figure 6 This is the sensor fault alarm diagram of the example of the present invention; at the same time, the number of abnormal sampling in the sensor definition is updated to 0.

[0110] Figure 7 Schematic diagram of original data in the classification method used in the present invention;

[0111] Figure 8 This is a schematic diagram of the processing results of the binary classification method used in the present invention.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A mine multi-sensor data collection, cleaning and discrimination system, characterized by: include: A data acquisition subsystem for cloud configuration of rock mass mechanical response monitoring equipment information and real-time acquisition and storage of mechanical response information during mining production; Based on the data acquisition subsystem after the monitoring equipment information is configured, the rock mass mechanical response information data collected in real time during the mining production process is constructed to build a dynamic queue. According to the binary classification principle, the queue data classification is realized. According to the classification results of the tail element of the queue, the abnormal state of the latest collected monitoring data is judged, and the data cleaning subsystem of the multi-sensor abnormality is completed to clean the data of the multi-sensor; According to the binary classification principle, queue data classification is implemented, and according to the classification result of the tail element of the queue, the abnormal state of the latest collected monitoring data is judged as follows: Step 2.1, when the service is started for the first time, read the measurement point mapping code and real-time data storage location field in the sensor definition table, and build the sensor definition matrix SensorList; Step 2.2, traverse the SensorList matrix, read m normal data in the specified data table as initial data, and form a dynamic queue matrix SensorQueue, where m is the dynamic queue length, which can be specified independently; Step 2.3, after the above steps are completed, with the help of data transmission service, real-time data collection is started. For the latest transmission data collected, the measurement point mapping code and perception data in the data are extracted, and the index index for this sensor in SensorList is found based on this as the search condition, and the queue D = SensorQueue[index] is set, and the perception data is pushed into the queue D; Step 2.4, according to formula 1-2, obtain two random and unequal queue center points k1', k2', and set k1, k2 to be equal to k1', k2' respectively; k1′=random(min(D),max(D)) (1) k2′=random(min(D),max(D)) (2) Step 2.5, calculate the absolute value of the Manhattan distance between the elements in queue D and the center points k1 and k2 according to formula 3-4, and determine the category of each element in the queue according to formula 5; dis1(i)=|k1-D(i)| (3) dis2 (i) = |k2 - D(i)| (4) category(i)=(dis1(i)<dis2(i)) +2(dis2(i)<dis1(i)) (5) Step 2.6, calculate the new classification center points k1', k2' after binary classification according to formula 6-7, and determine whether k1', k2' are equal to k1, k2. If they are equal, execute step 2.8, otherwise set k1, k2 to be equal to k1', k2' respectively, and execute step 2.5; Step 2.7, calculate the abnormal state judgment variable noise of the latest sensed data according to formula 8. In this example, the abnormal state evaluation threshold k of the monitoring data is taken as 2. When noise≤2, it means that the latest sensed data is abnormal outlier data, and it is directly marked as deleted and directly removed from the queue; when noise>2, it means that the latest sensed data is in a normal state, and it is marked as a normal state, and the queue is allowed to perform the dequeue operation normally; A sensor fault detection subsystem that dynamically updates the operating status of the multi-sensor based on the data cleaning subsystem, adopts a dynamic time window, and uses a data threshold or a data acquisition time interval to determine the fault status of the sensor during operation; The process of dynamically updating the operating status of the multi-sensor based on the data cleaning subsystem, using the form of a dynamic time window, and using a data threshold or a data acquisition time interval to judge the fault status of the sensor operation is as follows: Step 2.1, set the dynamic time window, use the timed task method to dynamically read the data of the person in charge, sampling cycle, latest sampling time, and abnormal sampling quantity in the sensor definition table; Step 2.2, based on the latest sampling time and sampling cycle data obtained dynamically, combined with the current time, calculate the interval between the last sampling time and the current time. When this interval exceeds n times the sampling cycle, it means that the sensor is in a disconnected state; where n is a user-specified parameter used to characterize the sensitivity of the system's sensor disconnection detection; Step 2.3, based on the abnormal sampling quantity data in step 2.1, determine whether it is greater than the abnormal sampling threshold k. If it is greater, it means that the sensor has data collection abnormality, where k is a user-defined parameter used to characterize the sensitivity of abnormal data detection; Step 2.4, based on the judgment results of step 2.2 and step 2.3, when the sensor is detected to be disconnected or the data collection is abnormal, combined with the information of the person in charge obtained in step 2.1, send SMS and email alarm information to inform them to check and repair the sensor, and update the number of abnormal sampling in the sensor definition to 0.

2. A mine multi-sensor data acquisition, cleaning and identification system according to claim 1, characterized in that: The data acquisition subsystem comprises: Multi-environment sensing equipment used for real-time perception of rock mass mechanical response information during mining production; A cloud database used for persistent storage of basic sensor information and diversified monitoring data, where the basic sensor information table includes the transmission protocol, measurement point mapping code, sensor name, monitoring area, person in charge, sampling cycle, latest sampling time, number of abnormal sampling, and monitoring data storage location; The diversified monitoring data table stores the sensor number, data collection time, collected data and data status; It is used to visualize the configuration of the sensor's data transmission protocol, measurement point mapping code, sensor sampling period, monitoring area, and responsible person parameters and persist them in the sensor basic information table. On this basis, the perception data of multiple sensors is stored in real time to the above-mentioned diversified monitoring data table to realize the data transmission parameter configuration cloud service of real-time monitoring data upload to the cloud database.

3. The mine multi-sensor data acquisition, cleaning and identification system according to claim 1, characterized in that: The multi-sensor is configured in the cloud through a data transmission protocol to achieve real-time transmission of multi-sensor data.

4. The mine multi-sensor data acquisition, cleaning and identification system according to claim 1, characterized in that: After the data acquisition subsystem is configured based on the monitoring equipment information, the rock mass mechanical response information data collected in real time during the mining production process is constructed to build a dynamic queue. According to the binary classification principle, the queue data classification is realized. According to the classification structure of the tail element of the queue, the abnormal state of the latest collected monitoring data is judged. The process of completing the data cleaning of the multi-sensor is as follows: Step 1.1, read the measurement point mapping code and real-time data storage location field in the sensor definition table to build a sensor definition matrix; Step 1.2, traverse the sensor definition matrix, read m normal data in the specified data table as initial data, and form a dynamic queue matrix, where m is the dynamic queue length, which can be specified independently; Step 1.3, start real-time data collection, extract the measurement point mapping code and perception data in the data for the latest transmission data collected, and use the mapping code as the search condition to find the index for this sensor in the sensor definition matrix, and push the perception data into the dynamic queue matrix index queue; Step 1.4, find the maximum and minimum values ​​in the dynamic queue matrix index queue, generate two unequal random center points k1 and k2 between the maximum and minimum values ​​as the initial center points, calculate the Manhattan distance of each element in the queue from the center points k1 and k2, and divide the queue elements into two categories based on the minimum Manhattan distance to obtain the classification result; Step 1.5, combine the classification results and generate new center points k1', k2' according to the midpoint calculation formula. When k1', k2' are completely equal to k1, k2, go to step 1.6, otherwise assign the values ​​of k1', k2' to k1, k2 respectively, and return to step 1.4; Step 1.6, according to the classification results, calculate the number of elements in the category of the latest monitoring data, so as to determine the status of the latest collected monitoring data. If the number of elements in the category of the tail element is less than k, it is an abnormal state and it is removed from the queue matrix index queue, otherwise it is a normal state; the head element of the queue matrix index queue is dequeued and the monitoring data is queued to realize dynamic update of the time window; k is the evaluation threshold of the abnormal state of the monitoring data. When k=1, it means that the data is in a completely abnormal state. As k increases, the judgment standard of abnormal data becomes lower.

5. The mine multi-sensor data acquisition, cleaning and identification system according to claim 1, characterized in that: The value of the abnormal data is between [1, 4].

6. The mine multi-sensor data acquisition, cleaning and identification system according to claim 1, characterized in that: The optimal value of the abnormal data is 2.

7. The mine multi-sensor data acquisition, cleaning and identification system according to claim 2, characterized in that: It also includes obtaining the current latest data collection time to update the latest sampling time data in the sensor definition table; When it is determined that the monitoring data is in an abnormal state, the number of abnormal samples in the sensor definition table is updated to the original number plus one, and the monitoring data is stored in the data table recorded in the queue matrix index queue, and its deleted field is set to 1, indicating that it is in a logically deleted state; When it is determined that the monitoring data is in a normal state, the data is stored in the data table recorded in the queue matrix index queue, and its deleted field is set to 0.