CN113094247A - Storm-based real-time prediction method for running state of coal mining machine - Google Patents

Storm-based real-time prediction method for running state of coal mining machine Download PDF

Info

Publication number
CN113094247A
CN113094247A CN202110438420.8A CN202110438420A CN113094247A CN 113094247 A CN113094247 A CN 113094247A CN 202110438420 A CN202110438420 A CN 202110438420A CN 113094247 A CN113094247 A CN 113094247A
Authority
CN
China
Prior art keywords
data
time
bolt
early warning
mining machine
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110438420.8A
Other languages
Chinese (zh)
Other versions
CN113094247B (en
Inventor
黄玉鑫
闫振国
范京道
刘睿卿
王延平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian University of Science and Technology
Original Assignee
Xian University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian University of Science and Technology filed Critical Xian University of Science and Technology
Priority to CN202110438420.8A priority Critical patent/CN113094247B/en
Publication of CN113094247A publication Critical patent/CN113094247A/en
Application granted granted Critical
Publication of CN113094247B publication Critical patent/CN113094247B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

The application relates to the technical field of coal mine safety, and discloses a Storm-based real-time prediction method for the operation state of a coal mining machine. And combining various data predicted by the prediction model GRU with the threshold value of each data based on Storm to realize early warning of the data state. And the GRU is trained from the RMSE, MAE and R after training in the training set2And evaluating the performance of the GRU in the test set respectively, and proving that the prediction model GRU can be suitable for predicting the state data of the coal mining machine. The invention breaks through the traditional monitoring method for the running state of the coal mining machine, integrates the prediction and early warning method for the real-time running state of the coal mining machine, and adopts the Storm distributed real-time processing frame in the seaThe parallel processing of the coal mine data has great practical value.

Description

Storm-based real-time prediction method for running state of coal mining machine
Technical Field
The application relates to the technical field of coal mine safety, in particular to a method for predicting the running state of a coal mining machine in real time based on Storm.
Background
The coal mining machine is one of the three fully mechanized coal mining machines, the working environment is complex, along with the development of coal mine intellectualization, the coal mining machine is provided with a plurality of sensors, the sampling frequency is high, the data collected every day is increased by PB magnitude, the change of the data at the next moment and the abnormity of the coal mining machine state are predicted through the real-time collection and analysis of the coal mining machine operation state data, the data can be effectively utilized, the safety of the coal mining machine and personnel can be guaranteed to a certain degree, and therefore the real-time monitoring and prediction of the coal mining machine state data are significant to the intellectualized operation of the coal mining.
In the process of researching the operation state of the coal mining machine in the past, simple mathematical statistical analysis is mostly carried out on monitoring data of the coal mining machine, and the research on how to predict the operation state of the coal mining machine in real time is less, so that the method has the following defects: (1) in the practical process, the conventional preprocessing method is difficult to judge whether abnormal data is useful data capable of extracting equipment state information or useless data capable of being cleaned, cannot adapt to the dynamic characteristic of time series data, and needs to establish a preprocessing cleaning model capable of adapting to the characteristics of the time series, so that the useless abnormal value of the coal mining machine can be identified and dynamically repaired; (2) when the time sequence data is predicted by traditional machine learning methods such as ARIMA, SVR, hidden Markov models and other algorithms, although the accuracy can meet the requirement, the accuracy cannot meet the requirement of large data, and the method is not easy to apply to practice. Although the deep learning method is excellent in performance in big data, the deep learning method is less applied to coal mine data, the coal mine data are complex and variable, a large amount of unreal data exist, and reasonable preprocessing is needed to train and predict through a deep learning model. (3) The coal mining machine has huge data volume, the coal mine intellectualization needs to realize the parallelization real-time processing of the data of each sensor, the existing coal mine data research is based on off-line data processing, and the data value is suddenly reduced due to the hysteresis of the data processing.
Storm is an open-source distributed real-time computing framework, can easily process infinite data flow, is widely applied to the aspects of power grids and big data, but is less used in the real-time processing of coal mine data, and is used for processing the coal mine equipment data, aiming at the problems of large data quantity of the coal mine fully-mechanized mining equipment operation state, noise, missing values and the like of the data, a MapReduce-based coal mine fully-mechanized mining equipment operation state big data cleaning model is established, but the MapReduce of Hadoop cannot meet the real-time requirement; the Caochai Steel and the like establish a data real-time cleaning platform based on Storm aiming at the problem that the running state data of the coal mining machine has noise points and missing values, but the prediction and early warning of the state data of the coal mining machine are not realized through Storm.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a method for predicting the running state of a coal mining machine in real time based on Storm.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for predicting the running state of a coal mining machine in real time based on Storm comprises the following steps:
step one, designing a data storage structure
The running state data of the coal mining machine belongs to time sequence data, and a data storage structure is designed based on a Hadoop Database distributed storage Database and comprises an index table and a data table;
the index table comprises a monitoring point position (Location), a GRU Prediction model (Prediction model) and early warning Threshold Upper and Lower bounds (Threshold Upper and Lower Bound);
the data table simulates a time sequence in a row increasing mode, each row represents sensor time sequence data acquired by a certain monitoring point, and in actual production, real-time operation data of each monitoring point is stored into a Hadoop Database at fixed time intervals, and then the Hadoop Database transfers the data into a Spout for processing through the deployment of a main control node Nimbus;
step two, Spout design
Simulating flow data collected in production through a Hadoop Database distributed storage Database, transmitting the flow data to different message queues Spout, extracting information of corresponding monitoring points from the data stream by calling a nextTuple () method by the Spout, and packaging and transmitting the information to the Bolt, wherein the method comprises the following specific steps:
21) reading n data from the data stream of the corresponding monitoring point;
22) judging whether the sequence length formed by the data reaches N (predicted historical data sample capacity), if so, performing step 24, otherwise, turning to step 23;
23) reading new data from the data stream, adding the new data to the end of the sequence, jumping back to step 22;
24) sending the encapsulation component group tuple form to a preprocessing Bolt;
step three, Bolt design
The method comprises the following steps that a Bolt receives a tuple transmitted by the Spout, and executes () methods are called by a plurality of bolts to respectively realize data preprocessing, prediction and early warning, and the method comprises the following specific steps:
31) embedding a python packet of a preprocessing algorithm in an execute () method in a preprocessing Bolt, when the preprocessing Bolt receives N data of a tuple analysis corresponding to a monitoring point transmitted from a Spout, automatically calling the python packet of the preprocessing algorithm to preprocess original data, and sending the preprocessed data to a GRU prediction Bolt in a tuple form;
32) embedding a trained GRU model python packet in an execute () method in a GRU prediction Bolt, automatically calling the trained GRU model when the prediction Bolt receives data transmitted by a preprocessing Bolt, predicting data at the next moment, sending a predicted value to an early warning Bolt, waiting for actual data at the moment, and sending the actual data to the early warning Bolt;
33) the early warning Bolt designs the early warning of the state data of the coal mining machine by using a time sliding window model:
given time t and span d, at t-d, t]The data stream arriving in the time period is a time basic window marked as W, and the jth time basic window is Wj(ii) a A sequence of successive time basic windows constitutes a time sliding window WS，WSi＝Wi-n+j,Wi-n+j+1,…,WiAnd the number of the time sliding windows is the number of the ith basic window after the ith basic window arrives, n represents the number of the basic windows accommodated by one time sliding window, each basic window judges the early warning level of the data in parallel, only the tuples judged by the threshold of each basic window are cached, each tuple of the original window is not required to be cached one by one, the early warning result is stored in a storage bolt database, and early warning is carried out when the predicted value of a certain sensor exceeds the threshold.
Further, GRU model training is carried out, and df is used as input data after original data are preprocessed by missing values, abnormal values and noise; input data is divided into 7: 3, the hidden layer is trained aiming at the training set, the super-parameters of the model are adjusted through an optimization function Adam and a loss function MSE, the model is optimized by taking the loss value loss minimum as an optimization criterion, the time with the best comprehensive performance in the training set is found out, and the loading model is stored to obtain the optimal result of the epoch; GRU parameters conforming to the data characteristics are found through training, a test set simulates real-time data flow on Hadoop Database, data of various test sets are input into various Workers in parallel under the control of a master control node Nimbus, and prediction and early warning of the data are completed in respective topologies.
Further, after the GRU is trained in the training set, the performance of the GRU in the test set is evaluated from three aspects of RMSE, MAE and R2, and the calculation formulas are respectively:
Figure BDA0003033964060000031
Figure BDA0003033964060000041
Figure BDA0003033964060000042
wherein
Figure BDA0003033964060000043
To predict value, yiAre true values.
Further, in the first step, after the Hadoop Database is deployed by the main control node Nimbus, the hypervisor node is responsible for receiving tasks allocated by Nimbus, managing and starting all work workers, and transmitting data to Spout for processing.
In a preferred embodiment of the present application, the shearer operating state data includes a cutting section motor current, a cutting section motor temperature, a traction section motor current, a traction section motor speed, an increase pump operating pressure, an increase pump operating speed, a cooling water pressure, and a frequency converter current in the shearer.
Further, in the step 33), the sliding distance of the sliding window model is 1s, the size of the sliding window is 1min, the sliding window is divided into 60 basic time windows, the attribute set represents a set of state data of the coal mining machine, each basic window judges the early warning level of the data in parallel, only 60 tuples judged by the threshold value of each basic window are cached, each tuple of the original windows does not need to be cached one by one, and the early warning result is stored in the storage bolt database.
In a preferred embodiment of the present application, the threshold value is set based on past experience and error requirements.
The prediction framework principle of the prediction method of the application is as follows: the method comprises the steps that running state data of the coal mining machine are continuously generated at fixed time intervals, Storm provides a platform for online real-time processing of the coal mining machine state detection data, flow data collected in production are simulated through a Hadoop Database distributed storage Database, the flow data are transmitted to different message queues Spout, each Spout sends the data to corresponding bolts in a tuple flow mode, preprocessing, prediction, error calculation and storage of the data are achieved through a plurality of bolts, and finally real-time prediction of the coal mining machine running state data is achieved.
The method breaks through the traditional monitoring method for the running state of the coal mining machine, is integrated with a prediction and early warning method for the real-time running state of the coal mining machine, adopts a GRU (Gate Current Unit) model in deep learning to predict time sequence data of the coal mining machine aiming at a large amount of long-term time sequence data, and utilizes the method to monitor the predicted value of each sensor in real time after the stability of the prediction model is ensured through training, and early warning is carried out when the predicted value of a certain sensor exceeds a threshold value; through the inspection of the accuracy and the processing efficiency of early warning, the accuracy of various data state predictions all reaches more than 85%, the realization duration of the whole early warning process is only about 10s and is far lower than 1min of the interval of measuring point data, and the Storm distributed real-time processing frame has great practical value in the parallel processing of mass coal mine data.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a Storm data flow model base structure;
FIG. 2 is a Storm frame diagram;
FIG. 3 is a schematic view of a sliding window;
FIG. 4 is a graph showing the comparison between the predicted state and the actual state of each data in example 2;
FIG. 5 is a comparison graph of the real values and predicted results of 300 points in each test set in example 2.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications can be made by persons skilled in the art without departing from the spirit of the invention. All falling within the scope of the present invention.
Example 1
Storm, as shown in FIG. 1, is an open source, distributed real-time computing framework that can easily handle unlimited data flows. The core component of Storm mainly comprises a master node Nimbus and a slave node Supervisor. The method mainly comprises the steps that a Nimbus node is mainly responsible for resource allocation and task scheduling, a Supervisor node is responsible for receiving tasks allocated by Nimbus, managing and starting all working Workers, 1 Supervisor corresponds to 4 Workers, 1 Worker corresponds to 1 Topology Topology, and the Topology Topology comprises Stream, Spout and Bolt. Stream, i.e. a data Stream; spout acts as a collector, connected to the data source; bolt is a service logic operation node, subscribes a plurality of spits, and realizes operations such as service processing, connection operation and the like.
As shown in fig. 2, which is a frame diagram of Storm of the present invention, the prediction frame principle of the prediction method is as follows: the method comprises the steps that running state data of the coal mining machine are continuously generated at fixed time intervals, Storm provides a platform for online real-time processing of the coal mining machine state detection data, flow data collected in production are simulated through a Hadoop Database distributed storage Database, the flow data are transmitted to different message queues Spout, each Spout sends the data to corresponding bolts in a tuple flow mode, preprocessing, prediction, error calculation and storage of the data are achieved through a plurality of bolts, and finally real-time prediction of the coal mining machine running state data is achieved.
The method for predicting the running state of the coal mining machine in real time based on Storm comprises the following steps:
step one, designing a data storage structure
The method comprises the steps that operation state data of the coal mining machine belong to time series data, and a data storage structure is designed based on a Hadoop Database distributed storage Database to facilitate management of massive operation data, wherein the data storage structure comprises an index table and a plurality of data tables;
the index table includes a Location of a monitoring point (Location), a GRU Prediction model (Prediction model), and Upper and Lower bounds of a pre-warning Threshold (Threshold Upper and Lower Bound), and is shown in table 1:
table 1 index table
Figure BDA0003033964060000061
The data table simulates a time sequence in a row increasing mode, each row represents sensor time sequence data acquired by a certain monitoring point, and in actual production, real-time operation data of each monitoring point is stored into a Hadoop Database at fixed time intervals, and then the Hadoop Database transfers the data into a Spout for processing through the deployment of a main control node Nimbus; the data table is shown in table 1:
TABLE 2 data sheet
Figure BDA0003033964060000062
Step two, Spout design
Simulating flow data collected in production through a Hadoop Database distributed storage Database, transmitting the flow data to different message queues Spout, extracting information of corresponding monitoring points from the data stream by calling a nextTuple () method by the Spout, and packaging and transmitting the information to the Bolt, wherein the method comprises the following specific steps:
21) reading n data from the data stream of the corresponding monitoring point;
22) judging whether the sequence length formed by the data reaches N (predicted historical data sample capacity), if so, performing step 24, otherwise, turning to step 23;
23) reading new data from the data stream, adding the new data to the end of the sequence, jumping back to step 22;
24) sending the encapsulation component group tuple form to a preprocessing Bolt;
step three, Bolt design
The method comprises the following steps that a Bolt receives a tuple transmitted by the Spout, and executes () methods are called by a plurality of bolts to respectively realize data preprocessing, prediction and early warning, and the method comprises the following specific steps:
31) embedding a python packet of a preprocessing algorithm in an execute () method in a preprocessing Bolt, when the preprocessing Bolt receives N data of a tuple analysis corresponding to a monitoring point transmitted from a Spout, automatically calling the python packet of the preprocessing algorithm to preprocess original data, and sending the preprocessed data to a GRU prediction Bolt in a tuple form;
32) embedding a trained GRU model python packet in an execute () method in a GRU prediction Bolt, automatically calling the trained GRU model when the prediction Bolt receives data transmitted by a preprocessing Bolt, predicting data at the next moment, sending a predicted value to an early warning Bolt, waiting for actual data at the moment, and sending the actual data to the early warning Bolt;
33) the early warning Bolt designs the early warning of the state data of the coal mining machine by using a time sliding window model:
given time t and span d, at t-d, t]The data stream arriving in the time period is a time basic window marked as W, and the jth time basic window is Wj(ii) a A sequence of successive time basic windows constitutes a time sliding window WS，WSi＝Wi-n+j,Wi-n+j+1,…,WiIs the time sliding window after the arrival of the ith basic window, wherein n represents the number of basic windows accommodated by one time sliding window, and the time sliding window is shown in fig. 3.
The sliding distance of the sliding window model is 1s, the size of the sliding window is 1min, the sliding window is divided into 60 basic time windows, the attribute set represents a set of state data of the coal mining machine, all the basic windows judge the early warning level of the data in parallel, only 60 tuples judged by all basic window thresholds are cached, all tuples of original windows do not need to be cached one by one, and early warning results are stored in a storage bolt database.
Example 2
In the embodiment, 3 PCs with the same configuration are selected to build the Storm distributed cluster environment, and each machine is deployed with one virtual machine. The three virtual machine operating systems are CentOS6.8, one virtual machine operating system serves as a Master and is provided with a Nimbus node, the other two virtual machines are provided with Supervisor nodes, and after receiving tasks of a Storm cluster, the Nimbus realizes resource allocation to the Supervisor through Zookeeper, and a main node Nimbus dual-core single processor, a 4GB memory and a 40G hard disk; a secondary node single-core single processor, a 2GB memory and a 20G hard disk.
Taking data of a MG400930-WD electric traction coal mining machine on a certain mine fully-mechanized mining face as an example, 1000 pieces of monitoring data of cutting part motor current, cutting part motor temperature, traction part motor current, traction part motor rotating speed, heightened pump working pressure, heightened pump working rotating speed, cooling water pressure and frequency converter current in the coal mining machine are taken as experimental data.
GRU model training is performed first. The original data is preprocessed by missing values, abnormal values and noise, and then df is used as input data. Input data is divided into 7: and 3, the hidden layer is trained aiming at the training set, the super-parameters of the model are adjusted through an optimization function Adam and a loss function MSE, the model is optimized by taking the loss value loss minimum as an optimization criterion, the time with the best comprehensive performance in the training set is found out, and the loading model is stored to obtain the optimal result of the epoch.
GRU parameters conforming to the data characteristics are found through training, the test set simulates real-time data flow on Hadoop Database, a basic time window is set to be 1min, under the regulation and control of a master control node Nimbus, the data of the eight test sets are input into 8 workers in parallel, and prediction and early warning of the data are completed in respective topologies.
Prediction results of the prediction model:
cutting part motor current, cutting part motor temperature, traction part motor current, traction part motor speed, pump working pressure is increased, pump working speed is increased, cooling water pressure and frequency converter current eight kinds of monitoring data are respectively represented by 1-8, optimizing training times epoch, learning rate, neuron number high size, weight attenuation weight decay, time step time, training sample number N each time, hidden layer number layer, fitting goodness R2With C1～C8The result of the hyper-parameter optimization of the GRU model in the training set of the experiment is shown in Table 3.
TABLE 3 GRU Superparametric optimization results
Figure BDA0003033964060000081
And (3) importing the trained GRU into Bolt, simulating real-time data flow by the test set, and comparing the real value and the prediction result of 300 points in each test set, such as the graph of FIG. 5.
This embodiment uses Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and goodness of fit (R)2) As an evaluation index.The calculation formulas are respectively as follows:
Figure BDA0003033964060000091
Figure BDA0003033964060000092
Figure BDA0003033964060000093
wherein
Figure BDA0003033964060000095
To predict value, yiFor the true values, the test aggregation results are shown in table 4 below:
TABLE 4 evaluation index comparison Table
Figure BDA0003033964060000094
R2 characterizes the goodness of fit, the closer to 1, the better the effect. The mean absolute error and root mean square error of MAE and RMSE characterization, the closer to 0, the better the effect. As can be seen from the data in the table 4, the goodness of fit of the predicted value and the actual value represented by R2 reaches more than 90%, the magnitude of MAE and RMSE is small compared with the magnitude of experimental data, and the model can be suitable for predicting the running state data of the coal mining machine.
The accuracy of early warning:
in this example, threshold settings were made for each data of the experiment based on past experience and error requirements, and the threshold settings for each data are shown in table 5 below.
TABLE 5 data threshold settings
Figure BDA0003033964060000101
According to the threshold setting, the states of each data can be divided into three categories: normal, attention and fault. When the coal mining machine runs, the data does not reach the attention value, the state is normal running, the working performance is stable, and no measures need to be taken; the data reaches the attention value but does not reach the failure threshold value, and the state is attention monitoring; and when the data reaches the fault threshold value, the state is that the fault is to be repaired. And predicting the data state and making corresponding early warning before the prediction Bolt obtains the prediction data and the actual data does not arrive. The comparison result of the predicted state and the actual state of each data of the test set is shown in FIG. 4; normal, attention and failure are respectively represented by 1, 2 and 3, and the accuracy of each data prediction is shown in Table 6:
TABLE 6 early warning accuracy
Figure BDA0003033964060000102
Except for the cooling water pressure, the state early warning accuracy rate of other data all reaches more than 95%, and the practical requirements are met. The cooling water pressure has small data, small interval between the attention value and the fault threshold value, large error during prediction, higher accuracy and certain practical value, and the error is more than 85 percent.
Storm platform based processing efficiency:
in the embodiment, data obtained from a sensor is simulated and transmitted into a Database Hadoop Database, Storm reads data from Hadoop, a main control node Nimbus monitors and distributes tasks through zookeeper, and a specific processing logic is handed to Worker through a Supervisor node to complete. The embodiment respectively measures the time required by the database to transmit the stream data into the Spout; the time required for the Spout to process data for distribution to Bolt; preprocessing Bolt, predicting Bolt, pre-warning Bolt, and total time required to store Bolt. The treatment times for each fraction are shown in table 7 below:
TABLE 7 processing schedules
Figure BDA0003033964060000111
The processing speed of the Spout and each Bolt in the database and the Worker for the stream data is very high, the realization duration of the whole early warning process is only about 10s, and the requirement of the measuring point data can be met for 1min interval.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes and modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention.

Claims (6)

1. A method for predicting the running state of a coal mining machine in real time based on Storm is characterized by comprising the following steps:
step one, designing a data storage structure
The running state data of the coal mining machine belongs to time sequence data, and a data storage structure is designed based on a Hadoop Database distributed storage Database and comprises an index table and a data table;
the index table comprises monitoring point positions, a GRU prediction model and upper and lower early warning threshold boundaries;
the data table simulates a time sequence in a row increasing mode, each row represents sensor time sequence data acquired by a certain monitoring point, and in actual production, real-time operation data of each monitoring point is stored into a Hadoop Database at fixed time intervals, and then the Hadoop Database transfers the data into a Spout for processing through the deployment of a main control node Nimbus;
step two, Spout design
Simulating flow data collected in production through a Hadoop Database distributed storage Database, transmitting the flow data to different message queues Spout, extracting information of corresponding monitoring points from the data stream by calling a nextTuple () method by the Spout, and packaging and transmitting the information to the Bolt, wherein the method comprises the following specific steps:
21) reading n data from the data stream of the corresponding monitoring point;
22) judging whether the sequence length formed by the data reaches N, if so, performing a step 24, otherwise, turning to a step 23;
23) reading new data from the data stream, adding the new data to the end of the sequence, jumping back to step 22;
24) sending the encapsulation component group tuple form to a preprocessing Bolt;
step three, Bolt design
The method comprises the following steps that a Bolt receives a tuple transmitted by the Spout, and executes () methods are called by a plurality of bolts to respectively realize data preprocessing, prediction and early warning, and the method comprises the following specific steps:
31) embedding a python packet of a preprocessing algorithm in an execute () method in a preprocessing Bolt, when the preprocessing Bolt receives N data of a tuple analysis corresponding to a monitoring point transmitted from a Spout, automatically calling the python packet of the preprocessing algorithm to preprocess original data, and sending the preprocessed data to a GRU prediction Bolt in a tuple form;
32) embedding a trained GRU model python packet in an execute () method in a GRU prediction Bolt, automatically calling the trained GRU model when the prediction Bolt receives data transmitted by a preprocessing Bolt, predicting data at the next moment, sending a predicted value to an early warning Bolt, waiting for actual data at the moment, and sending the actual data to the early warning Bolt;
33) the early warning Bolt designs the early warning of the state data of the coal mining machine by using a time sliding window model:
given time t and span d, at t-d, t]The data stream arriving in the time period is a time basic window marked as W, and the jth time basic window is Wj(ii) a A sequence of successive time basic windows constitutes a time sliding window WS，WSi＝Wi-n+j,Wi-n+j+1,…,WiAnd the number of the time sliding windows is the number of the ith basic window after the ith basic window arrives, n represents the number of the basic windows accommodated by one time sliding window, each basic window judges the early warning level of the data in parallel, only the tuples judged by the threshold of each basic window are cached, each tuple of the original window is not required to be cached one by one, the early warning result is stored in a storage bolt database, and early warning is carried out when the predicted value of a certain sensor exceeds the threshold.
2. The Storm-based real-time prediction method for the operating state of the coal mining machine according to claim 1, characterized in that the method further comprises the steps of carrying out GRU model training, preprocessing raw data by missing values, abnormal values and noise, and taking df as input data; input data is divided into 7: 3, the hidden layer is trained aiming at the training set, the super-parameters of the model are adjusted through an optimization function Adam and a loss function MSE, the model is optimized by taking the loss value loss minimum as an optimization criterion, the time with the best comprehensive performance in the training set is found out, and the loading model is stored to obtain the optimal result of the epoch; GRU parameters conforming to the data characteristics are found through training, a test set simulates real-time data flow on Hadoop Database, data of various test sets are input into various Workers in parallel under the control of a master control node Nimbus, and prediction and early warning of the data are completed in respective topologies.
3. The Storm-based real-time prediction method for the operating state of the coal mining machine according to claim 1, wherein after the GRU is trained in the training set, the performance of the GRU in the testing set is evaluated from three aspects of RMSE, MAE and R2, and the calculation formulas are as follows:
Figure FDA0003033964050000021
Figure FDA0003033964050000022
Figure FDA0003033964050000023
wherein
Figure FDA0003033964050000024
To predict value, yiAre true values.
4. The Storm-based real-time prediction method for the operation state of the coal mining machine according to claim 1, characterized in that the real-time operation data of each monitoring point is allocated by a master control node Nimbus by a Hadoop Database, and then the Supervisor node is responsible for receiving tasks distributed by Nimbus, managing and starting all working workers, and transmitting the data into Spout for processing.
5. The Storm-based real-time prediction method of shearer operating conditions according to claim 1, wherein the shearer operating condition data comprises cutting unit motor current, cutting unit motor temperature, haulage unit motor current, haulage unit motor speed, turn-up pump operating pressure, turn-up pump operating speed, cooling water pressure and frequency converter current in the shearer.
6. The Storm-based real-time prediction method for the operation state of the coal mining machine according to claim 1, wherein in the step 33), the sliding distance of the sliding window model is 1s, the size of the sliding window is 1min, the sliding window is divided into 60 basic time windows, the attribute set represents a set of state data of the coal mining machine, the basic windows judge the early warning level of the data in parallel, only 60 tuples after threshold judgment of the basic windows are cached, one-by-one caching of each tuple of the original windows is not needed, and the early warning result is stored in a storage bolt database.
CN202110438420.8A 2021-04-22 2021-04-22 Real-time prediction method for running state of coal mining machine based on Storm Active CN113094247B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110438420.8A CN113094247B (en) 2021-04-22 2021-04-22 Real-time prediction method for running state of coal mining machine based on Storm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110438420.8A CN113094247B (en) 2021-04-22 2021-04-22 Real-time prediction method for running state of coal mining machine based on Storm

Publications (2)

Publication Number Publication Date
CN113094247A true CN113094247A (en) 2021-07-09
CN113094247B CN113094247B (en) 2023-08-15

Family

ID=76679779

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110438420.8A Active CN113094247B (en) 2021-04-22 2021-04-22 Real-time prediction method for running state of coal mining machine based on Storm

Country Status (1)

Country Link
CN (1) CN113094247B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017092180A1 (en) * 2015-12-01 2017-06-08 中国矿业大学 Combined inertial navigation and laser scanning coal shearer positioning device and method
CN110400448A (en) * 2019-07-29 2019-11-01 精英数智科技股份有限公司 A kind of coalcutter operating status judgment method, system and storage medium
CN111768022A (en) * 2020-05-09 2020-10-13 中国煤矿机械装备有限责任公司 Equipment detection method and device for coal machine production equipment

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017092180A1 (en) * 2015-12-01 2017-06-08 中国矿业大学 Combined inertial navigation and laser scanning coal shearer positioning device and method
CN110400448A (en) * 2019-07-29 2019-11-01 精英数智科技股份有限公司 A kind of coalcutter operating status judgment method, system and storage medium
CN111768022A (en) * 2020-05-09 2020-10-13 中国煤矿机械装备有限责任公司 Equipment detection method and device for coal machine production equipment

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
BRODNY, J等: "Analysis of Availability of Longwall-Shearer Based On Its Working Cycle", 《WORLD MULTIDISCIPLINARY EARTH SCIENCES SYMPOSIUM》 *
曹现刚;姜韦光;张国祯;: "采煤机运行状态数据实时清洗技术研究", 煤炭工程, no. 03 *

Also Published As

Publication number Publication date
CN113094247B (en) 2023-08-15

Similar Documents

Publication Publication Date Title
CN105354614B (en) A kind of electric network information O&M active forewarning method based on big data
CN107330056B (en) Wind power plant SCADA system based on big data cloud computing platform and operation method thereof
CN102882969B (en) A kind of safety production cloud service platform of industrial and mining enterprises
CN102932419B (en) A kind of data-storage system for the safety production cloud service platform towards industrial and mining enterprises
CN106952028A (en) Dynamoelectric equipment failure is examined and health control method and system in advance
CN102917032B (en) A kind of safety production cloud service platform of industrial and mining enterprises
CN106447107B (en) Maintenance method based on aircraft structure health monitoring
CN102903011A (en) Mass data processing system used for safety production cloud service platform facing industrial and mining enterprises
CN102880802A (en) Fatal danger fountainhead analysis and evaluation method for safety production cloud service platform system facing industrial and mining enterprises
CN102903010A (en) Support vector machine-based abnormal judgment method for safety production cloud service platform orientating industrial and mining enterprises
CN110865924B (en) Health degree diagnosis method and health diagnosis framework for internal server of power information system
CN102929827A (en) Wireless sensor data acquisition cluster for industrial-and-mining-enterprise-oriented safety production cloud service platform
CN102930372A (en) Data analysis method for association rule of cloud service platform system orienting to safe production of industrial and mining enterprises
CN104407688A (en) Virtualized cloud platform energy consumption measurement method and system based on tree regression
CN109491339B (en) Big data-based substation equipment running state early warning system
CN104361088A (en) Congestion data processing method based on real-time weight analysis in SCADA (supervisory control and data acquisition) system
CN106570567A (en) Main network maintenance multi-constraint multi-target evaluation expert system and optimization method
CN105967063A (en) Failure analyzing and handling system and method of maintenance platform
CN110930072A (en) Reliability research method of traction power supply system
CN116308304A (en) New energy intelligent operation and maintenance method and system based on meta learning concept drift detection
CN107103425B (en) Intelligent energy evaluation system for power generation equipment running state computer
CN110941558B (en) Intelligent office remote operation and maintenance method and system
Ju et al. The use of edge computing-based internet of things big data in the design of power intelligent management and control platform
CN102903009A (en) Malfunction diagnosis method based on generalized rule reasoning and used for safety production cloud service platform facing industrial and mining enterprises
CN102915482A (en) Safety production process control and management method for cloud service platforms of industrial and mining enterprises

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant