Safety monitoring system based on internet of things tailings dam

Abstract

The tailings dam safety monitoring system is a system that plays an important role in disaster prevention and reduction. This paper divides the whole network of mine dam safety monitoring systems into two parts, that is, the basic wireless network and GPRS network from the gateway to the monitoring center. First, it’s the hardware of the design module, the network is divided into uplink communication and downlink communication, uplink communication is to upload the dam data collected by the terminal node to the gateway through the network. Downlink communication is the instruction sent by the gateway to the monitoring center, send to terminal nodes through data conversion between protocols. Secondly, the gateway also needs to solve the data conversion between the network and GPRS network protocol, so that the entire security monitoring system can communicate accurately. The communication between gateway and monitoring center is realized through the GPRS network, this requires adding the GPRS communication module to the gateway module. To increase the diversity of communications, communication between GSM short message communication methods and monitoring centers has also increased. The experimental results prove that the system in this paper can be applied to the mine dam safety monitoring system, meet the design requirements.

Keywords

Internet of things safety monitoring wireless sensor networks tailing dam

1 Introduction

Due to hydrology, geology, construction quality, operating conditions, and aging of the dam, many mine dams are subject to water seepage, deformation, damage, and other hazards, both in whole and in part [1]. According to statistics, most mines currently have safety hazards, which will greatly threaten the lives of thousands of miners [2]. At the same time, it has an impact on the normal operation of mineral exploration projects. In recent years, accidents such as the collapse of the mining dam and water seepage have occurred, causing huge economic losses [3]. The accident is mainly caused by the following two reasons: First, the management level of technicians is not high, there is no systematic training, and there are few experienced technicians. Secondly, the mine dam safety monitoring system is very imperfect and does not pay attention to the use of advanced equipment for monitoring. If these two points can be done well, the safety hazards of the dam can be monitored to a large extent so that they can be discovered and processed in time [4].

It is well known that dam safety monitoring instruments are the ears and eyes that understand the state of operation of the dam. Therefore, in some respects, compared with other industrial safety monitoring industries, the requirements are more stringent and more complex [5 –7]. Observing the static level from the outside, the laser permeation instrument until internal observation, the settler, the inclinometer, the soil strain gauge, the earth pressure gauge, and its automatic telemetry technology is all based on the high-reliability sensor IoT technology. With the increase of large-scale mine dam construction and the application of high-tech in recent years, the mine dam safety monitoring industry is moving towards integration, automation, digitization, and intelligence. IoT technology is a new technology that integrates sensors, controllers, computing power, and communication functions. They interact with external physical environments and transmit collected information to other computing devices, such as traditional computers, through a sensor network. With the rapid development of sensor technology, embedded computing technology, communication technology, and semiconductor and MEMS manufacturing technology, the manufacture of small and small, flexible, and low-power Internet of Things technology has gradually become a reality [8].

Sun E and other scholars use the cloud platform to predict the future state of the immersion line and obtain the saturation line equation [9]. A numerical simulation model was established by considering the saturation line prediction equation, the limit equilibrium state parameters, the reservoir water level, and rainfall. Then use the cloud platform to solve the safety factor, random reliability, and interval non-probabilistic reliability. Combined with the real-time monitoring of the deformation trend, combined with the calculated dynamic safety factor, random reliability, interval non-probabilistic reliability, the remote real-time early warning system can be used to obtain the warning signal of stable or dangerous tailings dam [9]. Wei Z and other scholars designed a monitoring and alarm system based on ZigBee wireless sensor network and LabVIEW. The system selects the single-chip CC2530 as the RF transceiver, completes the hardware and software design of the solar sensor node, and realizes the collection, transmission, and processing of the tailings library data. The upper computer monitoring alarm interface is completed by LabVIEW software, which displays the monitoring parameters in real-time and alarms in real-time. It uses the VISA serial resource module and the SQL access database to transfer and store monitoring data. Through the Matlab script node, the data processing module of the PC calls Matlab software, and the regression model of Support Vector Machine (SVR) is established on the back panel of Labview [10].

This paper divides the network of the entire mine dam safety monitoring system into two parts, namely the basic wireless network and the GPRS network from the gateway to the monitoring center. First, it is the hardware of the design module. The network is divided into uplink communication and downlink communication. The uplink communication is to upload the mine dam data collected by the terminal node to the gateway through the network. The downlink communication is an instruction sent by the gateway to the monitoring center and is sent to the terminal node through data conversion between the protocols. Secondly, the gateway also needs to solve the data conversion between the network and the GPRS network protocol, so that the entire security monitoring system can communicate accurately. The communication between the gateway and the monitoring center is implemented via a GPRS network, which requires the addition of a GPRS communication module to the gateway module. To increase the diversity of communication, communication between the GSM short message communication method and the monitoring center has also increased. The experimental results show that the system can be applied to the mine dam safety monitoring system and meet the design requirements.

2 Proposed method

2.1 Internet of things technology

A network that interconnects all projects with the Internet by agreed communication protocols for intelligent identification, location, analysis, monitoring, and management. The connection between objects and objects is the most obvious feature of the Internet of Things, which extends a service object from one object to another. From the perspective of technical definition understanding, the Internet of Things has its distinctive characteristics compared with the traditional Internet, which can be summarized into comprehensive sensing, reliable transmission, and intelligent processing.

(1) Key technologies of the Internet of Things

The Internet of Things satisfies objects that can be closely connected with other people and objects. They realize information sharing between them, and can automatically process the transmitted information through the device. The Internet of Things application can fully acquire data information. The system consists of the sensing layer. The network layer and intelligent processing are composed of three layers. The Internet of Things technologies applied in the three layers is different. Knowing the key technologies will accelerate the rapid development of the Internet of Things in various fields.

1) Key technologies of the perception layer. As the foundation layer of the Internet of Things architecture, the perception layer is to complete item identification and data acquisition. Key technologies include radio frequency identification, fast positioning, sensing, and laser technology. RFID is an automatic identification technology that uses radiofrequency technology, information transfer between equipment and objects, and can identify scanned objects. Its advantages include: strong anti-interference ability, can work normally in harsh environments, and errors occur. The possibility is very small, and the application method of this device is easy to grasp. It includes two parts: hardware and software components. The hardware components include RFID electronic tags and scanning devices, software components are divided into application software and middleware. Wireless sensor network technology includes many small wireless sensor devices scattered on the control object. The device obtains pressure, temperature, and other information to form a multi-hop wireless self-built network. The objects are connected wirelessly and the acquired data information is transmitted in real-time. Give the required personnel. Geographic information system, based on computer software technology and hardware equipment, provides geographic information data management for geographic information data operations, can meet customer needs for maps in real-time and can update electronic maps in time. The global positioning system is an efficient positioning method that emerges with the advancement of technology. Communication equipment can obtain real-time information through satellite positioning. Its advantages include: real-time, high accuracy, wide-coverage, and use satellite to calculate The position of a certain moment, through the corresponding calculation formula, can then determine the specific location of the object.

2) Key technologies in the network layer. Based on the existing Internet, mobile communication network, and radio and television network, the network layer transmits the information data collected by the sensing layer through the incoming and outgoing tools to transmit the information data. The technology used in the process is the third generation of mobile communication technology. Compared with the previous two generations of technology, it has the advantages of fast transmission speed, high precision, and strong compatibility.

3) Key technologies of the intelligent layer. The technology of this layer mainly includes cloud computing, data mining, and middleware. Based on these technologies, devices can control things autonomously and complete the functions of providing customers with the required services. Cloud computing technology is the core technology of this layer. This advanced computer technology can share all the information in the whole network, and then use the computer software to centrally manage the shared information, and provide different information according to the different needs of customers. For example, functions such as high accuracy of information transmission, fast transmission of information, a high degree of resource sharing, and the ability to provide services according to customer needs can be accomplished through cloud computing.

(2) Combination of Internet of Things and cloud computing

Cloud computing provides distributed technologies such as distributed systems, parallel computing, load balancing, network storage, data mining, and elastic computing, all of which can be tightly integrated with IoT systems.

Distributed and Parallel Computing: Distributed is a network-based computer processing technology that is a large-scale server cluster that can solve unreliable IoT server nodes.

Load balancing: Load balancing refers to the fact that the initial independent application is assigned to multiple operating units for execution, thereby solving the performance bottleneck problem of a single operating unit or module. The load balancing technology has deep application in the data acquisition module, storage module, calculation module, and API interface module of the Internet of Things system, and can provide stable and reliable technical support for the Internet of Things platform.

Elastic calculation: The server has the risk of crashing at any time. It only increases the server to increase the cost. When the amount of data is small, it will cause a waste of resources. Therefore, the flexibility of the computing power of cloud computing can solve this problem well.

Countless terminal devices in the Internet of Things are collecting, transmitting, and exchanging data. The huge amount of data requires a powerful cloud storage platform to meet application needs.

Data mining: The Internet of Things can quickly extract useful and understandable information from large amounts of data through highly automated data analysis and inductive reasoning.

Service Model: Cloud computing can maximize information sharing on the Internet of Things and combine it with cloud computing distributed across large server clusters around the world. Information on the Internet of Things can be limited by geographic location to maximize information sharing on the “cloud”.

2.2 Wireless sensor network

(1) Wireless sensor network structure

The so-called sensor network system is shown in Fig. 1, mainly includes management nodes, aggregation nodes, and sensor nodes. A large number of sensor nodes are randomly distributed inside and around the area to be monitored, and a network can be formed by self-organization. The data monitored by the sensor node is first transmitted hop by hop along with another sensor node. And after multiple hops, it is routed to the aggregation node and eventually reaches the management node via the Internet and satellite. The user then manages and configures the sensor network through the management node, issues monitoring tasks, and collects monitoring data.

Fig. 1

Sensor network structure.

Sensor nodes are typically micro-embedded systems with relatively low storage, processing, and communication capabilities. From the perspective of network function, in addition to local information collection and data processing, there are also sensor nodes that need to store, manage and integrate data forwarded by other nodes, and cooperate with other nodes to complete certain tasks. In other words, it takes into account the dual functions of the terminals and routers of traditional network nodes.

(2) Characteristics of wireless sensor networks

1) Application-based network

Wireless sensor networks can be used in a variety of applications through a combination of various sensing technologies, communication technologies, and processing computing technologies. For different applications, wireless sensor networks need to adjust their configuration. That is, the specific application requires a specific configuration.

2) Interact with the physical world

By combining wireless sensor networks with the physical world, physical phenomena can become the center of the network, making it a replacement for people. In the unattended situation, the network should be able to proactively perceive the external environment and reflect changes in the external environment in real-time. The ability of its system state depends on changes in the physical environment, application scenarios, and conscious calculations. It is centered on the monitoring object, and the ability to dynamically perform corresponding actions in real-time is related to changes in surrounding scenes.

3) Self-organizing network

The sensor nodes can form a wide monitoring area by randomly expanding the self-assembled network and self-configuring through the Adhoc method to form a network. At any time, the number of hops between a node and a server can be calculated by a sensor node carrying a special algorithm. And at any given time, it is determined which neighboring node can provide the best way to reach the data collection point. Due to environmental changes (such as node failures, detection of object changes, or arbitrary node movements), the network topology changes dynamically over time, so the network must be able to automatically maintain dynamic routes.

4) Data-centric network

It not only ensures the transmission of data but also reduces traffic. Due to the high node density and a large number, the global address is not practical. The distributed data flow management, distribution, query, and mining methods should be studied.

2.3 Gaussian process regression mine dam monitoring model

(1) Gaussian process

Suppose there is a training set S = { (x_i, y_i) |i = 1, 2, . . . , n } = (X, y), where x_i ∈ R^p is the p-dimensional input vector, X = [x₁, x₂, . . . , x_n] is the p×n-dimensional input matrix, y_i ∈ R is the corresponding output scalar, and y is the output vector. The main purpose of the regression is to learn the relationship between the input matrix X and the output vector y according to the training set, that is, to give a new input vector x^* to obtain the predicted distribution of the corresponding observation value y^* based on the training set S.

From the perspective of function space, the Gaussian process f (x) can be determined by the corresponding mean function m (x) and the covariance function, as defined by Equations (1) and (2): $m (x) = E (f (x))$ (1) $k (x, x^{'}) = E ((f (x) - m (x)) (f (x^{'}) - m (x^{'})))$ (2)

In the formula, x and x′ ∈ R^p are arbitrary random variables.

Gaussian process functions can be defined as Equation (3): $f (x) \sim GP (m (x), k (x, x^{'}))$ (3)

But to simplify the process, we can preset the mean function m (x) to zero.

(2) Gaussian process regression

In the problem of Gaussian process regression, we assume that the relationship between the input vector and the output scalar is as shown in Equation (4): $y_{i} = f (x_{i}) + ɛ$ (4)

In formula (4), f (x_i) is the value of any regression function, y_i is the output observation of noise pollution, noise ɛ satisfies the mean value of 0, and the variance is the Gaussian distribution of σ², ɛ ∼ N (0, σ²). Furthermore, we assume that f = [f (x) , f (x₂) , . . . , f (x_n)] ^T satisfies the definition requirement of the Gaussian process, that is, p (f|x₁, x₂, . . . , x_n) = N (0, K), where K is the covariance matrix of the element f (x_i). $K (X, X) = (\begin{matrix} k (x_{1}, x_{1}) & k (x_{1}, x_{2}) & . . . & k (x_{1}, x_{n}) \\ k (x_{2}, x_{1}) & k (x_{2}, x_{2}) & . . . & k (x_{2}, x_{n}) \\ \begin{matrix} . \\ . \\ . \end{matrix} & \begin{matrix} . \\ . \\ . \end{matrix} & \begin{matrix} . \\ . \\ . \end{matrix} & \begin{matrix} . \\ . \\ . \end{matrix} \\ k (x_{n}, x_{1}) & k (x_{n}, x_{2}) & . . . & k (x_{n}, x_{n}) \end{matrix})$ (5)

In Equation (5), K (X, X) = K_n ={ K_ij }, matrix element K_ij is the covariance between the f (x_i) and f (x_j) values.

The Gaussian process regression is used to calculate the predicted distribution of the function values of the f^* corresponding to the test point $X^{*} = [x_{1}^{*}, x_{2}^{*}, . . ., x_{m}^{*}]$ . Among them, f_i represents the f (x_i) function value, which is completely connected by the index set. Each connection represents the correlation between two variables, which are defined by the covariance function. The distribution of y is conditional on the value of f, which can be derived from an isotropic Gaussian distribution: $P (y | f, X) = N (f, σ_{n}^{2} |)$ (6)

In the formula (6), represents an identity matrix.

By the nature of the Gaussian distribution, we can get the marginal distribution: $P (y | X) = \int P (y | f, X) P (f | X) df = N (0, K + σ_{n}^{2} |)$ (7)

The joint probability distribution of the observed value and the function value of the test point can be written as: $[\begin{matrix} y \\ f \end{matrix}] \sim N (0, [\begin{matrix} K (X, X) + σ_{n}^{2} | & K (X, X) \\ K (X, X) & K (X, X) \end{matrix}])$ (8)

According to the Bayesian principle and the conditional probability characteristics of the joint normal distribution, we can get the posterior probability distribution: $P (f_{*} | X, y, X_{*}) \sim N ({\bar{f}}_{*}, cov (f_{*}))$ (9) ${\bar{f}}_{*} = K (X_{*}, X) [K (X, X) + σ^{2} |]^{- 1} y$ (10)

$\begin{matrix} cov (f_{*}) = K (X_{*}, X_{*}) - K (X_{*}, X) \\ [K (X, X) + σ^{2} |]^{- 1} K (X, X_{*}) \end{matrix}$ (11)

In the above formula, ${\bar{f}}_{*}$ stands for posterior mean and cov (f_*) stands for posterior covariance. The estimated value of the output not only contains the expected prediction of f_*, but also has a reference value for the uncertainty of the data.

For a particular problem, choosing the appropriate covariance function is an important part of the Gaussian process, because the choice of the covariance function determines the nature of the sample function extracted from the Gaussian process, such as smoothness, length scale, amplitude, and so on. We use the most commonly used square exponential covariance function, as shown in Equation (12).

$\begin{matrix} k (x_{p}, x_{q}) = & σ_{f}^{2} exp {(- \frac{1}{2 |^{2}} (x_{p} - x_{q}) (x_{p} - x_{q}))}^{T} \\ + σ_{n}^{2} δ (p, q) \end{matrix}$ (12)

In Equation (12), θ = (ln |, ln σ_f, ln σ_n) stands for hyperparameter and $σ_{n}^{2}$ stands for a noise variance. δ (p, q) stands for Roneck function, which is equal to 1 if p = q, otherwise equal to 0.

3 Design of mine terminal monitoring system for safety monitoring

3.1 The overall design of the field terminal system

The dam safety monitoring field terminal system is an important part of the whole system. It is mainly responsible for the summary and forwarding of field data, as well as the necessary data processing and display, which is convenient for on-site staff to monitor the operation status of the mine dam in real-time. From the application point of view, the terminal system is installed in the field, which requires the advantages of small size, low power consumption, flexible configuration, etc., and also need to adapt to various bad weather guarantee systems to work reliably. Therefore, consider using embedded systems for development and design. The overall design of the field terminal system is shown in Fig. 2. It mainly includes the embedded system software and hardware platform construction, SQLite database design, GPRS, Beidou remote communication module design, and human-computer interaction interface design.

Fig. 2

Block diagram of the mine terminal system for mine safety monitoring.

3.2 Hardware and software platform construction

Before developing, you need to build a corresponding software and hardware platform. The field terminal system development platform is divided into three parts: ARM Cortex-A8 processor and peripherals, embedded Linux operating system, and Qt/Embedded development environment.

When designing the field terminal system hardware platform, you need to consider the following functional requirements: able to communicate with ZigBee, GPRS, Beidou module through the serial port; able to run embedded Linux operating system; can store and manage collected data, need to transplant SQLite embedded The database adopts a touch screen LCD, which requires rich interface functions and smooth program running. At the same time, it also needs to consider the stability, power consumption, cost, and wide application of the system.

According to the above functional requirements, there are usually two ways to develop embedded systems: one is to design the hardware platform from scratch and then to transplant the operating system; the other is to use the existing development board for preliminary design, waiting for technology after the maturity, the development board was tailored. To shorten the development cycle and concentrate on the bursting of software functions, this paper chooses the latter. Adopting the OK210 development board of Feiling Embedded Co., Ltd., which is based on Samsung’s ARM CorteX-A8 architecture processor S5PV210, integrates rich peripheral resources. The OK210 development board adopts the “core board+backplane” structure. The core board is the smallest system board of the main processor S5PV210, including 512MB DDR2, 512MB NAND Flash, power management chip ATC8937, and so on. The core board is connected to the bottom plate by two 100-pin double-row sockets. The backplane is mainly for various peripheral interfaces, including RGB liquid crystal interface, Jtag interface, Ethernet-interface, RS-232 interface, UART interface, USB interface, and the like.

The S5PV210 is an application processor from Samsung that is suitable for multimedia devices such as smartphones and tablets. Supporting DDR2 and multiple types of NAND Flash, the S5PV210 has powerful internal resources and video processing capabilities and can play a 1080P large-size HD video. The main features of the S5PV210 chip are as follows:

(1) ARMCORtexTM-A8 core, ARMV7 instruction set, main frequency up to 1 GHz, 64/32 bit internal bus structure, 32/32KB data/instruction level 1 cache, 512KB L2 cache, 2000DMIPS per second (200 Mbps) High-performance computing power for computing 200 million instruction sets.

(2) contains many powerful hardware codec functions, integrated MFC (Multi-Format Codec), supports MPEG4/MPEG2, H.264/H.263, VC-1 format video codec, supports JPEG hardware codec, maximum Support 8000×8000 resolution, with high-performance Power VR SGX540 3D graphics engine and 2D vector graphics, support 2D / 3D graphics acceleration, with TFT LCD controller.

(3) Provides a wealth of peripheral interfaces, including a 3-channel PCM serial interface, three 24-bit IIS interfaces, USB Host 2.0/OTC 2.0, LCD interface, three IIC interfaces, four UART interfaces, and two SPI interfaces., a modem interface and a large number of GPIO interfaces, etc., these interfaces greatly facilitate the connection of peripheral devices and facilitate the expansion of external functions.

3.3 Mine dam safety monitoring information management software

The real-time monitoring system consists of a monitoring center station and six monitoring stations. The monitoring center station consists of a computer and wireless digital radio station, which is the command center of the entire real-time monitoring system. The mine dam safety monitoring information management software can be divided into data acquisition software, analysis processing software, database management software, and engineering safety management software according to different functions. The entire management software system adopts a graphical and modular design to facilitate user’s use and system expansion.

1) Data acquisition software

The data acquisition software consists of two parts: manual acquisition and automatic acquisition.

For manually collected data, the software provides a human-machine interface window that is directly input into the library through the keyboard; the automatic acquisition software is a set of graphics window software. At the same time, each monitoring point in the layout communicates with the corresponding instrument of the field data acquisition unit. The automatic acquisition software can be used for single-machine acquisition and network acquisition. Any MCU machine can control the monitoring points for data acquisition and transmit the collected data to the monitoring center station.

2) Analysis and processing software The analysis and processing software is used to process and calculate the monitored seepage data, and then evaluate the actual operating state of the mine dam. The measured value can be converted into standard monitoring quantity, and the monitoring quantity is calculated and analyzed, and an optional analysis factor is provided for the user to select. Provides rich graphics capabilities to window the entire analysis process and graphically analyze the results. The liquid level height is calculated from the osmometer value and the water immersion surface is fitted.

3) Database management software

The database management software uses the SQL Server database platform for data conversion and connection and has security management functions such as database access level and database access record.

4) Engineering safety management software

The contents of engineering safety management software include engineering document management and graphic report production. Among them, graphic production mainly includes process diagrams, distribution diagrams, related diagrams, contour maps, histograms, and pie charts. Reports mainly include customized comprehensive monthly reports, annual reports, etc., to suit the different needs of users.

5) Network system management software

Network system management software includes system installation, system initialization, system configuration, system security management, system monitoring subnet management, system and LAN, and remote network connection.

6) Remote monitoring and remote assistant software

Strengthen the management and maintenance of the automatic monitoring system. The system provides remote monitoring and remote service functions. The superior monitoring department can use the remote monitoring function to remotely collect the automatic monitoring system to understand the safety status of the mine.

4 Discussion

4.1 Safety monitoring system for tailings dam

Figure 3 shows the main monitoring interface of the system, in which the specific positions of different sensors on the dam are marked. The sensor types include the seam gauge, the three-direction strain gauge set, the stress-free gauge, and the osmometer, and further, The button at the bottom of the interface can realize the monitoring of the sensor of each dam section. Also, the interface can realize the connection and query of the configuration and monitoring software data. Click the “Database Settings” button on the screen to enter the database setting screen to realize the configuration. The connection between the software and the dam-based database and the new database.

Fig. 3

Main monitoring interface.

This system is based on the safety monitoring system designed and operated by the dam-based data acquisition system. Therefore, the dam-based data acquisition software should be run in the background before running this software. Otherwise, the data cannot be updated. To enter the system, the database must be set first. The database is the system. The core component provides strong support for the system display. Therefore, the path of the database applied in this system cannot be changed. As shown in Fig. 4, the upper part of the interface is the operation mode selection, manual operation, and automatic query time setting. The operation modes are divided into automatic mode and manual mode.

Fig. 4

Database settings interface.

4.2 Testing based on the safety monitoring system of the internet of things tailings dam

As shown in Fig. 5, the table shows the accuracy of the different prediction models. It can be seen from the following data: the multiple linear regression model (MLR) and the stepwise regression model (SR) have almost the same accuracy. The error of the Gaussian process regression model (GPR) is smaller than that of the multiple linear regression model and the stepwise regression model. It can be concluded that the Gaussian process regression model performs better than the multiple linear regression model and the stepwise regression model inaccuracy. Comparing the GPR-1 model with the GPR-2 model, it can be found that the performance of two Gaussian process regression models with different quantitative variables is almost the same. According to the data in Fig. 5, the GPR-1 model with more variables is more than GPR- 2 The model performed slightly better.

Fig. 5

Comparison of the accuracy of different models.

From Table 1 and Figs. 6 and 7, we can see that the Gaussian Process Regression Model (GPR) shows higher precision and generalization ability than the artificial neural network. The parameters of the Gaussian process regression model are automatically determined, while the RBF neural network model requires multiple attempts to achieve the purpose of selecting relatively appropriate parameters. The prediction result of the Gaussian process regression model also includes the standard deviation value to judge the uncertainty of the prediction result. Figure 6 shows the displacement prediction value and its confidence range of the GPR-1 model, and since the GPR-2 model has almost no gap with the GPR-1 model in displacement prediction, this paper only gives the displacement prediction of the GPR-1 model. Figure from Fig. 6, it can be concluded that in the results of the test samples, the actual measured displacement values are within the 95% confidence interval or fairly close to the 95% confidence interval.

Table 1

Accuracy of different models

Algorithm	Train				Test
	E_max	A_rmenian	RMSE	R	E_max	A_rmenian	RMSE	R
MLR	1.77	0.51	0.62	0.98	1.76	0.71	0.81	0.97
SR	1.73	0.48	0.63	0.97	1.74	0.69	0.80	0.97
RBF-1	1.11	0.29	0.38	0.99	1.31	0.65	0.73	0.98
RBF-2	1.07	0.34	0.41	0.98	1.24	0.60	0.69	0.98
GPR-1	0.54	0.17	0.21	0.99	1.25	0.53	064	0.98
GPR-2	0.51	0.14	0.19	0.99	1.31	0.59	0.72	0.98

Fig. 6

GPR-1 model residual normal distribution test chart (training sample).

Fig. 7

GPR-1 model residual normal distribution test chart (test sample).

5 Conclusions

This paper analyzes the necessity of current mine safety monitoring and introduces the overall framework and main functions of the mine dam safety monitoring system. The mine dam safety monitoring system uses multi-sensor data fusion technology, and the ultimate goal is to more accurately analyze mine dam monitoring data.

This paper focuses on the more common concrete dams. Based on the traditional monitoring model, the Gaussian process regression algorithm is introduced to establish the Gaussian process regression monitoring model, and the actual engineering data. The performance comparison between the Gaussian process regression model and several other more traditional statistical models is carried out and summarized.

Footnotes

Acknowledgments

This research has been supported by the Natural Science Foundation of Fujian Province, China (No. 2017Y0066).

References

Kovaliova

, Kadnikov

V.V.

, Antsiferov

D.V.

, et al. Genome sequence of the acid-tolerant Desulfovibrio sp. DV isolated from the sediments of a Pb-Zn mine tailings dam in the Chita region, Russia. Genomics Data 11(C) (2017), 125–127.

Wang

, Li

, Jiao

Y.Y.

, et al. A relationship-based and object-oriented software for monitoring management during geotechnical excavation, Advances in Engineering Software 71(71) (2014), 34–45.

Gan-wu

ZHOU

, Neng-hui

, Ning

, et al. Automatic monitoring system for the safety of earth-rockfill dams based on the Internet of Things, Yantu Gongcheng Xuebao/Chinese Journal of Geotechnical Engineering 36(12) (2014), 2330–2334.

Zhang

, Wang

L.G.

and Yang

X.C.

, Research on tailings pond online safety monitoring system based on zigbee sensor network, Applied Mechanics and Materials 602–605 (2014), 1728–1732.

Yaya

, Tikou

and Cheng

L.Z.

, Numerical analysis and geophysical monitoring for stability assessment of the Northwest tailings dam at Westwood Mine, International Journal of Mining Science and Technology 27(4) (2017), 701–710.

Guangjin

, Xiangyun

, Liwei

, et al. Study on the flow characteristics of mud fluid in the process of the tailings dam break, Journal of Computational & Theoretical Nanoscience 13(4) (2016), 2623–2628.

Zhou

Shuren

, Ke

Maolin

and Luo

Peng

, Multi-camera transfer GAN for person re-identification, J. Vis. Commun. Image Represent 59 (2019), 393–400.

, Chen

, Wu

, et al. Effect of seepage control on stability of a tailings dam during its staged construction with a stepwise-coupled hydro-mechanical model, International Journal of Surface Mining Reclamation & Environment 29(2) (2015), 125–140.

Sun

, Zhang

and Li

, The internet of things (IoT) and cloud computing (CC) based tailings dam monitoring and pre-alarm system in mines, Safety Science 50(4) (2014), 811–815.

10.

Wei

, Yin

, Ling

, et al. A case study on a geotechnical investigation of drainage methods for heightening a tailings dam, Environmental Earth Sciences 75(2) (2016), 106.