CN117990162B - Inspection well monitoring device and method based on convolutional neural network - Google Patents
Inspection well monitoring device and method based on convolutional neural network Download PDFInfo
- Publication number
- CN117990162B CN117990162B CN202410400652.8A CN202410400652A CN117990162B CN 117990162 B CN117990162 B CN 117990162B CN 202410400652 A CN202410400652 A CN 202410400652A CN 117990162 B CN117990162 B CN 117990162B
- Authority
- CN
- China
- Prior art keywords
- data
- monitoring
- inspection well
- model
- sound
- 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.)
- Active
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 178
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012806 monitoring device Methods 0.000 title claims abstract description 24
- 238000012544 monitoring process Methods 0.000 claims abstract description 165
- 238000012545 processing Methods 0.000 claims abstract description 101
- 230000006870 function Effects 0.000 claims abstract description 34
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 24
- 239000007788 liquid Substances 0.000 claims abstract description 21
- 238000004458 analytical method Methods 0.000 claims abstract description 19
- 230000002159 abnormal effect Effects 0.000 claims description 75
- 238000005457 optimization Methods 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 24
- 238000013461 design Methods 0.000 claims description 20
- 239000011159 matrix material Substances 0.000 claims description 18
- 238000004891 communication Methods 0.000 claims description 17
- 230000008447 perception Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 14
- 230000006978 adaptation Effects 0.000 claims description 13
- 230000005540 biological transmission Effects 0.000 claims description 13
- 238000005516 engineering process Methods 0.000 claims description 13
- 238000007405 data analysis Methods 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 10
- 238000013480 data collection Methods 0.000 claims description 8
- 230000003321 amplification Effects 0.000 claims description 7
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000012098 association analyses Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 239000000463 material Substances 0.000 claims description 5
- 229910001335 Galvanized steel Inorganic materials 0.000 claims description 4
- 239000000084 colloidal system Substances 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 238000004146 energy storage Methods 0.000 claims description 4
- 239000008397 galvanized steel Substances 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 4
- 239000004743 Polypropylene Substances 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 3
- -1 polypropylene Polymers 0.000 claims description 3
- 229920001155 polypropylene Polymers 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 abstract description 28
- 230000008859 change Effects 0.000 abstract description 10
- 238000010586 diagram Methods 0.000 description 4
- 230000009977 dual effect Effects 0.000 description 4
- 238000012423 maintenance Methods 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 238000009434 installation Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 239000010865 sewage Substances 0.000 description 2
- 238000009423 ventilation Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 238000000746 purification Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 238000004148 unit process Methods 0.000 description 1
Landscapes
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention discloses an inspection well monitoring device and method based on a convolutional neural network, and relates to the technical field of inspection well monitoring. According to the invention, the data sensing unit is arranged, so that various sensors can be integrated into a whole in a multifunctional and multi-mode manner, the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time are realized, the water flow condition of the pipeline in the inspection well is accurately monitored, a decision basis is provided for subsequent processing and analysis, and the intelligent scheduling of a drainage system is realized.
Description
Technical Field
The invention relates to the technical field of inspection well monitoring, in particular to an inspection well monitoring device and method based on a convolutional neural network.
Background
With the acceleration of the urban infrastructure, construction and maintenance of urban infrastructure has become an important task, wherein inspection shafts are an important component of urban infrastructure, and the security and functionality of the inspection shafts are critical to the normal operation of cities, and the inspection shafts generally refer to underground water storage facilities or shafts for ventilation and descent into underground spaces (such as basements, tunnels and the like). In modern urban infrastructure, inspection wells may refer to underground pipelines, ventilation shafts, drainage wells, etc., which are very important for the arrangement and maintenance of urban drainage systems, communication lines.
In the field of safety monitoring and management, the safety of the inspection well is always an important point of urban management, however, the traditional inspection well monitoring mode has a plurality of problems, such as incomplete monitoring, difficult maintenance and the like, so that research on a monitoring scheme in inspection well monitoring is needed.
Meanwhile, the traditional inspection well has low mechanization degree, the pipeline is silted and blocked, damaged and leaked, and the inspection and the judgment are completely carried out by human, so that manpower and material resources are wasted, real-time monitoring cannot be achieved, a real-time monitoring system is imperfect, equipment perception is limited, daily maintenance and operation of the equipment are difficult, harmful gas in the pipeline cannot be timely monitored, and physical health of an maintainer can be possibly endangered.
For the problems in the related art, no effective solution has been proposed at present.
Disclosure of Invention
(One) solving the technical problems
Aiming at the defects of the prior art, the invention provides an inspection well monitoring device and an inspection well monitoring method based on a convolutional neural network.
(II) technical scheme
The invention adopts the following specific technical scheme:
according to one aspect of the invention, an inspection well monitoring device based on a convolutional neural network is provided, and the inspection well monitoring device comprises an inspection well, wherein an upright post is arranged at the top end of the inspection well, an inspection well cover is arranged on one side of the upright post, a safety monitoring unit and a data sensing unit are respectively arranged at the top end and the bottom end of the inspection well cover, the data sensing unit is arranged on the side wall of the inspection well, a data processing unit is arranged in the upright post, a solar power supply unit is arranged at the top end of the upright post, and a communication unit is arranged at the top end of the data processing unit and penetrates through the upright post.
Furthermore, in order to integrate various sensors into a whole in a multifunctional and multi-mode manner, realize the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time, accurately monitor the water flow condition of the pipeline in the inspection well, provide decision basis for subsequent processing analysis, realize intelligent scheduling of a drainage system, and realize the intelligent scheduling of the drainage system, wherein the data sensing unit comprises a sensor shell arranged on the inner side wall of the inspection well, a turbidity sensor is arranged at the inner bottom end of the sensor shell, a pH value sensor is arranged at the top end of the turbidity sensor, and the bottom end of the turbidity sensor is connected with a water quality sensor; the data sensing unit further comprises a bell jar type shell arranged at the top end of the sensor shell, a gas sensor is arranged in the bell jar type shell, a liquid level sensor is arranged on one side of the gas sensor, an image sensor is arranged on one side of the liquid level sensor, and the image sensor is arranged on the outer side of the bell jar type shell.
Further, in order to avoid the outage condition appearing in the in-process of carrying out the monitoring, solar power supply unit is provided with a plurality of solar cell panels including installing in the photovoltaic controller on stand top in the outside of photovoltaic controller, and one side of stand is provided with the energy storage colloid battery that photovoltaic controller connects.
Further, in order to ensure that the sensor housing and the sensor inside the bell jar type housing are integrally provided with stronger corrosion resistance, so that pollution is prevented better, the measurement accuracy is effectively ensured, the stand column is made of galvanized steel pipes, the inside of the stand column is of a hollow structure, and the sensor housing and the bell jar type housing are made of polypropylene waterproof materials.
Further, the data processing unit is used for carrying out processing and analysis operation on the data acquired by the data sensing unit, and generating a decision instruction according to a preset algorithm to predict the state in the inspection well;
the data processing unit comprises a data collection marking module, a data dividing and processing module, a processing model building module, a deployment monitoring implementation module and an alarm decision processing module;
The data collection marking module is used for collecting parameter data in the inspection well, which is collected by the data sensing unit, and performing time marking operation on the parameter data according to a time point, wherein the parameter data comprises water quality data, gas concentration data, water flow data and image data;
The data dividing and processing module is used for dividing the marked parameter data into a training set, a verification set and a test set according to a preset proportion;
The processing model building module is used for building a processing model by utilizing parameter data and a convolutional neural network technology to realize prediction of the abnormal condition inside the inspection well;
The deployment monitoring implementation module is used for inputting the processing model into a designated inspection well monitoring area to predict the condition in the inspection well;
And the alarm decision processing module is used for implementing comparison operation by using the prediction result and the threshold value, judging whether an abnormal condition exists in the inspection well according to the comparison result and adopting corresponding measures.
Further, the method for predicting the internal abnormal condition of the inspection well by using the parameter data and the convolutional neural network technology to establish a processing model comprises the following steps:
selecting vectors from the test set by utilizing a sliding search and time point marking mode to construct an input matrix with time characteristics;
combining an input matrix with a convolutional neural network to establish a processing model, and optimizing model parameters by adopting an optimization algorithm and a training set;
Simulating the optimized processing model by using the verification set to output a prediction result of the inspection well;
the method for constructing the input matrix with time characteristics by selecting vectors from the test set by utilizing a sliding search and time point marking mode comprises the following steps:
Selecting a characteristic vector of an abnormal condition monitored by the inspection well and an output abnormal category of the processing model according to the test set, and performing abnormal value removal and normalization processing on the characteristic vector;
Setting a sliding search length value, sequentially carrying out feature vectors according to a time-stamped result, and carrying out sliding operation according to the sliding length value;
Acquiring time sequence characteristics in the time mark according to the operation result, and marking the time mark on the sliding window to acquire an overlapped subset sequence;
Traversing the subset sequences of all the feature vectors, combining the subset sequences into a sample set, and taking the sample set and the corresponding abnormal category as an input matrix.
Further, combining the input matrix with the convolutional neural network to build a processing model, and optimizing the model parameters by adopting an optimization algorithm and a training set comprises:
Taking the input matrix as an input variable, taking a corresponding abnormal class as an output variable, and capturing a nonlinear relation among a plurality of input variables by utilizing a convolution layer in a convolution neural network;
Combining nonlinear relations of input variables by using a full-connection layer, and selecting the maximum value by using a pooling layer to obtain an abnormal class output model;
Judging a loss function of the output model, updating model parameters by using an optimization algorithm to obtain a processing model, and establishing a model evaluation standard by using a training set to judge a model parameter optimization result;
And when the optimization result is smaller than the standard value, the optimization is failed, and parameter optimization is continued until the optimization result is larger than the standard value.
Further, the safety monitoring unit is used for utilizing the internet of things technology to design a remote automatic monitoring model to monitor the inspection well cover, analyzing whether the inspection well cover has abnormal conditions, and judging the stability of the data sensing unit;
The safety monitoring unit comprises a monitoring data acquisition module, a data transmission and receiving module, a monitoring frame design module, an abnormal condition judgment module, an alarm design generation module and a monitoring integrated connection module;
The monitoring data acquisition module is used for installing a sound sensor on the side wall of the top end of the inspection well cover to acquire sound data of the inspection well cover within a preset distance;
The data transmission and receiving module is used for establishing a perception framework by utilizing a decibel amplification mode and importing a sound template under the condition of normal monitoring of the system design of the internet of things, wherein the sound template comprises knocking noise sound and carrying moving sound within a preset distance of the inspection well cover;
The monitoring framework design module is used for adding an automatic updating monitoring weight function in the sensing framework according to the sound template, and establishing an automatic monitoring model by utilizing the sound template and the sound data analysis fitness function;
The abnormal condition judging module is used for inputting preset sound data in the automatic monitoring model and judging the state of the sound template of the inspection well cover under the abnormal condition according to the output result;
the alarm design generation module is used for analyzing the stability of the data sensing unit according to the abnormal condition of the inspection well cover and setting a corresponding alarm reminding mode by utilizing the stability result;
and the monitoring integrated connection module is used for implementing integrated connection operation of the automatic monitoring system and the sound sensor, analyzing whether the inspection well cover has abnormal conditions according to real-time sound data, and judging the stability of the data sensing unit according to an analysis result.
Further, adding an automatic update monitoring weight function in the perception architecture according to the sound template, and establishing an automatic monitoring model by utilizing the sound template and the sound data analysis fitness function comprises the following steps:
Performing conversion operation on the data in the designed sound template in a spectrum centroid mode to obtain a feature vector, and utilizing a preset monitoring requirement to define a fitness function to evaluate the matching degree between the feature vector and the sound template;
transmitting the matching degree to a sensing framework to construct an automatic monitoring model, and judging the fitness score of preset sound data by combining an association analysis algorithm with the automatic monitoring model;
And formulating a weight updating rule according to the fitness score result, and adding an automatic updating monitoring weight function into the perception framework based on the updating rule.
Further, transmitting the matching degree to the sensing architecture to construct an automatic monitoring model, and determining the fitness score of the preset sound data by combining the automatic monitoring model with the association analysis algorithm comprises:
packaging the matching degree into a preset transmission format, transmitting the preset transmission format into a sensing framework according to a preset communication protocol, and receiving data from each matching degree by the sensing framework;
presetting a monitoring frame based on a perception framework, establishing a risk factor set as a sample by combining the matching degree, and setting the number of the samples and the risk factor set;
Constructing an expression frame according to the risk factor set to describe a sample state when the expression frame belongs to the risk state, and obtaining the rough membership of the sample to judge the occupation proportion of the risk factors;
Judging the adaptation degree between each group of risk factors according to the occupation proportion of each group of risk factors and a preset ideal value, constructing an identification frame by using the adaptation degree, and combining a monitoring frame, an expression frame and the identification frame to construct an automatic monitoring model;
and inputting preset sound data into an automatic monitoring model to obtain risk occupation ratios and adaptation values, and sensing the adaptation degree scores of the sound data under different risk occupation ratios by using a correlation analysis algorithm.
According to another aspect of the present invention, there is also provided a method for monitoring an inspection well based on a convolutional neural network, the method comprising the steps of:
s1, a data sensing unit is arranged in the inspection well to collect parameter data in the inspection well and transmit the parameter data to a data processing unit;
s2, the data processing unit processes and analyzes the acquired parameter data, and a processing model is utilized to monitor whether an abnormal condition exists in the inspection well;
S3, utilizing a safety monitoring unit to remotely and automatically monitor the condition of the inspection well cover within a preset distance, and judging whether the stability of the data sensing unit is affected by the abnormal condition;
and S4, uploading the monitoring results of the data processing unit and the safety monitoring unit to a cloud platform by the communication unit, and publishing the monitoring results to the mobile terminal and the computer terminal by the cloud platform.
(III) beneficial effects
Compared with the prior art, the inspection well monitoring device and method based on the convolutional neural network provided by the invention have the following beneficial effects:
(1) According to the invention, the data sensing unit is arranged, so that various sensors can be integrated into a whole in a multifunctional and multi-mode manner, the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time are realized, the water flow condition of the pipeline in the inspection well is accurately monitored, a decision basis is provided for subsequent processing and analysis, and the intelligent scheduling of a drainage system is realized.
(2) According to the invention, the data processing unit is arranged to process and analyze the data acquired by the data sensing unit, a decision instruction is generated according to a preset algorithm to predict the state of the inspection well, sensing equipment is arranged in the inspection well, the underground liquid level change can be sensed by using the underground liquid level meter, a liquid level analysis model is established, the flow of the pipe network is monitored in real time by using the flowmeter, meanwhile, the methane concentration in the pipe network environment is detected by using the methane monitor, the methane gas risk is prevented, and the like, the data monitored in real time are subjected to predictive analysis by using a big data analysis technology, the water flow condition in the pipe is accurately monitored, and the scientificity and the operation management level of the reconstruction decision of the drainage system are improved.
(3) According to the invention, the safety monitoring unit is arranged to monitor the sound state near the inspection well in real time by installing the sound real-time monitoring model at the top end of the inspection well cover, and judge whether the abnormal condition exists near the inspection well according to the sound state, so that the early identification of potential danger or safety problem is realized, the influence of artificial theft damage on various sensors in the inspection well is avoided, and the accuracy of the inspection in the inspection well is reduced.
(4) The solar power supply unit is arranged to provide a stable power supply for the whole monitoring device, and the power supply is supplemented when the electric quantity is exhausted, so that the condition that the power supply is powered off in the monitoring process is avoided.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic structural diagram of an inspection well monitoring device based on a convolutional neural network according to an embodiment of the present invention;
FIG. 2 is an enlarged view of a portion of FIG. 1 at A;
FIG. 3 is a partial enlarged view at B in FIG. 1;
FIG. 4 is an enlarged view of a portion of FIG. 1 at C;
FIG. 5 is a schematic block diagram of a data processing unit in a convolutional neural network-based inspection well monitoring device in accordance with an embodiment of the present invention;
FIG. 6 is a schematic block diagram of a safety monitoring unit in a convolutional neural network-based inspection well monitoring device in accordance with an embodiment of the present invention;
FIG. 7 is a control schematic diagram of an inspection well monitoring device based on a convolutional neural network according to an embodiment of the present invention;
FIG. 8 is a flow chart of a data processing technique in a manhole monitoring apparatus based on a convolutional neural network according to an embodiment of the present invention;
fig. 9 is a flowchart of a method for inspection well monitoring based on convolutional neural network in accordance with an embodiment of the present invention.
In the figure:
1. Inspection well; 2. a column; 3. an inspection well cover; 4. a data sensing unit; 401. a sensor housing; 402. a turbidity sensor; 403. a pH value sensor; 404. a water quality sensor; 405. a bell-type housing; 406. a gas sensor; 407. a liquid level sensor; 408. an image sensor; 5. a data processing unit; 501. a data collection marking module; 502. a data dividing and processing module; 503. a processing model building module; 504. deploying a monitoring implementation module; 505. an alarm decision processing module; 6. a solar power supply unit; 601. a photovoltaic controller; 602. a solar cell panel; 603. an energy storage colloid storage battery; 7. a communication unit; 8. a security monitoring unit; 801. a monitoring data acquisition module; 802. a data transmission and reception module; 803. a monitoring frame design module; 804. an abnormal condition judgment module; 805. an alarm design generation module; 806. monitoring the integrated connection module; 9. a mains supply unit; 10. a cloud platform; 11. a computer terminal; 12. and a mobile terminal.
Detailed Description
For the purpose of further illustrating the various embodiments, the present invention provides the accompanying drawings, which are a part of the disclosure of the present invention, and which are mainly used to illustrate the embodiments and, together with the description, serve to explain the principles of the embodiments, and with reference to these descriptions, one skilled in the art will recognize other possible implementations and advantages of the present invention, wherein elements are not drawn to scale, and like reference numerals are generally used to designate like elements.
According to the embodiment of the invention, an inspection well monitoring device and an inspection well monitoring method based on a convolutional neural network are provided.
The invention will now be further described with reference to the accompanying drawings and the specific embodiments, as shown in fig. 1 to 8, an inspection well monitoring device based on a convolutional neural network according to an embodiment of the invention includes an inspection well 1, a stand column 2 is provided at the top end of the inspection well 1, an inspection well cover 3 is provided at one side of the stand column 2, a security monitoring unit 8 and a data sensing unit 4 are respectively provided at the top end and the bottom end of the inspection well cover 3, the data sensing unit 4 is mounted on the side wall of the inspection well 1, a data processing unit 5 is provided inside the stand column 2, a solar power supply unit 6 is provided at the top end of the stand column 2, and a communication unit 7 is provided at the top end of the data processing unit 5 and penetrates through the stand column 2.
As shown in fig. 7, it should be explained that the present invention further includes a mains supply unit 9 when in use, the mains supply unit 9 includes a power conversion module for supplying power to the monitoring system, the communication unit 7 is an NB-loT wireless data uploading device, and has a function of storing data in a cloud computing data center, and transmitting data to the computer terminal 11 and the mobile terminal 12.
The stand column 2 is used for installing the solar power supply unit 6 and the data processing unit 5, galvanized steel pipe materials are empty in the middle, meanwhile, a cable penetrates through the stand column 2 to reach the data sensing unit 4 and the data processing unit 5, and the cable is used for connecting the data sensing unit 4, the solar power supply unit 6, the mains supply unit 9 and the communication unit 7, supplying power for all equipment and transmitting monitoring signals to the intelligent measuring equipment.
The dual-power-supply working principle is that when weather is good, the solar power supply unit 6 is used for generating power, daily operation of each data sensing unit 4 and the intelligent measurement sensor in the inspection well 1 is maintained, the data sensing unit 4 can collect and collect data change parameters such as flow, liquid level and methane concentration in the inspection well 1, and the data change parameters and images are uploaded to the data processing unit 5, and the intelligent measurement sensor can switch the commercial power supply unit 9 to supply power for the whole monitoring equipment in overcast and rainy weather, so that the dual-power-supply scheme of the whole device is realized.
The solar power supply unit 6 and the commercial power supply unit 9 supply power to the monitoring system, real-time conversion is carried out under different conditions, the solar power supply unit 6 can be used for remote areas far away from urban areas, and the whole coverage of the whole underground pipe network detection can be realized.
In one embodiment, for the data sensing unit 4, the data sensing unit 4 includes a sensor housing 401 mounted on an inner sidewall of the inspection well 1, a turbidity sensor 402 is disposed at an inner bottom end of the sensor housing 401, a pH sensor 403 is disposed at a top end of the turbidity sensor 402, and a water quality sensor 404 is connected to a bottom end of the turbidity sensor 402; the data sensing unit 4 further comprises a bell jar type shell 405 arranged at the top end of the sensor shell 401, a gas sensor 406 is arranged in the bell jar type shell 405, a liquid level sensor 407 is arranged on one side of the gas sensor 406, an image sensor 408 is arranged on one side of the liquid level sensor 407, and the image sensor 408 is arranged on the outer side of the bell jar type shell 405, so that various sensors can be integrated into a whole in a multifunctional mode, the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time are realized, accurate monitoring is performed on the water flow condition of the internal pipeline of the inspection well 1, a decision basis is provided for subsequent processing analysis, and intelligent scheduling of a drainage system is realized.
Specifically, the data sensing unit 4 is installed on the side surface of the inspection well 1, the image sensor 408, the liquid level sensor 407 and the gas sensor 406 are located at the top of the inspection well 1, and the image sensor 408 is installed on the outer side of the bell-type shell 405 and is used for collecting images of the inner state of the inspection well 1 and transmitting the collected images to the data processing unit 5, and the liquid level sensor 407 adopts an ultrasonic ranging technology to accurately measure the liquid level height in the inspection well 1, and when the liquid level height exceeds a preset value, the device can automatically start an alarm function.
The gas sensor 406 adopts a methane detector to detect the methane concentration in the pipe network environment, prevents the excessive explosion of methane gas, and when the gas concentration exceeds a preset value, the device can automatically start an alarm function, the turbidity sensor 402 is used for detecting the pH value and the turbidity of sewage, the water quality sensor 404 floats in a pipeline near the inspection well 1 and is used for collecting sewage when the water quantity is smaller, and the data is better provided for the pH value sensor 403 and the turbidity sensor 402.
In one embodiment, for the solar power supply unit 6, the solar power supply unit 6 includes a photovoltaic controller 601 installed at the top end of the upright post 2, a plurality of groups of solar panels 602 are disposed on the outer side of the photovoltaic controller 601, and an energy storage colloid storage battery 603 connected with the photovoltaic controller 601 is disposed on one side of the upright post 2, so that the power supply can be supplemented to the data processing unit 5, and the situation that power is cut off in the monitoring process is avoided.
In particular, the solar power supply unit 6 is located at one side of the inspection well 1, at the highest position of the whole device, and 3.5 meters away from the ground.
In one embodiment, for the above-mentioned upright post 2, the upright post 2 is made of galvanized steel pipe, and the inside of the upright post 2 is of a hollow structure, and the sensor housing 401 and the bell jar housing 405 are made of polypropylene waterproof material, so that the sensor housing 401 and the sensor inside the bell jar housing 405 can be guaranteed to have stronger corrosion resistance integrally, so as to better prevent pollution, and effectively guarantee measurement accuracy.
In one embodiment, the data processing unit 5 is configured to perform processing and analysis operations on the data collected by the data sensing unit 4, and generate a decision instruction according to a preset algorithm to predict the state in the inspection well 1.
The data processing unit 5 has a dual power supply switching function, a microprocessor is arranged in the dual power supply switching function, collected data CAN be processed and analyzed in real time, a decision instruction is generated according to a preset algorithm, the dual power supply switching function comprises a dual-core ARM processor, FGPA, a memory, analog and digital I/O, communication interfaces such as Ethernet, CAN, USB, 232/485 and the like, a 4G communication module is integrated, and intelligent measuring equipment is used for switching dual power supplies between the solar power supply unit 6 and the mains supply unit 9, so that the power supply of the whole device is ensured to be sufficient.
The data processing unit 5 comprises a data collection marking module 501, a data division processing module 502, a processing model building module 503, a deployment monitoring implementation module 504 and an alarm decision processing module 505.
The data collection and marking module 501 is configured to collect parameter data inside the completed inspection well 1 collected by the data sensing unit 4, and perform a time marking operation on the parameter data according to a time point, where the parameter data includes water quality data, gas concentration data, water flow data and image data.
Specifically, the collected data sensing unit 4 collects the parameter data inside the completed inspection well 1, and performs a time marking operation on the parameter data according to a time point, including:
The data sensing unit 4 is used for collecting various parameter data in the inspection well 1, the collected data are sent to the storage device, and at the same time, each data point needs to be time-marked during data collection.
The time stamping is applied to the data, typically at the same time as the data acquisition, and the time stamped data is stored in a database or other storage system for subsequent data analysis and processing.
The method comprises the steps of collecting water quality, gas concentration and flow data and images in the inspection well 1 by utilizing a plurality of data sensing modules, marking time by the data, tracking the change of the data along with time, cleaning, denoising and standardizing the data, organizing the data into a time sequence format, screening and preprocessing the images in the inspection well 1, adding tags for the data at different time points in the inspection well 1, indicating the data state, such as normal, abnormal, alarm and the like, wherein threshold conditions are compared with historical data analysis, the deviation is normal within 5%, the deviation exceeds 5% is abnormal, the alarm is sent when the deviation exceeds 20%, and the alarm is sent when the maximum allowable concentration of methane exceeds 1%.
The data dividing processing module 502 is configured to divide the marked parameter data into a training set, a verification set and a test set according to a preset proportion, and generally divide the marked parameter data according to a proportion of 70% training, 15% verification and 15% testing.
And the processing model building module 503 is used for building a processing model by utilizing parameter data and convolutional neural network technology to realize the prediction of the internal abnormal condition of the inspection well 1.
Specifically, a deep learning frame is built, a model is built according to space-time correlation, a CNN framework suitable for processing multichannel data is designed, wherein each channel corresponds to water quality, gas concentration and flow, a convolution layer, a pooling layer and a full connection layer are built, data characteristic information and classification are extracted by a residual convolution neural network (ResNet) method, and the data characteristic information in the inspection well 1 is converted into water quality, gas concentration and flow by a convolution long and short memory neural network (ConvLSTM) method; the convolutional neural network comprises a convolutional layer, convolution kernel in convolution layer size 3 x 3.
The method for predicting the internal abnormal condition of the inspection well 1 by using the parameter data and the convolutional neural network technology to establish a processing model comprises the following steps:
and selecting vectors from the test set by utilizing a sliding search and time point marking mode to construct an input matrix with time characteristics.
It should be explained that, the method of selecting the vector from the test set by using the sliding search and the time point marking method to construct the input matrix with time characteristics includes:
selecting a characteristic vector of the abnormal condition monitored by the inspection well 1 and an output abnormal category of the processing model according to the test set, and performing abnormal value removal and normalization processing on the characteristic vector;
Setting a sliding search length value, sequentially carrying out feature vectors according to a time-stamped result, and carrying out sliding operation according to the sliding length value;
Acquiring time sequence characteristics in the time mark according to the operation result, and marking the time mark on the sliding window to acquire an overlapped subset sequence;
Traversing the subset sequences of all the feature vectors, combining the subset sequences into a sample set, and taking the sample set and the corresponding abnormal category as an input matrix.
And combining the input matrix with the convolutional neural network to establish a processing model, and optimizing model parameters by adopting an optimization algorithm and a training set.
It should be explained that, combining the input matrix with the convolutional neural network to build a processing model, and implementing optimization on model parameters by using an optimization algorithm and a training set includes:
Taking the input matrix as an input variable, taking a corresponding abnormal class as an output variable, and capturing a nonlinear relation among a plurality of input variables by utilizing a convolution layer in a convolution neural network;
Combining nonlinear relations of input variables by using a full-connection layer, and selecting the maximum value by using a pooling layer to obtain an abnormal class output model;
Judging a loss function of the output model, updating model parameters by using an optimization algorithm to obtain a processing model, and establishing a model evaluation standard by using a training set to judge a model parameter optimization result;
And when the optimization result is smaller than the standard value, the optimization is failed, and parameter optimization is continued until the optimization result is larger than the standard value.
And simulating the optimized processing model by using the verification set to output a prediction result of the inspection well 1.
Training the processing model by using a training set, updating model parameters by a deep learning neural network through a back propagation algorithm based on errors between model output and a true value until loss errors are minimized, selecting an optimal model by comprehensively comparing errors in a training period and a verification period, evaluating performance of the model by using the verification set, calculating performance indexes such as accuracy, recall, F1 score and the like, adjusting super parameters of the model according to verification results to optimize performance such as learning rate, batch processing size and iteration times, finally evaluating the performance of the model by using a testing set, and ensuring good generalization effect of the model on new data.
A deployment monitoring implementation module 504 is configured to input a processing model into a specified inspection well monitoring area to predict conditions in inspection well 1.
The alarm decision processing module 505 is configured to implement a comparison operation by using the prediction result and the threshold value, determine whether an abnormal condition exists in the inspection well 1 according to the comparison result, and take corresponding measures.
Specifically, a trained processing model is deployed into the inspection well 1 system for real-time monitoring and analysis. The data in the inspection well 1 is monitored in real time, the data are input into a processing model for prediction, a threshold value is set, an alarm is triggered when the model detects abnormal conditions, relevant personnel are notified or emergency measures are automatically taken, and corresponding decisions such as closing the inspection well, adjusting flow or executing water quality purification are taken according to the output of the model.
The safety monitoring unit 8 is used for designing a remote automatic monitoring model by utilizing the internet of things technology to monitor the inspection well cover 3, analyzing whether the inspection well cover 3 has abnormal conditions, and judging the stability of the data sensing unit 4.
The safety monitoring unit 8 includes a monitoring data acquisition module 801, a data transmission and reception module 802, a monitoring frame design module 803, an abnormal situation judgment module 804, an alarm design generation module 805, and a monitoring integration connection module 806.
The monitoring data acquisition module 801 is configured to install a sound sensor on a top side wall of the inspection well cover 3 to acquire sound data of the inspection well cover 3 within a preset distance.
Specifically, the sound sensor installed on the top side wall of the inspection well cover 3 collects sound data of the inspection well cover 3 within a preset distance, and the sound data includes:
According to the installation purpose, a sound sensor suitable for monitoring illegal opening of the inspection well cover 3 is selected, and necessary accessories and equipment such as an installation bracket, a protective cover, a connecting cable and the like are purchased in combination with the installation requirement.
And a proper position is selected on the side wall of the top end of the inspection well cover 3 for installing the sound sensor, so that the sound within a preset distance can be effectively captured.
The data transmission and receiving module 802 is configured to establish a sensing architecture by using a db amplification manner and introduce a sound template under the condition of normal monitoring of the system design of the internet of things, where the sound template includes a knocking noise sound and a carrying moving sound within a preset distance of the inspection well cover 3.
Specifically, the voice template for establishing a perception architecture by utilizing a decibel amplification mode and importing the system design of the internet of things under the condition of normal monitoring comprises the following steps:
the overall framework of the sensing framework is designed according to the decibel amplification technology, and comprises the data acquisition, processing, storage, analysis, response and the like, and the sensing framework and the Internet of things equipment are integrated to ensure smooth flow and processing of data.
Sensor equipment capable of carrying out sound decibel amplification is deployed in a monitoring area, sound data under normal conditions are collected as a reference template according to monitoring requirements, and the collected sound data are subjected to decibel amplification processing so as to improve the recognition accuracy of the sound data in a low-noise environment.
Key features such as frequency, rhythm, pitch and the like are extracted from the amplified sound data, and a sound template is established according to the sound features under the normal monitoring condition.
And accessing the sound sensor equipment into the Internet of things platform, and uploading data in real time to determine a data communication protocol between the Internet of things equipment and the platform.
The monitoring framework design module 803 is configured to add an automatic update monitoring weight function into the sensing framework according to the sound template, and build an automatic monitoring model by using the sound template and the sound data analysis fitness function.
Specifically, adding an automatic updating monitoring weight function in a perception architecture according to a sound template, and establishing an automatic monitoring model by utilizing the sound template and a sound data analysis fitness function comprises the following steps:
Performing conversion operation on the data in the designed sound template in a spectrum centroid mode to obtain a feature vector, and utilizing a preset monitoring requirement to define a fitness function to evaluate the matching degree between the feature vector and the sound template;
And transmitting the matching degree to a sensing framework to construct an automatic monitoring model, and judging the fitness score of preset sound data by combining the automatic monitoring model through a correlation analysis algorithm.
The step of transmitting the matching degree to the sensing architecture to construct an automatic monitoring model, and the step of judging the fitness score of the preset sound data by combining the automatic monitoring model through the association analysis algorithm comprises the following steps:
packaging the matching degree into a preset transmission format, transmitting the preset transmission format into a sensing framework according to a preset communication protocol, and receiving data from each matching degree by the sensing framework;
presetting a monitoring frame based on a perception framework, establishing a risk factor set as a sample by combining the matching degree, and setting the number of the samples and the risk factor set;
and constructing an expression frame according to the risk factor set to describe a sample state when the expression frame belongs to the risk state, and obtaining the rough membership of the sample to judge the occupation proportion of the risk factors.
Wherein, the calculation formula of the rough membership degree is as follows:
；
Wherein L represents the rough membership degree, A represents the conditional attribute of a sample, particularly the conditional attribute of the kth risk factor of the ith sample, x represents the risk factor, and card represents the cardinality of the set, namely the number of elements in the set, particularly the cardinality of A, namely the number of elements in A.
Judging the adaptation degree between each group of risk factors according to the occupation proportion of each group of risk factors and a preset ideal value, constructing an identification frame by using the adaptation degree, and combining a monitoring frame, an expression frame and the identification frame to construct an automatic monitoring model;
the calculation formula of the adaptation degree is as follows:
；
Where Q represents the fitness, r ik represents the sample state of the ith sample, x ik represents the kth risk factor of the ith sample, x 0k represents the ideal value of the corresponding kth risk factor, and C represents the fitness.
And inputting preset sound data into an automatic monitoring model to obtain risk occupation ratios and adaptation values, and sensing the adaptation degree scores of the sound data under different risk occupation ratios by using a correlation analysis algorithm.
And formulating a weight updating rule according to the fitness score result, and adding an automatic updating monitoring weight function into the perception framework based on the updating rule.
The abnormal condition judging module 804 is configured to input preset sound data in the automatic monitoring model, and judge a state of the sound template of the inspection well cover 3 under an abnormal condition according to an output result.
Specifically, inputting preset sound data in the automatic monitoring model, and judging the state of the sound template of the inspection well cover 3 under the abnormal condition according to the output result includes:
The sound templates are used as a reference for comparison, and sound data of the inspection well cover 3 in various preset abnormal states are collected and analyzed, including sound that is knocked, moved or destroyed to establish the sound templates in the abnormal states.
And inputting the sound templates in the abnormal state into a trained automatic monitoring model, analyzing the input sound data by the model, comparing the input sound data with preset normal and abnormal sound templates, and identifying the state of the sound templates.
Judging the current state of the inspection well cover 3 according to the output result of the model, and judging the inspection well cover to be in an abnormal state if the sound data is matched with the abnormal template; if the template is matched with the normal template, the normal state is judged.
The alarm design generating module 805 is configured to analyze the stability of the data sensing unit 4 according to the abnormal condition of the inspection well cover 3, and set a corresponding alarm reminding mode according to the stability result.
Specifically, analyzing the stability of the data sensing unit 4 according to the abnormal condition of the inspection well cover 3, and setting the corresponding alarm reminding mode by using the stability result includes:
It is ensured that sufficient abnormal situation data, including sound sensor data, is collected from the real-time monitoring of the inspection well lid 3 and the type and severity of the abnormal situation in the collected data is determined by means of an automatic monitoring model.
The state of the manhole cover 3 is judged based on the abnormal condition data, and the stability of the data sensing unit 4 is evaluated according to the state data thereof, for example, when the manhole cover 3 is moved, the data sensing unit 4 is exposed to the outside, and when the manhole cover 3 is moved, the data sensing unit 4 is at risk of falling into the manhole 1.
Different alarm levels are defined according to the result of the stability analysis, including slightly unstable triggering early warning, moderately unstable triggering alarm, and severely unstable triggering emergency response.
The monitoring integrated connection module 806 is configured to implement integrated connection operation of the automatic monitoring system and the sound sensor, analyze whether the inspection well cover 3 has an abnormal condition according to the real-time sound data, and determine stability of the data sensing unit 4 according to the analysis result.
Specifically, the integrated connection operation of the automatic monitoring system and the sound sensor is implemented, whether the inspection well cover 3 has an abnormal condition is analyzed according to the real-time sound data, and the judging of the stability of the data sensing unit 4 according to the analysis result includes:
the physical connection of the sound sensor and the automatic monitoring system is ensured to be correct, the sound sensor comprises a power supply, a data wire and the like, and when the inspection well cover 3 is specifically used, the sound sensor captures sound data around the inspection well cover 3 in real time and sends the data to the monitoring model.
The monitoring model preprocesses the collected sound data to identify whether an abnormal sound mode exists, such as prying, knocking and the like, judges whether the inspection well cover 3 is in an abnormal state based on the abnormal sound identification result, judges the stability of the data sensing unit 4, and ensures the high efficiency and the reliability of the system.
Therefore, by installing the sound real-time monitoring model at the top end of the inspection well cover 3, monitoring the sound state near the inspection well 1, judging whether abnormal conditions exist according to the sound state, abnormal sounds such as artificial damage and illegal opening of the inspection well cover 3 can be captured in real time, so that early identification of potential hazards or safety problems is realized, the inspection well cover 3 and the internal sensors can respond quickly by detecting the abnormal sounds in time, and measures are taken to prevent or reduce the risk of theft or damage of the inspection well cover 3 and the internal sensors.
As shown in fig. 9, according to another embodiment of the present invention, there is also provided a method for monitoring an inspection well based on a convolutional neural network, the method for monitoring an inspection well based on a convolutional neural network comprising the steps of:
S1, a data sensing unit 4 is arranged in the inspection well 1 to collect parameter data in the inspection well 1 and transmit the parameter data to a data processing unit 5;
S2, the data processing unit 5 processes and analyzes the acquired parameter data, and a processing model is utilized to monitor whether an abnormal condition exists in the inspection well 1;
S3, utilizing a safety monitoring unit 8 to remotely and automatically monitor the condition of the inspection well cover 3 within a preset distance, and judging whether the stability of the data sensing unit 4 is affected by abnormal conditions;
s4, the communication unit 7 uploads the monitoring results of the data processing unit 5 and the security monitoring unit 8 to the cloud platform 10, and the cloud platform 10 is utilized to issue the monitoring results to the mobile terminal 12 and the computer terminal 11.
In summary, by means of the technical scheme, the data sensing unit 4 is arranged, so that various sensors can be integrated into a whole in a multifunctional and multi-mode, the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time are realized, the water flow condition of the pipeline inside the inspection well 1 is accurately monitored, a decision basis is provided for subsequent processing analysis, and intelligent scheduling of a drainage system is realized. According to the invention, the data processing unit 5 is arranged to perform processing and analysis operation on the data acquired by the data sensing unit 4, and a decision instruction is generated according to a preset algorithm to predict the state in the inspection well 1. The solar power supply unit 6 is arranged, so that a stable power supply can be provided for the whole monitoring device, and the power supply is supplemented when the electric quantity is exhausted, so that the condition that the power supply is powered off in the monitoring process is avoided. The inspection well monitoring equipment provided by the invention integrates multiple functions and modes, has the functions of sensing underground liquid level change, monitoring pipe network flow, methane concentration and the like in real time, performs predictive analysis on data monitored in real time through a big data analysis technology, accurately monitors the water flow condition in a pipeline, improves the scientificity of reconstruction decisions of a drainage system and the operation management level of the system, and realizes intelligent scheduling of the drainage system. According to the invention, the safety monitoring unit 8 is arranged, the sound state near the inspection well 1 is monitored in real time by installing the sound real-time monitoring model at the top end of the inspection well cover 3, and whether the abnormal condition exists near the inspection well 1 is judged according to the sound state, so that the early identification of potential danger or safety problem is realized, the influence of artificial theft damage on various sensors in the inspection well 1 is avoided, and the accuracy of monitoring in the inspection well 1 is reduced.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "configured," "connected," "secured," "screwed," and the like are to be construed broadly and may be, for example, fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intermediaries, or in communication with each other or in interaction with each other, unless explicitly defined otherwise, the meaning of the terms described above in this application will be understood by those of ordinary skill in the art in view of the specific circumstances.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (8)
1. The utility model provides an inspection shaft monitoring devices based on convolutional neural network, includes inspection shaft (1), the top of inspection shaft (1) is provided with stand (2), its characterized in that, one side of stand (2) is provided with inspection shaft lid (3), the top and the bottom of inspection shaft lid (3) are provided with safety monitoring unit (8) and data perception unit (4) respectively, just data perception unit (4) install in the lateral wall of inspection shaft (1), the inside of stand (2) is provided with data processing unit (5), the top of stand (2) is provided with solar energy power supply unit (6), the top of data processing unit (5) just runs through stand (2) is provided with communication unit (7);
The safety monitoring unit (8) is used for implementing monitoring operation on the inspection well cover (3) by utilizing an Internet of things technology to design a remote automatic monitoring model, analyzing whether the inspection well cover (3) has an abnormal condition, and judging the stability of the data sensing unit (4);
The safety monitoring unit (8) comprises a monitoring data acquisition module (801), a data transmission and reception module (802), a monitoring frame design module (803), an abnormal condition judgment module (804), an alarm design generation module (805) and a monitoring integrated connection module (806);
The monitoring data acquisition module (801) is used for installing a sound sensor on the side wall of the top end of the inspection well cover (3) to acquire sound data of the inspection well cover (3) within a preset distance;
the data transmission and reception module (802) is used for establishing a perception architecture by utilizing a decibel amplification mode and importing a sound template under the condition of normal monitoring of the system design of the internet of things, wherein the sound template comprises knocking noise sound and carrying moving sound within a preset distance of the inspection well cover (3);
The monitoring framework design module (803) is used for adding an automatic updating monitoring weight function in the perception framework according to the sound template, and establishing an automatic monitoring model by utilizing the sound template and the sound data analysis fitness function;
the abnormal condition judging module (804) is used for inputting preset sound data in the automatic monitoring model and judging the state of the sound template of the inspection well cover (3) under the abnormal condition according to the output result;
The alarm design generation module (805) is configured to analyze stability of the data sensing unit (4) according to an abnormal condition of the inspection well cover (3), and set a corresponding alarm reminding mode according to a stability result;
The monitoring integrated connection module (806) is used for implementing integrated connection operation of the automatic monitoring system and the sound sensor, analyzing whether the inspection well cover (3) has an abnormal condition according to real-time sound data, and judging the stability of the data sensing unit (4) according to an analysis result;
The adding the function of automatically updating the monitoring weight in the perception framework according to the sound template, and establishing the automatic monitoring model by utilizing the sound template and the sound data analysis fitness function comprises the following steps:
Performing conversion operation on the data in the designed sound template in a spectrum centroid mode to obtain a feature vector, and utilizing a preset monitoring requirement to define a fitness function to evaluate the matching degree between the feature vector and the sound template;
transmitting the matching degree to a sensing framework to construct an automatic monitoring model, and judging the fitness score of preset sound data by combining an association analysis algorithm with the automatic monitoring model;
And formulating a weight updating rule according to the fitness score result, and adding an automatic updating monitoring weight function into the perception framework based on the updating rule.
2. The inspection well monitoring device based on the convolutional neural network according to claim 1, wherein the data sensing unit (4) comprises a sensor housing (401) mounted on the inner side wall of the inspection well (1), a turbidity sensor (402) is arranged at the inner bottom end of the sensor housing (401), a pH value sensor (403) is arranged at the top end of the turbidity sensor (402), and a water quality sensor (404) is connected to the bottom end of the turbidity sensor (402);
The data sensing unit (4) further comprises a bell jar type shell (405) arranged at the top end of the sensor shell (401), a gas sensor (406) is arranged in the bell jar type shell (405), a liquid level sensor (407) is arranged on one side of the gas sensor (406), an image sensor (408) is arranged on one side of the liquid level sensor (407), and the image sensor (408) is arranged on the outer side of the bell jar type shell (405).
3. The inspection well monitoring device based on the convolutional neural network according to claim 2, wherein the solar power supply unit (6) comprises a photovoltaic controller (601) arranged at the top end of the upright post (2), a plurality of groups of solar panels (602) are arranged on the outer side of the photovoltaic controller (601), and an energy storage colloid storage battery (603) connected with the photovoltaic controller (601) is arranged on one side of the upright post (2);
The stand column (2) is made of galvanized steel pipes, the inside of the stand column (2) is of a hollow structure, and the sensor shell (401) and the bell jar type shell (405) are made of polypropylene waterproof materials.
4. A manhole monitoring device based on a convolutional neural network according to claim 3, wherein the data processing unit (5) is configured to perform processing and analysis operations on the data collected by the data sensing unit (4), and generate a decision instruction according to a preset algorithm to predict the state in the manhole (1);
the data processing unit (5) comprises a data collection marking module (501), a data division processing module (502), a processing model building module (503), a deployment monitoring implementation module (504) and an alarm decision processing module (505);
The data collection and marking module (501) is used for collecting parameter data in the inspection well (1) which is collected by the data sensing unit (4), and performing time marking operation on the parameter data according to a time point, wherein the parameter data comprises water quality data, gas concentration data, water flow data and image data;
the data dividing and processing module (502) is used for dividing the marked parameter data into a training set, a verification set and a test set according to a preset proportion;
The processing model building module (503) is used for building a processing model by utilizing parameter data and a convolutional neural network technology to realize the prediction of the internal abnormal condition of the inspection well (1);
the deployment monitoring implementation module (504) is used for inputting a processing model into a designated inspection well monitoring area to predict the condition in the inspection well (1);
and the alarm decision processing module (505) is used for implementing comparison operation by using the prediction result and the threshold value, judging whether an abnormal condition exists in the inspection well (1) according to the comparison result and adopting corresponding measures.
5. The inspection well monitoring device based on the convolutional neural network according to claim 4, wherein the method for predicting the internal abnormal condition of the inspection well (1) by using the parameter data and the convolutional neural network technology to build a processing model comprises the following steps:
selecting vectors from the test set by utilizing a sliding search and time point marking mode to construct an input matrix with time characteristics;
combining an input matrix with a convolutional neural network to establish a processing model, and optimizing model parameters by adopting an optimization algorithm and a training set;
Simulating the optimized processing model by using a verification set to output a prediction result of the inspection well (1);
the method for constructing the input matrix with time characteristics by selecting vectors from the test set by utilizing a sliding search and time point marking mode comprises the following steps:
selecting a characteristic vector of the abnormal condition monitored by the inspection well (1) and an output abnormal category of the processing model according to the test set, and performing abnormal value removal and normalization processing on the characteristic vector;
Setting a sliding search length value, sequentially carrying out feature vectors according to a time-stamped result, and carrying out sliding operation according to the sliding length value;
Acquiring time sequence characteristics in the time mark according to the operation result, and marking the time mark on the sliding window to acquire an overlapped subset sequence;
Traversing the subset sequences of all the feature vectors, combining the subset sequences into a sample set, and taking the sample set and the corresponding abnormal category as an input matrix.
6. The inspection well monitoring device based on the convolutional neural network according to claim 5, wherein the combining the input matrix with the convolutional neural network to build a processing model and optimizing the model parameters using an optimization algorithm and a training set comprises:
Taking the input matrix as an input variable, taking a corresponding abnormal class as an output variable, and capturing a nonlinear relation among a plurality of input variables by utilizing a convolution layer in a convolution neural network;
Combining nonlinear relations of input variables by using a full-connection layer, and selecting the maximum value by using a pooling layer to obtain an abnormal class output model;
Judging a loss function of the output model, updating model parameters by using an optimization algorithm to obtain a processing model, and establishing a model evaluation standard by using a training set to judge a model parameter optimization result;
And when the optimization result is smaller than the standard value, the optimization is failed, and parameter optimization is continued until the optimization result is larger than the standard value.
7. The inspection well monitoring device based on the convolutional neural network according to claim 1, wherein the step of transmitting the matching degree to the sensing architecture to construct an automatic monitoring model, and the step of determining the fitness score of the preset sound data by combining the automatic monitoring model with the association analysis algorithm comprises the steps of:
packaging the matching degree into a preset transmission format, transmitting the preset transmission format into a sensing framework according to a preset communication protocol, and receiving data from each matching degree by the sensing framework;
presetting a monitoring frame based on a perception framework, establishing a risk factor set as a sample by combining the matching degree, and setting the number of the samples and the risk factor set;
Constructing an expression frame according to the risk factor set to describe a sample state when the expression frame belongs to the risk state, and obtaining the rough membership of the sample to judge the occupation proportion of the risk factors;
Judging the adaptation degree between each group of risk factors according to the occupation proportion of each group of risk factors and a preset ideal value, constructing an identification frame by using the adaptation degree, and combining a monitoring frame, an expression frame and the identification frame to construct an automatic monitoring model;
and inputting preset sound data into an automatic monitoring model to obtain risk occupation ratios and adaptation values, and sensing the adaptation degree scores of the sound data under different risk occupation ratios by using a correlation analysis algorithm.
8. A method for monitoring an inspection well based on a convolutional neural network, for implementing monitoring of the inspection well monitoring device based on a convolutional neural network according to any one of claims 1 to 7, characterized in that the inspection well monitoring method based on a convolutional neural network comprises the following steps:
S1, installing the data sensing unit (4) in the inspection well (1) to collect parameter data in the inspection well (1) and transmitting the parameter data to the data processing unit (5);
s2, the data processing unit (5) processes and analyzes the acquired parameter data, and a processing model is utilized to monitor whether an abnormal condition exists in the inspection well (1);
S3, utilizing the safety monitoring unit (8) to remotely and automatically monitor the condition of the inspection well cover (3) within a preset distance, and judging whether the stability of the data sensing unit (4) is influenced by abnormal conditions or not;
s4, the communication unit (7) uploads the monitoring results of the data processing unit (5) and the safety monitoring unit (8) to a cloud platform, and the cloud platform is utilized to issue the monitoring results to a mobile terminal and a computer terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410400652.8A CN117990162B (en) | 2024-04-03 | 2024-04-03 | Inspection well monitoring device and method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410400652.8A CN117990162B (en) | 2024-04-03 | 2024-04-03 | Inspection well monitoring device and method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117990162A CN117990162A (en) | 2024-05-07 |
| CN117990162B true CN117990162B (en) | 2024-06-14 |
Family
ID=90902374
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410400652.8A Active CN117990162B (en) | 2024-04-03 | 2024-04-03 | Inspection well monitoring device and method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117990162B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104697587A (en) * | 2015-04-01 | 2015-06-10 | 成都城联科技有限公司 | Monitoring device for underground drainage pipe network and working mechanism of monitoring device |
| CN115471982A (en) * | 2021-12-28 | 2022-12-13 | 盐城工学院 | Inspection well cover intelligent management system based on Internet of things and big data technology |
| CN117007122A (en) * | 2023-08-03 | 2023-11-07 | 深圳市安天科技有限公司 | Monitoring method, device and system for well lid abnormality |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104131581A (en) * | 2013-11-26 | 2014-11-05 | 成都科创佳思科技有限公司 | Electrically-controlled antitheft inspection well cover |
| CN104677418A (en) * | 2015-01-28 | 2015-06-03 | 上海市政工程设计研究总院（集团）有限公司 | Intelligent drainage manhole cover monitoring system and monitoring method |
| CN206817179U (en) * | 2017-05-03 | 2017-12-29 | 南京朗坤自动化有限公司 | Drainage pipeline networks on-line monitoring system |
| CN109142518A (en) * | 2017-06-27 | 2019-01-04 | 田海飞 | A kind of inspection shaft and well lid working status monitoring system based on noise measuring |
| CN108564190A (en) * | 2017-12-14 | 2018-09-21 | 佛山科学技术学院 | A kind of Consultative System for Urban Disaster |
| CN109839156A (en) * | 2019-03-08 | 2019-06-04 | 安徽中科中涣防务装备技术有限公司 | A kind of three-dimensional urban road drainage safety on-line monitoring early warning system |
| CN114895060A (en) * | 2022-04-21 | 2022-08-12 | 中山大学 | Municipal drainage pipeline flow speed and flow intelligent assessment method and system |
| CN116822352A (en) * | 2023-06-20 | 2023-09-29 | 天地（常州）自动化股份有限公司 | GRNN neural network-based underground waterproof sealing wall working condition prediction method and system |
-
2024
- 2024-04-03 CN CN202410400652.8A patent/CN117990162B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104697587A (en) * | 2015-04-01 | 2015-06-10 | 成都城联科技有限公司 | Monitoring device for underground drainage pipe network and working mechanism of monitoring device |
| CN115471982A (en) * | 2021-12-28 | 2022-12-13 | 盐城工学院 | Inspection well cover intelligent management system based on Internet of things and big data technology |
| CN117007122A (en) * | 2023-08-03 | 2023-11-07 | 深圳市安天科技有限公司 | Monitoring method, device and system for well lid abnormality |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117990162A (en) | 2024-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105571644B (en) | A kind of insulator metallic dust filth on-Line Monitor Device and its monitoring method | |
| AU2020102518A4 (en) | A method of gas sensor calibration based on linear optimization | |
| CN117371337B (en) | Water conservancy model construction method and system based on digital twin | |
| CN115118333A (en) | Antenna health management system and method for satellite ground station | |
| CN115022187B (en) | Situation awareness method and device for electric-gas comprehensive energy system | |
| CN116595062A (en) | Park carbon emission monitoring and early warning method | |
| CN116295637A (en) | Intelligent monitoring method and system for safety of high-altitude building | |
| CN117990162B (en) | Inspection well monitoring device and method based on convolutional neural network | |
| CN109374513A (en) | Pipeline external anti-corrosion layer testing and analysis system | |
| CN117494009A (en) | Electrical equipment state evaluation method based on insulating material pyrolysis analysis and cloud platform | |
| CN116822352A (en) | GRNN neural network-based underground waterproof sealing wall working condition prediction method and system | |
| CN117108458A (en) | Marine wind power partition corrosion safety assessment method, system, equipment and storage medium | |
| CN116311775B (en) | Wearable near-electricity alarm equipment and monitoring system | |
| CN115479219B (en) | Intelligent pipeline state monitoring method, monitoring device and intelligent pipeline system | |
| CN117191147A (en) | Flood discharge dam water level monitoring and early warning method and system | |
| CN116777088A (en) | Power supply emergency repair environment monitoring method and system for guaranteeing life safety | |
| CN114544668B (en) | CCTV pipeline on-line measuring system | |
| CN114563040A (en) | Hydropower station hydraulic tunnel monitoring system and working method thereof | |
| CN112560916B (en) | Wind power tower barrel overturning intelligent diagnosis method based on tilt angle sensor information | |
| CN112378480B (en) | Intelligent pit data analysis method and system | |
| CN103472192A (en) | Intelligent positioning method of gas sensor | |
| KR20180120413A (en) | System for processing and analyzing magnetotellurics monitoring data using deep learning algorithm | |
| CN113917287A (en) | Substation bus joint discharge heating fault monitoring system | |
| CN113156543A (en) | Remote-measuring automatic weather station system and weather forecasting method thereof | |
| CN112395278B (en) | Environment-friendly detection system based on environment-friendly Internet of things and working method thereof |
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 |