Landslide early warning monitoring method and system based on NBIOT

Abstract

The invention relates to the technical field of geological disaster monitoring and detection, in particular to a landslide early warning and monitoring method and system based on NBIOT. Comprising the following steps: collecting data including rainfall data, soil humidity data and displacement data; carrying out data processing and analysis, carrying out integration and time synchronization processing on the collected data, and comparing the data with a preset alarm threshold value; when the data exceeds a preset alarm threshold value, a local alarm mechanism is triggered, and otherwise, the data is uploaded to the cloud; performing cloud data processing, including receiving and storing data, preprocessing the data, and establishing and training a BP neural network model; randomly initializing a weight value and a threshold value of the neural network; historical data is extracted as training sample input, a back propagation algorithm is adopted to train the BP neural network, the weight and the threshold are adjusted, and the network structure or parameters are adjusted to meet performance requirements; and performing landslide probability prediction by using the model. According to the technical scheme, the monitoring precision and timeliness can be improved, and the early warning accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring and detection, and in particular to a landslide early warning monitoring method and system based on NBIOT. Background Art

[0002] Landslides, a highly destructive geological disaster, occur frequently worldwide, particularly in areas of heavy rainfall. Their suddenness and powerful destructive power pose a serious threat to people's lives and property. Currently, landslides are ranked alongside volcanoes and earthquakes as one of the three major sources of geological hazards worldwide, and their hazard cannot be underestimated. Due to the widespread distribution of slopes, landslides are more prone to occur, are more common, and are more frequent than volcanoes and earthquakes.

[0003] Given the severity and complexity of landslide disasters, scientific and effective monitoring and early warning are particularly urgent and important. Traditional landslide monitoring and early warning methods have problems such as low monitoring accuracy, untimely data transmission, and low early warning accuracy, which are difficult to meet actual needs. Summary of the Invention

[0004] The purpose of the present invention is to propose a landslide early warning monitoring method and system based on NBIOT, which can improve the monitoring accuracy and timeliness and enhance the early warning accuracy.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a landslide early warning monitoring method based on NBIOT, comprising: Collect data, including rainfall data, soil moisture data and displacement data; Data processing and analysis: the collected data is integrated, time-synchronized, and compared with the preset alarm threshold. When the data exceeds the preset alarm threshold, the local alarm mechanism is triggered; otherwise, the data is uploaded to the cloud. Perform cloud data processing, including receiving and storing data, data preprocessing, building and training BP neural network models, including: According to the landslide early warning requirements, the BP neural network structure is designed, including input layer, hidden layer and output layer; the number of neurons in the input layer is at least 3; the mathematical model of the jth neuron is expressed as: Net input value:

[0006] Output value:

[0007] in, is the weight from the input layer to the hidden layer, is the threshold, is the activation function; Randomly initialize the weights and thresholds of the neural network; Extract historical data as training sample input, use back propagation algorithm to train BP neural network, adjust weights and thresholds, and the error function is as follows:

[0008] The global error is the sum of all sample errors:

[0009] By minimizing , the network gradually approaches the ideal output; Back propagation calculates the error gradient layer by layer through the chain rule and adjusts the weights and thresholds; The output layer weight adjustment formula is:

[0010]

[0011] Among them, η is the learning rate. To adjust the hidden layer weights, the error signal needs to be transmitted back to the previous layer and the weights need to be updated using the error derivative. Use test samples to evaluate the trained model and adjust the network structure or parameters to meet performance requirements; Use the model to predict landslide probability.

[0012] Beneficial effects of the basic scheme: This technical scheme simultaneously collects rainfall data, soil moisture data and displacement data, and monitors landslide-related factors from multiple dimensions. The fusion of multi-source data can more comprehensively and accurately reflect the possibility of landslide occurrence, avoiding misjudgment or omission that may be caused by a single data source.

[0013] This application integrates and time-synchronizes the collected data and compares it with pre-set alarm thresholds. When data exceeds the pre-set alarm threshold, it can immediately trigger a local alarm mechanism for a quick and timely response. Data that does not exceed the pre-set alarm threshold is uploaded to the cloud for further processing and prediction.

[0014] By extracting historical data as training sample input, the BP neural network is trained using a back-propagation algorithm. As time passes and data accumulates, the model can be continuously updated and optimized, making it better able to adapt to different geological conditions and environmental changes, and improving the accuracy of landslide warnings.

[0015] As an implementable and preferred solution, a model is used to predict the probability of landslides, including the following: Input the latest monitoring data into the trained BP neural network model to predict the probability of landslide occurrence at the current moment; Based on the prediction results, the risk level is divided into three levels: low, medium and high. The probability less than 10% is low risk, 10% - 50% is medium risk, and greater than 50% is high risk.

[0016] As an implementable preferred solution, rainfall data is collected through a tipping bucket rain sensor. Each time the tipping bucket rain sensor flips over, the microcontroller detects the level change through an interrupt and accumulates and counts the rainfall. The data is stored in a memory buffer. The buffer data is read and processed regularly to verify rationality. If the data is abnormal, it is corrected or marked as invalid. Soil moisture data is collected through a soil moisture sensor, which uses an ADC or directly reads the digital output to obtain soil moisture information. If analog output is used, the microcontroller activates the AD conversion module to convert the analog voltage signal output by the sensor into a digital signal. Based on the calibration parameters of the sensor, the digital signal is converted into the actual soil moisture value. If digital output is used, the microcontroller directly reads the level status of the DO pin; when the level is high, it means that the soil moisture is lower than the set threshold; when the level is low, it means that the soil moisture is higher than the set threshold; the acquired soil moisture data is stored in the memory of the microcontroller, and the data source and timestamp are marked; The displacement data acquisition MPU6050 sensor communicates with the microcontroller through the IIC interface. The microcontroller sends a data read instruction to the MPU6050 according to the set sampling frequency. After receiving the instruction, the MPU6050 sends the collected gyroscope and accelerometer data to the microcontroller through the IIC interface. After receiving the data, the microcontroller analyzes and processes the data; calculates the displacement change of the landslide area based on the calibration parameters and algorithm of the MPU6050; stores the displacement data in the memory of the microcontroller, and records the time information of the data acquisition.

[0017] As an implementable preferred solution, including system initialization, it includes the following contents: Initialize the power circuit, check the power supply mode and ensure the power supply voltage is stable; Initialize the reset circuit to ensure that the microcontroller can be reset correctly; Initialize the sensor module and configure the working parameters of each sensor; Initialize the NBIOT wireless communication module, configure communication parameters and check the network connection status; Initialize the microcontroller and configure internal registers, timers, and interrupts.

[0018] In a second embodiment, the present invention provides a landslide early warning monitoring system based on NBIOT, which utilizes the above-mentioned landslide early warning monitoring method based on NBIOT, including: The main control module uses the STC8H8K64U low-power microcontroller as the main control chip; The power supply module integrates a USB power supply interface, a lithium battery charging management module, and a solar charging interface, and supports automatic switching among three power supply modes; The sensor module includes a rainfall monitoring unit, a soil moisture monitoring unit, and a displacement monitoring unit, which are used to collect landslide factors and send them to the main control module; The communication module is based on NBIOT narrowband IoT technology and adopts the YNH-MN316 wireless communication module to realize data exchange between terminal devices and cloud platforms through a standardized communication protocol stack; The cloud processing platform uses the ONE-NET IoT cloud platform to receive and store data collected by sensors and provide an interface for data processing and analysis.

[0019] As an implementable preferred solution, the rainfall monitoring unit uses a tipping bucket rain sensor with a resolution of 0.2mm, which outputs rainfall information through a pulse signal; The soil moisture monitoring unit uses a sensor with digital switching output and analog output, and is equipped with a potentiometer to adjust the soil moisture threshold; The displacement monitoring unit uses the MPU6050 three-dimensional angular velocity sensor, which integrates a 3-axis MEMS gyroscope and a 3-axis MEMS accelerometer, as well as an expandable digital motion processor DMP, and outputs data through the IIC interface.

[0020] As an implementable preferred solution, the peripheral circuit includes a power supply circuit, a reset circuit and a debugging interface; The power circuit integrates USB power supply interface, lithium battery charging management module and solar charging interface; The reset circuit utilizes the built-in reset function of the chip. Externally, only a 10KΩ pull-up resistor needs to be connected to VCC to achieve power-on reset and manual reset. The debug interface reserves a USB to TTL download circuit and uses the CH340 chip, which supports program burning and debugging through PC software. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a hardware architecture diagram of the present invention.

[0022] Figure 2 This is a logic diagram of a landslide early warning monitoring method based on NBIOT.

[0023] Figure 3 This is a comparison chart between model prediction and actual results.

[0024] Figure 4FIG. 2 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

[0025] Reference numerals: electronic device 500 , processor 501 , communication interface 502 , memory 503 , bus 504 . DETAILED DESCRIPTION

[0026] In order to make the technical solution and advantages of the present application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It will be understood that the specific embodiments described herein are only partial embodiments of the present invention, which are only used to explain the present application, rather than to limit the present application. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered to be isolated, and they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the drawings of the following embodiments represent the same features or components, which can be applied to different embodiments.

[0027] In addition, unless otherwise defined, technical or scientific terms used in the description of the present invention should have the common meanings understood by those skilled in the art in the art to which the present invention belongs.

[0028] The present invention will be further described in detail below with reference to the accompanying drawings: Reference Figure 1 ,The hardware of this system includes the main control module, ,power supply module, sensor module, communication module, and cloud ,processing platform.

[0029] The main control module preferably uses the STC8H8K64U low-power microcontroller as the main control chip. The chip is based on the enhanced 8051 architecture, has a built-in high-precision clock source (frequency range 12-48MHz) and reset circuit, does not require an external crystal oscillator and reset components, has an operating voltage of 3.3-5.5V, and has a static power consumption as low as 1.8mA@5V.

[0030] The peripheral circuits include power supply circuit, reset circuit and debug interface.

[0031] The power circuit integrates a USB power supply interface (Type-C), a lithium battery charge management module (TP4056 chip), and a solar charging interface (input voltage 5-18V), supporting automatic switching among three power supply modes. A 22Ω impedance matching resistor is connected in series with the USB interface (D+ and D- pins) to ensure high-speed communication stability.

[0032] The reset circuit utilizes the built-in reset function of the chip. Externally, only a 10KΩ pull-up resistor needs to be connected to VCC to implement power-on reset and manual reset (triggered by a button).

[0033] The debug interface reserves a USB to TTL download circuit (CH340 chip), which supports program burning and debugging through PC software (AIapp-ISP-v6.94S).

[0034] The sensor module includes a rainfall monitoring unit, a soil moisture monitoring unit and a displacement monitoring unit, which can realize real-time capture and dynamic monitoring of environmental parameters and send the collected landslide factors to the main control module.

[0035] The rain gauge uses a tipping bucket-type rain gauge with a resolution of 0.2mm. It outputs rainfall information via a pulse signal. Each time the bucket flips, the voltage level at the IO port changes, indicating 0.2mm of rain has fallen. The sensor's positive and negative terminals are directly connected to the microcontroller's IO ports. An interrupt system monitors the IO port voltage changes and accumulates the rainfall amount.

[0036] The soil moisture sensor uses a sensor with both digital and analog outputs. The digital output (DO) can be directly connected to a microcontroller, which detects high and low levels to determine whether the soil moisture exceeds a threshold. The analog output (AO) can be connected to an analog-to-analog module for more accurate soil moisture readings through AD conversion. The sensor also features a potentiometer for adjusting the soil moisture threshold. Clockwise adjustment increases the moisture content, while counterclockwise adjustment decreases it.

[0037] The displacement sensor uses the MPU6050 3D angular velocity sensor, which integrates a 3-axis MEMS gyroscope and a 3-axis MEMS accelerometer, along with a scalable digital motion processor (DMP). The MPU-6050 uses three 16-bit ADCs for the gyroscope and accelerometer, respectively, to convert the analog measurements into digital outputs. These data are then output via an IIC interface for monitoring displacement changes in the landslide area.

[0038] The communication module utilizes the transmission channels of NBIOT narrowband IoT technology, enabling reliable data exchange between terminal devices and the cloud platform through a standardized communication protocol stack. The preferred NBIOT wireless communication module, the YNH-MN316, uses AT commands to communicate with the microcontroller via the serial port and transmit data to the ONE-NET cloud. NBIOT technology offers low power consumption and wide coverage, allowing monitoring equipment to operate on batteries for extended periods, meeting the communication needs of landslide monitoring systems in remote mountainous areas.

[0039] The power module includes USB power supply circuit, battery power supply circuit, solar charging circuit, etc. Through the coordinated power supply of USB, battery and solar energy, the device can ensure continuous and stable operation, so as to maintain continuous and reliable operation of the device.

[0040] The cloud processing platform uses the ONE-NET IoT cloud platform to receive and store data collected by sensors and provide an interface for data processing and analysis. The present invention proposes a landslide early warning monitoring system based on NBIOT technology, which can make up for the problems of insufficient communication coverage, high power consumption, unstable data transmission, etc. in traditional landslide monitoring systems. This system uses multiple sensors such as rainfall sensors, soil moisture sensors, displacement sensors, etc. to collaboratively monitor and collect key landslide parameters in real time, and relies on the low-power and wide-coverage network characteristics of NBIOT to achieve remote and reliable transmission of monitoring data. The system uses the STC8H8K64U low-power microcontroller to process data, and triggers an alarm when the parameter exceeds the threshold; if it does not exceed the threshold, the data is sent to the ONE-NET cloud through the NBIOT module, and finally the data is processed on the PC side, and the probability of future landslides is predicted. This system can effectively solve the problems of difficult deployment of monitoring equipment and high maintenance costs in remote mountainous areas, and provides a new technical path and intelligent solution for the application of Internet of Things technology in the field of disaster prevention and mitigation.

[0041] Reference Figure 2 The embodiment of the present invention discloses a landslide early warning monitoring method based on NBIOT, comprising: Step S100, system initialization, includes: In step S101, after the system is powered on, the power supply circuit is initialized. The USB power supply circuit is checked for proper connection. If so, 5V power is provided to the system via the USB interface. If not, the system switches to battery power. Simultaneously, the solar charging circuit is checked for proper operation to ensure that the solar panel can charge the battery. Voltage monitoring points in the power supply circuit are sampled to ensure that the power supply voltage is stable within the required system range (this system operates within the 5V voltage range).

[0042] Step S102: Initialize the reset circuit to ensure the MCU resets correctly upon power-up or an abnormality. Check the reset pin voltage to ensure it is within the normal operating range. If a system failure occurs or a restart is required, the reset circuit resets the MCU to its initial state.

[0043] Step S103 initializes the rainfall sensor, soil moisture sensor, displacement sensor, and other sensor modules. For the rainfall sensor, configure its IO port to interrupt mode, and set the interrupt trigger condition to a level change. For the soil moisture sensor, adjust the potentiometer to set the soil moisture threshold based on actual conditions, and initialize the A / D converter module (if precise values are required). For the displacement sensor MPU6050, perform initialization and configuration via the IIC interface, including setting parameters such as the sensor's operating mode, sampling frequency, and range.

[0044] Step S104: Initialize the NBIOT wireless communication module YNH-MN316. Send AT commands to the communication module through the serial port to configure communication parameters such as baud rate, data bits, stop bits, and parity. Check the communication module's signal strength and network connection status to ensure it can communicate properly with the ONE-NET cloud.

[0045] Step S105: Initialize the STC8H8K64U microcontroller, including configuring internal registers, timers, interrupts, etc. Initialize the USB download and debug circuit to ensure that the microcontroller can be downloaded and debugged through the USB interface.

[0046] Step S200, collecting data, includes: Step S201, collecting rainfall data, includes: When rainfall occurs, the IO level of the tipping bucket rain gauge changes with each flip of the bucket. The microcontroller detects this IO level change through the interrupt system and immediately enters the interrupt service routine. Within the interrupt service routine, the rainfall amount is accumulated. Each flip of the bucket represents 0.2 mm of rainfall, and the accumulated rainfall data is stored in the microcontroller's memory buffer.

[0047] Regularly read rainfall data from the memory buffer, process and verify the data, check the rationality of the data, and if the data exceeds the normal range (such as negative numbers or abnormally large values), correct the data or mark it as invalid.

[0048] Step S202, soil moisture data collection, includes: The soil moisture sensor acquires soil moisture information through an ADC (if using analog output) or directly reading the digital output (DO). If using analog output, the microcontroller activates the AD converter module to convert the sensor's analog voltage output into a digital signal. Based on the sensor's calibration parameters, the digital signal is converted into the actual soil moisture value.

[0049] If digital output is used, the microcontroller directly reads the level status of the DO pin. A high level indicates that the soil moisture is below the set threshold; a low level indicates that the soil moisture is above the set threshold. The acquired soil moisture data is stored in the microcontroller's memory, with the data source and timestamp.

[0050] Step S203, displacement data collection, includes: The MPU6050 sensor communicates with the microcontroller via the I2C interface. The microcontroller sends data read commands to the MPU6050 at a set sampling frequency. Upon receiving these commands, the MPU6050 sends the collected gyroscope and accelerometer data to the microcontroller via the I2C interface.

[0051] After receiving the data, the microcontroller analyzes and processes it. Based on the calibration parameters and algorithms of the MPU6050, it calculates the displacement change in the landslide area. The displacement data is stored in the microcontroller's memory, and the time of data acquisition is recorded.

[0052] Step S300, data processing and analysis, includes: In step S301, the main control module integrates the collected data such as rainfall, soil moisture, and displacement to form a comprehensive data set containing multiple parameters. The data is time-synchronized to ensure the consistency of the data of each parameter over time.

[0053] In step S302, each parameter in the comprehensive dataset is compared with a preset alarm threshold. When a parameter exceeds its corresponding threshold, a local alarm mechanism is triggered. The threshold settings in this embodiment are shown in Table 1. Local alarms can be implemented using an audible and visual alarm, which emits a warning sound and flashes lights to alert on-site personnel to the landslide hazard.

[0054] Table 1

[0055] If all parameters do not exceed the threshold, the data will be further processed and analyzed.

[0056] In step S303, the locally processed data is uploaded to the ONE-NET cloud via the NBIOT communication module. Prior to upload, the data is packaged and encrypted to ensure security and integrity. The format of the data packet can be defined according to the ONE-NET cloud interface requirements.

[0057] Step S400, performing cloud data processing and landslide probability prediction, includes: Step S401: Receive and store data After receiving the data uploaded from the NBIOT communication module, the ONE-NET cloud parses and verifies the data, checking its integrity and correctness. If there are any errors or data loss, feedback is sent to the monitoring system requesting the data to be re-uploaded.

[0058] The verified data is stored in a cloud-based database. This database can be a relational database (such as MySQL) or a non-relational database (such as MongoDB), depending on the data characteristics and application requirements. Corresponding data tables are created in the database to store different types of monitoring data, such as rainfall, soil moisture, and displacement.

[0059] Step S402 involves data preprocessing, extracting historical monitoring data from the cloud database, including parameters such as rainfall, soil moisture, and displacement. The data is cleaned to remove outliers and missing values. Outliers can be addressed using statistical methods (such as the 3σ principle) or business rule-based judgment methods; missing values can be addressed using methods such as interpolation and mean filling. The cleaned data is normalized to convert data of different dimensions into a unified range.

[0060] Step S403: Establish and train a BP (Back Propagation) neural network model. The causative factors of landslide disasters (such as rainfall, soil moisture content, and crack displacement) are highly nonlinearly correlated with the disaster trend. Based on a layered neuron topology, the BP neural network can fit complex data relationships through nonlinear activation functions, effectively extracting potential features from multi-source monitoring data and overcoming the model bias caused by the pre-set function form of traditional statistical models (such as linear regression). This algorithm, with its advantages in nonlinear mapping, adaptive parameter update mechanism, and rapid inference capabilities, is suitable for the data modeling requirements of landslide warning scenarios. Combined with the parallel computing characteristics of the embedded hardware platform, this algorithm can reduce edge resource consumption and enhance the timeliness of warning responses, including: Step S403-1: Design the structure of the BP neural network based on the needs of landslide early warning. The core of the BP neural network is to dynamically adjust the network weights and thresholds through the gradient descent method to achieve a nonlinear mapping relationship between input and output. It includes an input layer, a hidden layer, and an output layer. The input layer receives external signals. The number of neurons in the input layer is determined by the number of input parameters. In this embodiment, the input parameters include rainfall, soil moisture, crack displacement, etc., so the number of neurons in the input layer is at least 3. The hidden layer extracts features through weighted summation and nonlinear transformation. The number of neurons can be selected based on empirical formulas or experimental results and can generally be set to 5-10. The number of neurons in the output layer is 1, which is used to output the probability of landslide occurrence.

[0061] The mathematical model of the jth neuron can be expressed as: Net input value:

[0062] Output value:

[0063] in, is the weight from the input layer to the hidden layer, is the threshold, is the activation function.

[0064] Forward propagation passes the input signal layer by layer to the output layer, calculating the actual output value. Taking landslide warning as an example, the input parameters Xp (such as rainfall, soil moisture content, etc.) are processed by the hidden layer, and the output layer provides the predicted risk level Op.

[0065] Step S403-2: Randomly initialize the weights and thresholds of the neural network.

[0066] Step S403-3: extract historical data from the cloud database to obtain historical data for landslide prediction, and input the normalized data as training samples: X i For {X i1 , X i2 ,……,X in}, output {Y1, Y2, ..., Y n In the data, X1 represents rainfall, X2 represents soil moisture, and X3 represents crack displacement. The data is divided into training samples and testing samples. Training samples are used to train the neural network, while testing samples are used to evaluate the neural network's performance. The ratio of training samples to testing samples can be adjusted based on actual conditions.

[0067] Step S403-4: Train the BP neural network using the backpropagation algorithm. During training, the input data of the training sample is fed into the neural network, and the error between the output and the actual result is calculated. Based on the error signal, the weights and thresholds of the neural network are adjusted to gradually reduce the error.

[0068] The error function is as follows:

[0069] The global error is the sum of all sample errors:

[0070] By minimizing E, the network gradually approaches the ideal output.

[0071] Back propagation calculates the error gradient layer by layer through the chain rule and adjusts the weights and thresholds. The weight adjustment formula for the output layer is:

[0072]

[0073] Among them, η is the learning rate. To adjust the hidden layer weights, the error signal needs to be transmitted back to the previous layer and the weights need to be updated using the error derivative.

[0074] The training process can set the maximum number of iterations and error threshold. When the maximum number of iterations is reached or the error is less than the error threshold, the training stops.

[0075] The data is used as the input of BP neural network and the probability of landslide occurrence is used as the output. Figure 3 , after model training, the actual value and the predicted value are compared.

[0076] The relationship between the input layer, hidden layer and output layer is:

[0077] in, is the output of the neural network, is the weight between the hidden layer and the input layer, i is the number of input variables, and b is the correction parameter.

[0078] Traditional BP networks usually process a single type of data (such as images and text), while this scheme integrates three key landslide factors: rainfall, soil moisture, and displacement data as the input layer (at least three neurons), realizing weighted fusion of multi-source data and covering the "water-soil-displacement" chain mechanism of landslide formation.

[0079] By utilizing the multi-layer nonlinear mapping capability of the BP network, the complex coupling relationship between multiple factors (such as heavy rainfall → soil saturation → displacement acceleration) is captured, breaking through the linear assumption limitations of traditional statistical models (such as linear regression).

[0080] This technical solution can dynamically adjust the type, weight or threshold of input features according to the geological conditions of different regions (such as the difference in moisture threshold of clay / sand), rather than fixing the network structure.

[0081] Using long-term time series data stored in the cloud (such as multi-year rainfall-displacement sequences), the model adapts to long-term environmental changes such as climate change and vegetation cover changes through backpropagation algorithms and thresholds. A real-time feedback correction mechanism: When local monitoring data exceeds a threshold and triggers an alarm, the system automatically marks the sample as a "high-risk case" and incorporates it into the cloud training set, enabling the model to quickly learn sudden disaster patterns.

[0082] Through NBIoT technology, real-time feedback of multi-sensor data in remote mountainous areas can be achieved, solving the problem of insufficient model timeliness caused by data collection lag in traditional BP networks, and forming a closed loop of "edge collection-cloud modeling". This architecture is innovative in the combination of the Internet of Things and the field of geological disasters.

[0083] Traditional BP networks adjust model outputs through weights w, while this technical solution allows for dual adjustment through hardware thresholds (such as soil moisture potentiometers) and algorithm weights w, forming a hierarchical early warning mechanism of "front-end rapid response (threshold-triggered alarm) + back-end in-depth prediction (BP probability output)". When the data accumulation of newly built monitoring points is insufficient, the system can use similar geological condition data stored in the cloud (such as historical data of areas with similar slopes and soil types) for transfer learning, rather than relying on single-site data training, breaking the traditional BP network's dependence on large-scale labeled data.

[0084] Through the chain derivation of the back-propagation algorithm, the model can automatically mine the abnormal correlation patterns of multiple factors before the disaster (such as the time series characteristics of a sudden increase in rainfall accompanied by accelerated displacement), thereby improving the feature extraction capability.

[0085] Step S403-5: Evaluate the trained BP neural network model using test samples. Calculate metrics such as the model's prediction accuracy, recall, and F1 score to assess model performance. If the model performance does not meet expectations, adjust the network structure, weight initialization method, and training parameters, and retrain.

[0086] Step S500, performing landslide probability prediction, includes: The latest monitoring data obtained from the cloud is input into the trained BP neural network model to predict the probability of a landslide at the current moment. Based on the prediction results, the landslide risk level is determined. The risk level is categorized as low, medium, and high. For example, a landslide probability of less than 10% is considered low risk; a probability between 10% and 50% is considered medium risk; and a probability greater than 50% is considered high risk.

[0087] Those skilled in the art will understand that all or part of the processes in a landslide early warning and monitoring method based on NBIOT can be implemented by instructing related hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of various embodiments of a landslide early warning and monitoring method based on NBIOT. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0088] This embodiment of the present application also provides an electronic device 500 that utilizes the aforementioned NBIOT-based landslide early warning and monitoring method. The device includes a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the program, the steps of the aforementioned NBIOT-based landslide early warning and monitoring method are implemented. In this embodiment of the present application, the processor serves as the control center of the computer method and can be a processor of a physical machine or a processor of a virtual machine.

[0089] Reference Figure 4 The electronic device 500 includes: at least one processor 501, at least one communication interface 502, at least one memory 503, and at least one bus 504. Bus 504 is used to implement communication between these components, communication interface 502 is used to communicate signaling or data with other node devices, and memory 503 stores machine-readable instructions executable by processor 501. When the electronic device 500 is running, processor 501 communicates with memory 503 via bus 504. When the machine-readable instructions are called by processor 501, the steps of the above-mentioned NBIOT-based landslide early warning and monitoring method are executed.

[0090] The above contents are merely embodiments of the present invention. Common knowledge such as the known specific structures and characteristics in the scheme is not described in detail here. A person of ordinary skill in the art is aware of all common technical knowledge in the technical field to which the invention belongs before the filing date or priority date, is able to obtain all existing technologies in the field, and has the ability to apply conventional experimental means before that date. A person of ordinary skill in the art can, under the guidance of this application, improve and implement this scheme in combination with his or her own abilities. Some typical known structures or known methods should not become an obstacle for a person of ordinary skill in the art to implement this application. It should be pointed out that for a person of ordinary skill in the art, several variations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the scope of protection of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection claimed in this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A landslide early warning monitoring method based on NBIOT, characterized in that: include: Collect data, including rainfall data, soil moisture data and displacement data; Data processing and analysis: the collected data is integrated, time-synchronized, and compared with the preset alarm threshold. When the data exceeds the preset alarm threshold, the local alarm mechanism is triggered; otherwise, the data is uploaded to the cloud. Perform cloud data processing, including receiving and storing data, data preprocessing, building and training BP neural network models, including: According to the landslide early warning requirements, the BP neural network structure is designed, including input layer, hidden layer and output layer; the number of neurons in the input layer is at least 3; the mathematical model of the jth neuron is expressed as: Net input value: Output value: in, is the weight from the input layer to the hidden layer, is the threshold, is the activation function; Randomly initialize the weights and thresholds of the neural network; Extract historical data as training sample input, use back propagation algorithm to train BP neural network, adjust weights and thresholds, and the error function is as follows: The global error is the sum of all sample errors: By minimizing , the network gradually approaches the ideal output; Back propagation calculates the error gradient layer by layer through the chain rule and adjusts the weights and thresholds; The output layer weight adjustment formula is: Among them, η is the learning rate. To adjust the hidden layer weights, the error signal needs to be transmitted back to the previous layer and the weights need to be updated using the error derivative. Use test samples to evaluate the trained model and adjust the network structure or parameters to meet performance requirements; Use the model to predict landslide probability.

2. The landslide early warning monitoring method based on NBIOT according to claim 1 is characterized in that: Use the model to predict landslide probability, including the following: Input the latest monitoring data into the trained BP neural network model to predict the probability of landslide occurrence at the current moment; Based on the prediction results, the risk level is divided into three levels: low, medium and high. The probability less than 10% is low risk, 10%-50% is medium risk, and greater than 50% is high risk.

3. The landslide early warning monitoring method based on NBIOT according to claim 1 is characterized in that: Rainfall data is collected through a tipping bucket rain gauge. When the tipping bucket rain gauge flips over once, the microcontroller detects the level change through an interrupt and accumulates the rainfall amount. The data is stored in the memory buffer. The buffer data is read and processed regularly to verify rationality. If the data is abnormal, it is corrected or marked as invalid. Soil moisture data is collected through a soil moisture sensor, which uses an ADC or directly reads the digital output to obtain soil moisture information. If analog output is used, the microcontroller activates the AD conversion module to convert the analog voltage signal output by the sensor into a digital signal. Based on the calibration parameters of the sensor, the digital signal is converted into the actual soil moisture value. If digital output is used, the microcontroller directly reads the level status of the DO pin; when the level is high, it means that the soil moisture is lower than the set threshold; When the level is low, it means that the soil moisture is higher than the set threshold; The acquired soil moisture data is stored in the memory of the microcontroller and marked with the data source and timestamp; The displacement data acquisition MPU6050 sensor communicates with the microcontroller through the IIC interface. The microcontroller sends a data read instruction to the MPU6050 according to the set sampling frequency. After receiving the instruction, the MPU6050 sends the collected gyroscope and accelerometer data to the microcontroller through the IIC interface. After receiving the data, the microcontroller analyzes and processes the data; calculates the displacement change of the landslide area based on the calibration parameters and algorithm of the MPU6050; stores the displacement data in the memory of the microcontroller, and records the time information of the data acquisition.

4. The landslide early warning monitoring method based on NBIOT according to claim 1 is characterized in that: Includes system initialization, including the following: Initialize the power circuit, check the power supply mode and ensure the power supply voltage is stable; Initialize the reset circuit to ensure that the microcontroller can be reset correctly; Initialize the sensor module and configure the working parameters of each sensor; Initialize the NBIOT wireless communication module, configure communication parameters and check the network connection status; Initialize the microcontroller and configure internal registers, timers, and interrupts.

5. A landslide early warning monitoring system based on NBIOT, characterized by: A landslide early warning monitoring method based on NBIOT as claimed in any one of claims 1 to 4 is used, comprising: The main control module uses the STC8H8K64U low-power microcontroller as the main control chip; The power supply module integrates a USB power supply interface, a lithium battery charging management module, and a solar charging interface, and supports automatic switching among three power supply modes; The sensor module includes a rainfall monitoring unit, a soil moisture monitoring unit, and a displacement monitoring unit, which are used to collect landslide factors and send them to the main control module; The communication module is based on NBIOT narrowband IoT technology and adopts the YNH-MN316 wireless communication module to realize data exchange between terminal devices and cloud platforms through a standardized communication protocol stack; The cloud processing platform uses the ONE-NET IoT cloud platform to receive and store data collected by sensors and provide an interface for data processing and analysis.

6. The NBIOT-based landslide early warning monitoring system according to claim 5, characterized in that: The rainfall monitoring unit uses a tipping bucket rain sensor with a resolution of 0.2mm, which outputs rainfall information through pulse signals; The soil moisture monitoring unit uses a sensor with digital switching output and analog output, and is equipped with a potentiometer to adjust the soil moisture threshold; The displacement monitoring unit uses the MPU6050 three-dimensional angular velocity sensor, which integrates a 3-axis MEMS gyroscope and a 3-axis MEMS accelerometer, as well as an expandable digital motion processor DMP, and outputs data through the IIC interface.

7. The NBIOT-based landslide early warning monitoring system according to claim 5, characterized in that: The peripheral circuit includes power supply circuit, reset circuit and debug interface; The power circuit integrates USB power supply interface, lithium battery charging management module and solar charging interface; The reset circuit utilizes the built-in reset function of the chip. Externally, only a 10KΩ pull-up resistor needs to be connected to VCC to achieve power-on reset and manual reset. The debug interface reserves a USB to TTL download circuit and uses the CH340 chip, which supports program burning and debugging through PC software.

Images