Slope radar data vpn private network transmission and remote early warning system

Abstract

The present application relates to the technical field of slope monitoring and early warning, in particular to a slope radar data VPN private network transmission and remote early warning system, which comprises a slope radar monitoring module, an intelligent data processing module, a VPN transmission module based on value priority and a B / S cloud early warning platform. The intelligent data processing module realizes multi-source heterogeneous data alignment and unified analysis through dynamic protocol feature self-learning and cross-protocol conflict resolution technology; meanwhile, a data value density model is constructed to integrate multi-dimensional factor quantitative data values such as deformation rate and geological conditions. Based on value density, the data is classified, and the VPN transmission module matches the link quality level to realize differentiated transmission, and the B / S cloud early warning platform completes hierarchical storage and multi-dimensional early warning. The system solves the problems of large compatibility difficulty of multi-source data analysis and high-value data transmission lag, and guarantees efficient disposal of slope risk data.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring and early warning technology, and more specifically, to a slope radar data VPN private network transmission and remote early warning system. Background Technology

[0002] Slope monitoring is a core technology in geological disaster prevention and control. By collecting data on slope deformation and the environment, it provides support for disaster early warning and is of great significance for ensuring engineering safety and the safety of people's lives and property. With the development of monitoring technology, slope radar equipment from different manufacturers is widely used, but inconsistent equipment protocols and data format differences have become common problems in the industry.

[0003] In existing technologies, the parsing of multi-source heterogeneous data relies on manually written adaptation rules. When facing unknown protocols or protocol iterations, repeated development is required, resulting in low adaptation efficiency and poor compatibility. Simultaneously, data transmission employs a one-size-fits-all approach, failing to prioritize data value. High-risk warnings, critical deformation data, and other core data share transmission links with regular scanning data, making them susceptible to transmission delays due to link congestion or poor quality, impacting the timeliness of warning responses. Furthermore, existing systems lack quantitative assessment of data value, hindering precise allocation of transmission resources and further reducing the overall reliability of slope monitoring.

[0004] To address the aforementioned shortcomings, this invention proposes a VPN private network transmission and remote early warning system for slope radar data, which solves the problems of multi-source data parsing compatibility and data transmission priority allocation through innovative technical solutions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a VPN private network transmission and remote early warning system for slope radar data.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The slope radar data VPN private network transmission and remote early warning system includes a slope radar monitoring module, an intelligent data processing module, a value-first VPN transmission module, and a B / S cloud early warning platform.

[0008] The slope radar monitoring module collects physical state data and environmental parameters of the slope surface, providing a raw data source for data value quantification.

[0009] The intelligent data processing module takes data value density calculation as its core and integrates a multi-source heterogeneous monitoring protocol adaptive parsing engine and a real-time data value density perception model: The multi-source heterogeneous monitoring protocol adaptive parsing engine realizes adaptive parsing of multi-vendor device protocols through dynamic protocol feature self-learning and cross-protocol conflict resolution technology, and outputs standardized data.

[0010] The real-time data value density perception model integrates deformation rate, geological vulnerability coefficient, historical risk weight, and real-time environmental factors, and combines dynamic weight coefficient and data timeliness decay coefficient to generate quantified data value densities.

[0011] The value-priority-based VPN transmission module first uses tools to evaluate link latency, packet loss rate, and bandwidth in real time and classify link quality levels. Then, based on the data value density output by the intelligent data processing module, it classifies the data into emergency alarms, key deformation trends, and routine scan data, and establishes a matching mechanism between data value levels and link quality levels, allowing high-value data to occupy high-quality VPN links for transmission.

[0012] The B / S cloud early warning platform implements hierarchical storage based on data value density, and combines AI trend prediction results with data value density to trigger multi-dimensional hierarchical early warnings, ensuring priority handling of high-value risk data.

[0013] Furthermore, the multi-source heterogeneous monitoring protocol adaptive parsing engine includes a dynamic protocol feature self-learning mechanism and cross-protocol conflict resolution technology;

[0014] The aforementioned dynamic protocol feature self-learning mechanism achieves adaptive adaptation to unknown protocols by capturing device data frames, extracting patterns of fixed identifiers and variable fields, classifying and semantically inferring variable fields, and updating and verifying the feature library.

[0015] The cross-protocol conflict resolution technology resolves protocol conflicts by constructing a field semantic ontology library, automatically identifying and converting data formats while preserving precision, adapting verification rules, and verifying data integrity.

[0016] Furthermore, the dynamic parsing execution process of the adaptive parsing engine for the multi-source heterogeneous monitoring protocol is as follows:

[0017] Step 1: After receiving binary data, extract the data start segment and match it with the preset protocol header feature library. If the match fails, trigger the dynamic protocol feature self-learning mechanism to analyze the protocol field structure.

[0018] Step 2: Perform field segmentation on the data frame based on the information entropy algorithm. Calculate the information entropy of each byte position, combine the preset sliding window parameters to traverse the data frame to generate a continuous entropy value sequence, and simultaneously calculate the entropy value mutation amplitude and mutation rate.

[0019] Step 3: Set a preset threshold for entropy value mutation amplitude and mutation rate. When a certain byte position meets the following conditions: the attenuation ratio of the entropy value compared with the previous byte position meets the standard and the entropy value mutation amplitude and entropy value mutation rate both reach the corresponding preset thresholds, the byte position is determined to be a field boundary, and the field segmentation of the data frame is completed.

[0020] Step 4: For each independent field after segmentation, associate the field mapping relationship with the system standard field based on the field mapping relationship obtained by the dynamic protocol feature self-learning mechanism, unify the data format according to the preset data format conversion rules in the cross-protocol conflict resolution technology, and resolve field naming conflicts through the field semantic ontology library.

[0021] Furthermore, the magnitude and rate of entropy value mutation are as follows:

[0022] Traverse the entire data frame to obtain a continuous sequence of entropy values. , m is the length of the data frame in bytes; define the magnitude of entropy mutation:

[0023] ;

[0024] Where k is the byte position index, 2≤k≤m;

[0025] Define the entropy mutation rate:

[0026] ;

[0027] in, This is the compensation value in the denominator.

[0028] Furthermore, the aforementioned real-time data value density perception model is as follows:

[0029] ;

[0030] in, This represents the normalized deformation rate, with a value range of [0,1]. This represents the geological vulnerability coefficient, with a value range of [0,1]. This represents the historical risk weight, with a value range of [0,1]. This represents a real-time environmental factor, with a value range of [0,1]. Indicates the dynamic weighting coefficient; This is the data timeliness decay coefficient.

[0031] Furthermore, the aforementioned real-time environmental factors are obtained by comprehensively considering rainfall and vibration intensity, as detailed below:

[0032] A linear weighted fusion method is used to construct the calculation formula. Among them, the rainfall factor R1 is assigned a value based on the real-time rainfall range; the vibration intensity factor R2 is assigned a value based on the real-time vibration intensity range. and The influence weights are derived from historical data analysis.

[0033] Furthermore, the dynamic weight coefficients are dynamically adjusted using the random forest algorithm in conjunction with the real-time environment, as detailed below:

[0034] Constructing a training dataset: Collect landslide event data from the past 5 years, add environmental and deformation correlation data, and form a four-dimensional training dataset including deformation rate, geological vulnerability, historical risk, and environmental factors;

[0035] Random forest model training: The number of trees, maximum tree depth, and feature selection method were set, and the model was trained through 50 rounds of iterations to ensure that the accuracy of the model on the validation set met expectations.

[0036] Real-time environment-driven adjustment: The weight coefficients output by the random forest algorithm are optimized a second time according to the preset adjustment rules to adapt to the weight scenarios under extreme environments or sudden scenarios. The adjustment rules are preset based on expert experience.

[0037] Furthermore, the deep binding mechanism between data type, data value density, and link quality established by the value-priority VPN transmission module is as follows:

[0038] Emergency alarm data: Business data scenarios include slope landslide warnings and monitoring equipment failures. The data value density V≥0.7, and the transmission must be completed within 1 second. The VPN link selection priority is to use the VPN link with the best link quality.

[0039] Key deformation trend data: Business data scenarios include slope deformation rate data in geologically vulnerable zones and slope monitoring data after rainfall. The data value density is 0.3≤V<0.7. Transmission is required to be completed within 5 seconds. The VPN link selection priority is to first use VPN links with good link quality, and secondly use VPN links with medium link quality.

[0040] Regular scanning data: Business data scenarios include slope point cloud data in large stable areas and periodic slope inspection data. The data value density V < 0.3, and the transmission time can be relaxed to within 30 seconds. The VPN link selection priority is to first use the VPN link with medium link quality, and secondly use the VPN link with poor link quality.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The multi-source heterogeneous monitoring protocol adaptive parsing engine proposed in this invention relies on a dynamic protocol feature self-learning mechanism to capture data frames, extract field patterns, infer semantics, and update the feature library. Combined with cross-protocol conflict resolution technology, it constructs a semantic ontology library, preserves accuracy while performing format conversion, and adapts validation rules. Simultaneously, it introduces an information entropy algorithm to calculate byte entropy values ​​and mutation parameters to achieve accurate field segmentation. It can adapt to unknown protocols and iterative versions without manual intervention, avoiding redundant development and significantly reducing the difficulty of integrating devices from multiple vendors. At the same time, it outputs standardized data, eliminating format and semantic differences, providing a reliable data source for subsequent data value quantification, and ensuring the accuracy of the entire data processing process.

[0043] 2. To address the issue of delays in high-value data caused by the traditional "one-size-fits-all" transmission approach, the system employs a dual-innovation optimization. Firstly, it constructs a data value density model, integrating normalized deformation rate, geological vulnerability coefficient, historical risk weights, and real-time environmental factors, combined with random forest dynamic weights and time-sensitivity decay coefficients to accurately quantify data value. Secondly, based on a value-priority VPN transmission module, it real-time assesses and classifies link quality (latency, packet loss rate, bandwidth), establishing a matching mechanism between data classification and links. This ensures that high-value data is transmitted promptly through high-quality links, avoiding delayed warnings, while allowing regular data to reuse secondary links, preventing resource waste and achieving precise allocation of transmission resources. Attached Figure Description

[0044] Figure 1 A block diagram of a VPN private network transmission and remote early warning system for slope radar data;

[0045] Figure 2 This is a flowchart of the dynamic protocol feature self-learning mechanism of the present invention;

[0046] Figure 3 This is a flowchart illustrating the implementation of the cross-protocol conflict resolution technology of the present invention. Detailed Implementation

[0047] Example, refer to Figure 1 The slope radar data VPN private network transmission and remote early warning system of this embodiment includes a slope radar monitoring module, an intelligent data processing module, a value-first VPN transmission module, and a B / S cloud early warning platform.

[0048] Slope radar monitoring module: As the sensing terminal of the system, it collects physical state data and environmental parameters of the slope surface, providing the original data source for subsequent intelligent processing, transmission and scheduling;

[0049] This module is equipped with a 24GHz FMCW slope radar, achieving a ranging accuracy of ±0.1mm, a 360° scanning angle, and a data acquisition frequency of 1Hz, enabling stable operation in all weather conditions. Simultaneously, the module integrates a rainfall sensor and a vibration sensor for synchronous environmental parameter acquisition. Data output uses a vendor-specific protocol for encapsulation, supporting both binary stream and custom JSON formats, transmitted via UDP port 5000, with each data entry kept under 1KB, providing multi-source, multi-format raw data support for subsequent module processing.

[0050] In terms of system collaboration, the module synchronously collects multi-dimensional data such as point coordinates (X / Y / Z), deformation rate, rainfall, and vibration intensity every second, and encapsulates them into a unified data frame before sending them to the intelligent data processing module. In addition, the module has flexible protocol adaptation capabilities and can dynamically adjust the data output format (e.g., from binary stream to JSON) according to the instructions of the intelligent data processing module to ensure compatibility with the backend parsing logic.

[0051] Compared to the limitations of existing technologies that rely solely on radar data collection, this module provides environmental dimension support for data value quantification by adding environmental parameters to synchronously collect rainfall and vibration data, making subsequent value judgments more aligned with real-world scenarios. It also supports dynamic adjustment of data output formats, adapting to the needs of multiple vendor protocol parsing, significantly reducing the workload and difficulty of protocol adaptation for backend modules.

[0052] Intelligent data processing module: Integrates two core engines - multi-source heterogeneous monitoring protocol adaptive parsing engine and data value density real-time perception model, to realize raw data parsing, value quantification and hierarchical labeling processing, and solve the core pain points of existing technologies such as difficulty in integrating multi-source devices and the disconnect between data value and transmission;

[0053] Multi-source heterogeneous monitoring protocol adaptive parsing engine: Introduces dynamic protocol feature self-learning and cross-protocol conflict resolution mechanism to achieve adaptive adaptation to unknown manufacturer equipment and protocol version iterations, without the need for manual parsing rule writing;

[0054] S21. Dynamic protocol feature self-learning mechanism:

[0055] The specific steps of dynamic protocol feature self-learning are as follows: Figure 2 As shown:

[0056] a. Device Data Frame Capture: When the system connects to an unknown protocol device, the data capture process is automatically triggered; three complete communication data frames of the device are captured continuously. The length of each frame is dynamically determined according to the device type, typically ranging from 128 to 1024 bytes, and is stored in a temporary buffer, such as a Redis list.

[0057] b. Extraction of Fixed Identifiers and Variable Field Patterns: A 16-byte sliding window algorithm is used to scan the captured data frame byte by byte. The window starts from the beginning of the data frame and slides 1 byte at a time to extract the byte sequence features within the window, such as the repetition patterns of ASCII characters and hexadecimal values. By statistically analyzing the frequency of different windows, the fixed protocol header and checksum fields of the data frame are identified, while the variable field areas whose length and content change with the communication cycle are marked.

[0058] c. Variable Field Classification and Semantic Inference: For variable field regions, the K-Means unsupervised clustering algorithm is used for classification, with the number of clusters K set to 5~10, dynamically adjusted according to the number of fields. After clustering, the meaning of the fields is inferred based on the numerical distribution characteristics, including but not limited to the following fields:

[0059] If the value of a certain field is stable in the range of 0~10mm / h and shows a continuous changing trend over time, it is determined to be a deformation rate field.

[0060] If the value is an integer and the range is 0~360°, it is determined to be the scanning angle field;

[0061] If it is a string and contains keywords such as "error" or "warning", it is determined to be a device status field.

[0062] d. Feature database update and review: Generate a review request based on the inferred field mapping relationship and push it to the system administrator interface; after the administrator approves it, write the mapping relationship into the MySQL feature database;

[0063] S22. Cross-protocol conflict resolution techniques: such as... Figure 3 As shown, the following steps are used to resolve conflicts such as field naming conflicts and data format differences between different vendor protocols.

[0064] a. Construction of field semantic ontology library and synonym mapping:

[0065] A semantic ontology library containing professional terms in the field of slope monitoring was constructed. It was described using the OWL ontology language and stored in the Neo4j graph database. Each term includes attributes such as synonyms, data type, and value range. For synonymous fields from different manufacturers, semantic similarity was calculated using the cosine similarity algorithm: the field names were converted into word vectors, and a Word2Vec pre-trained model with a vector dimension of 300 was used. If the similarity was ≥0.8, it was determined to be semantically consistent and uniformly mapped to the system's standard fields.

[0066] b. Automatic data format recognition and precision-preserving conversion:

[0067] To address the various data formats that may arise from multiple sources, a method combining regular expressions and bitwise operations is used to achieve automatic identification, and then a precision preservation algorithm is used to uniformly convert them into the system standard format (float64).

[0068] For int format, if the field length is 2 bytes and the value range is -32768 to 32767, it is determined as int16; if the field length is 4 bytes and the value range is -2147483648 to 2147483647, it is determined as int32. During conversion, int16 is directly converted to float64, while int32 is converted to int16 or int8. Range verification is required first. If it exceeds the target format range, it is scaled proportionally, and the scaling factor is recorded to ensure data traceability and ensure that the precision loss is ≤0.5%.

[0069] For floating-point formats, a byte length of 4 and conforming to the IEEE 754 single-precision encoding rules are classified as float32, and a byte length of 8 and conforming to the double-precision encoding rules are classified as float64. Float32 to float64 conversion directly preserves the original precision, while float64 to float32 conversion uses a truncation conversion, resulting in a precision loss of ≤0.0001%.

[0070] For non-numeric formats, if the field content is printable ASCII characters and has no fixed numerical range, it is determined to be a string format, and the original content is directly preserved and encapsulated into a system string type; if the field value is a binary discrete value, such as 0 / 1, true / false, it is determined to be a Boolean type and uniformly mapped to the system standard Boolean value.

[0071] c. Verification rule adaptation and data integrity verification:

[0072] Extract verification rules from mainstream vendor protocols and build a verification rule adaptation pool; for received data frames, automatically match the corresponding verification rules and execute the verification algorithm to verify data integrity; if the verification fails, trigger the retransmission mechanism; if the retransmission still fails, mark the data frame as invalid.

[0073] S23. Dynamic parsing execution process: The complete parsing process for unknown protocol data is coherent and reproducible, ensuring that the parsing accuracy and adaptation time meet preset standards. In this embodiment, the preset parsing accuracy is ≥98% and the adaptation time is ≤3min.

[0074] First, after receiving binary data, the first 16 bytes are extracted and matched with the protocol header features in the MySQL feature library. If the match fails, the dynamic feature self-learning mechanism is triggered to analyze the protocol field structure.

[0075] Then, the field is segmented based on the information entropy algorithm, and the information entropy of each byte position i is calculated, where 1 ≤ i ≤ m, as shown in the following formula:

[0076] ;

[0077] Where k1 represents the index of the possible values ​​at the byte position, and n1 represents the number of all possible values ​​at the byte position. The probability of byte position i taking the k1-th value is obtained by statistical calculation of the captured data frame; when the entropy value changes abruptly, it is determined as a field boundary, thereby dividing the data frame into multiple independent fields;

[0078] Abrupt changes in entropy are determined by the magnitude and rate of these changes. A 16-byte sliding window with a 1-byte step size is used to traverse the entire data frame, resulting in a continuous sequence of entropy values. (m is the length of the data frame in bytes); Define the magnitude of entropy mutation:

[0079] ;

[0080] Where k is the byte position index, 2≤k≤m;

[0081] Define the entropy mutation rate:

[0082] ;

[0083] in, This is a denominator compensation value to control the occurrence of extreme values; since the beginning and end of data frames are mostly fixed protocol headers, checksums, or end characters, therefore... The field boundary identification range is defined as the middle segment of the data frame; among them, 1.5 is the optimal threshold verified by statistics - it is higher than the mutation rate threshold of 1.0 for field boundary determination to ensure that no effective boundary is missed, and lower than the average mutation rate in extreme noise scenarios (about 2.0) to avoid misjudgment;

[0084] Setting a quantitative threshold: Based on the entropy distribution characteristics of slope monitoring data, and after calibration with over 1000 samples, the threshold for entropy value mutation amplitude is determined. Mutation rate threshold In practical applications, if the false positive rate is found to be too high, the threshold can be flexibly adjusted.

[0085] When satisfied and When this happens, the byte position k2 is determined to be the field boundary;

[0086] After segmentation, semantic mapping, format conversion, and conflict resolution operations are performed on each field—the mapping relationship obtained through self-learning is associated with the system's standard fields, the data format is unified according to the format conversion rules, and naming conflicts are resolved through the semantic ontology library;

[0087] Finally, all parsed fields are encapsulated into a system-wide JSON key-value pair format and sent to the radar data server of the intelligent data processing module via a RESTful API interface.

[0088] Real-time data value density perception model: The original formula is derived from a mathematical linear weighted model. Combined with the characteristics of slope monitoring operations, environmental factors and time-related decay coefficients are introduced. The formula is as follows:

[0089] ;

[0090] in, This represents the normalized deformation rate, with a value range of [0,1]. Represents the geological vulnerability coefficient, based on the RQD index mapping, with a value range of [0,1]. This represents the historical risk weight, which is mapped based on the frequency of landslides in the past 3 years, and its value ranges from [0,1]. This represents real-time environmental factors, taking into account both rainfall and vibration intensity, with a value range of [0,1].

[0091] The rainfall is divided into intervals, and the rainfall factor R1 is assigned a value: when the rainfall is ≥50mm / h, R1=0.9; when the rainfall is 30≤rainfall<50mm / h, R1=0.7; when the rainfall is 10≤rainfall<30mm / h, R1=0.5; when the rainfall is 0≤rainfall<10mm / h, R1=0.1.

[0092] Based on vibration intensity (unit: g, 1g = 9.8 m / s²) 2 Divide the intervals and assign values ​​to the vibration intensity factor R2: when the vibration intensity is ≥ 0.5g, R2 = 0.8; when 0.2 ≤ vibration intensity < 0.5g, R2 = 0.5; when 0 ≤ vibration intensity < 0.2g, R2 = 0.1.

[0093] A linear weighted fusion method was adopted, combining the influence weights of rainfall and vibration intensity on slope deformation. The influence weights were derived from historical landslide event data analysis, with rainfall as the primary factor. Vibration intensity weight The specific formula is as follows:

[0094] ;

[0095] Calculation example:

[0096] Heavy rainfall scenario: rainfall 60mm / h, R1=0.9, vibration intensity 0.1g, R2=0.1, then:

[0097] E = 0.6 × 0.9 + 0.4 × 0.1 = 0.58;

[0098] Strong vibration scenario: rainfall 5mm / h, R1=0.1, vibration intensity 0.6g, R2=0.8, then:

[0099] E = 0.6 × 0.1 + 0.4 × 0.8 = 0.38;

[0100] Heavy rain intensifies vibration scenario: rainfall 55mm / h, R1=0.9, vibration intensity 0.6g, R2=0.8, then:

[0101] E = 0.6 × 0.9 + 0.4 × 0.8 = 0.86;

[0102] Represents dynamic weighting coefficients, satisfying It uses the random forest algorithm combined with real-time environmental dynamic adjustments; The data timeliness attenuation coefficient ranges from [0.1, 1] and is calculated using the following formula:

[0103] ;

[0104] in, The time difference between data acquisition and calculation is given by k3=0.046, which is the time decay coefficient. hour, Reverse reasoning:

[0105] ;

[0106] Dimensionality Consistency: All variables are dimensionless coefficients, and the dynamic weighting coefficient and attenuation coefficient are also dimensionless;

[0107] Dynamic weight training and adjustment mechanism:

[0108] Training dataset: Collected over 2000 sets of landslide event data from a provincial geological disaster database over the past 5 years, and added over 1000 sets of related environmental and deformation data to construct a four-dimensional dataset including deformation rate, geology, historical risk, and environmental factors;

[0109] Random forest training parameters: 100 trees, maximum tree depth 12, Gini coefficient as the feature selection method, 50 training iterations, achieving a validation set accuracy of ≥93%;

[0110] Example of dynamic adjustment rules: Based on the weight coefficients output by the random forest algorithm, the weight coefficients are optimized a second time to meet the scene adaptation of the weights in extreme environments or sudden scenarios. The adjustment rules are preset according to expert experience. When the rainfall is ≥50mm / h, Adjusted from 0.1 to 0.3. The value judgment was adjusted from 0.55 to 0.35, indicating that environmental factors now dominate the value judgment.

[0111] When the geological vulnerability G ≥ 0.8, Adjusted from 0.3 to 0.4. The value was adjusted from 0.1 to 0.05, with geological factors now dominating the value judgment.

[0112] Data timeliness When ≥30s, =0.5, the overall data value density is reduced by 50%, avoiding outdated data from occupying high-quality transmission resources;

[0113] Example of data value density calculation:

[0114] High-value data examples (dominated by environmental factors):

[0115] Deformation rate After normalization Given G=0.7, HI=0.6, real-time rainfall of 60 mm / h, vibration intensity of 0.1 g, and corresponding E=0.58, ,correspond ;

[0116] The dynamic weights are adjusted as follows: ;but:

[0117] V=(0.35×0.4+0.3×0.7+0.15×0.6+0.2×0.58)×1=0.556;

[0118] Low-value data instances (time-sensitive):

[0119] Deformation rate After normalization G=0.3, HI=0.2, E=0.1 ,correspond The weight is ;but:

[0120] V=(0.55×0.1+0.3×0.3+0.15×0.2+0.0×0.1)×0.135=0.0236;

[0121] Value-based VPN transmission module: This module uses data value density V as its core criterion, introducing real-time link quality assessment and a deep binding mechanism between data type and value level. The core technical solutions are as follows:

[0122] S31. Real-time link quality assessment:

[0123] Latency and packet loss rate: ICMP probe packets are sent every 100ms using the Ping tool, and the average latency (unit: ms) and packet loss rate (dimensionless, range [0,1]) are calculated.

[0124] Bandwidth: Real-time measurement of the actual available bandwidth of the channel using the iPerf tool (unit: Mbps);

[0125] Quality grading: Based on the indicators output by the tool, the link quality is divided into four levels: excellent, good, medium, and poor. The grading thresholds are dynamically set according to the radar data transmission requirements in the actual slope scenario. The specific grading thresholds in this embodiment are as follows:

[0126] Advantages: Latency ≤30ms, packet loss rate ≤0.01%, bandwidth ≥80% of the channel's maximum bandwidth;

[0127] Good: 30ms < latency ≤ 50ms, 0.01 < packet loss rate ≤ 0.05, 60% ≤ bandwidth < 80%;

[0128] Medium: 50ms < latency ≤ 100ms, 0.05 < packet loss rate ≤ 0.1, 40% ≤ bandwidth < 60%;

[0129] Poor: Latency > 100ms, packet loss rate > 0.1%, bandwidth < 40%;

[0130] When classifying link quality, the "barrel effect" is followed, and the lowest level among all indicators is used as the overall quality level of the link.

[0131] S32. Deep binding of data type, value density, and link quality:

[0132] Based on the business characteristics and data value density V of the three core data types, the priority of VPN link selection is clarified:

[0133] Emergency alarms include, but are not limited to, landslide warnings (deformation rate ≥ 2 mm / h and geological vulnerability coefficient G > 0.6), equipment failures (radar power failure, communication interruption), and other data that needs to be transmitted within 1 second, with a data value density V ≥ 0.7; VPN link selection priority is top-priority.

[0134] Key deformation trend data: This includes, but is not limited to, deformation rate data of geologically vulnerable zones (G≥0.6), monitoring data after rainfall (rainfall ≥50mm / day), and other data that needs to be transmitted within 5 seconds, with a data value density of 0.3≤V<0.7; VPN link selection priority is first-good, second-medium.

[0135] Regular scan data includes, but is not limited to, point cloud data with large stable areas (G≤0.3) and historical risk weights (HI≤0.2), periodic inspection data, etc., with transmission time relaxed to 30 seconds, and data value density V<0.3; VPN link selection priority is first-medium, second-lower priority;

[0136] The thresholds for the above data range division were determined by combining historical data with expert experience;

[0137] After determining a suitable VPN tunnel, data from different local area networks in each mine is transmitted to the B / S cloud early warning platform through the established VPN private network via the domain controller.

[0138] B / S Cloud Early Warning Platform:

[0139] S41, Data Receiving and Hierarchical Storage Unit:

[0140] Receiving mechanism: It adopts a high-concurrency RESTful API interface cluster, supports the parallel reception of 1000+ JSON format data per second, and is adapted to the output protocol of the intelligent data processing module;

[0141] Tiered storage strategy: Strictly correlate data value density V to achieve differentiated storage - high-value data (V≥0.7) is stored in SSD arrays with a retention period of 1 year; critical data (0.3≤V<0.7) is stored in a hybrid storage pool (SSD+HDD) with a retention period of 6 months; and regular data (V<0.3) is stored in HDD distributed storage with a retention period of 3 months.

[0142] Database selection: A combined architecture of time-series database, relational database, and graph database is adopted—InfluxDB stores time-series monitoring data such as deformation rate and rainfall; MySQL stores structured data such as early warning records, equipment configuration, and user permissions; Neo4j associates semantic data such as geological vulnerability and historical risk.

[0143] S42, Lightweight AI Trend Prediction Unit:

[0144] S421. Algorithm Architecture: A fusion model of LSTM neural network and random forest is adopted. LSTM captures the time series trend of deformation rate, while random forest integrates multi-dimensional static features such as geology, environment, and historical risk. The specific model structure is as follows:

[0145] LSTM neural network submodule:

[0146] Network layer structure: Input layer (processes deformation rate time series, dimension is the time window length, default 60) → LSTM hidden layer (containing 128 memory units, capturing time series trends) → Output layer (linear activation, outputs one-dimensional time series trend score);

[0147] Input: Deformation rate time series (a one-dimensional sequence of length 60, representing the deformation rate at 60 consecutive time points).

[0148] Output: Time series trend score (0-1 continuous value, quantifying the significance of the upward trend of the deformation rate time series).

[0149] Random Forest submodule:

[0150] Network layer structure: Feature input layer (receives 5 types of static features) → Multi-decision tree layer (composed of 100 decision trees, each tree independently splits and fits the features) → Probability output layer (integrates the results of all decision trees and outputs a single-dimensional static risk probability).

[0151] Inputs: Geological vulnerability coefficient G, historical risk weight HI, real-time environmental factor E, data value density V, link quality level (5 types of static features, 5 dimensions);

[0152] Output: Static risk probability (a continuous value of 0-1, quantifying the degree of deformation risk under a combination of static features).

[0153] Fusion layer:

[0154] Network layer structure: Fully connected layer (containing 32 neurons, using ReLU activation function to achieve non-linear feature fusion) → Output layer (Sigmoid activation, outputting the final risk probability);

[0155] Input: Temporal trend score of LSTM submodule, static risk probability of random forest submodule (two single-dimensional features, dimension 2);

[0156] Output: Final deformation risk probability (a continuous value from 0 to 1, enhanced information linkage between the two sub-modules through nonlinear fusion to improve the accuracy of risk assessment).

[0157] S422, Input Features: Integrates 6 core features—normalized deformation rate, geological vulnerability coefficient G, historical risk weight HI, real-time environmental factor E, data value density V, and link quality level, with feature dimensions fully aligned with the output of the intelligent data processing module;

[0158] Risk assessment and command triggering: The default risk probability threshold is set to 85% (can be customized according to the actual slope site requirements). When the model assesses that the deformation risk probability of a certain area is greater than or equal to the risk probability threshold, a pre-scheduling command is automatically generated. Through the control interface of the VPN transmission module, a superior VPN link is pre-allocated to the monitoring equipment corresponding to the area to ensure the transmission and processing efficiency of high-risk data in advance.

[0159] S43, Multi-dimensional hierarchical early warning engine:

[0160] Early warning level construction: Based on data value density and risk level, an early warning system is constructed: emergency early warning (V≥0.7 and deformation risk probability≥85%), critical early warning (0.3≤V<0.7 and deformation risk probability≥70%), and routine early warning (V<0.3 and deformation risk probability≥60%).

[0161] Through the detailed description of the above embodiments, the slope radar data VPN private network transmission and remote early warning system of the present invention constructs a complete solution for data acquisition, intelligent analysis, value-based transmission, and tiered early warning through the coordinated operation of four major modules. The system breaks down device protocol barriers with multi-source heterogeneous data adaptive analysis technology to achieve unified data alignment; it uses a data value density model as its core to achieve precise allocation of transmission resources; and finally, through the AI ​​trend prediction and tiered early warning mechanism of the B / S cloud early warning platform, it ensures priority handling of high-risk data. This system effectively solves the pain points of multi-source data compatibility difficulties and core data transmission lag in existing technologies, providing more efficient and reliable technical support for slope monitoring and helping to improve the intelligent level of geological disaster prevention and control.

[0162] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.

[0163] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0164] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0165] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0166] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0167] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0168] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0169] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A side slope radar data VPN private network transmission and remote early warning system, characterized in that, The slope radar monitoring module, the intelligent data processing module, the VPN transmission module based on value priority and the B / S cloud early warning platform are comprised; The slope radar monitoring module collects slope surface physical state data and environmental parameters to provide original data source for data value quantification; The intelligent data processing module takes data value density calculation as the core, integrates a multi-source heterogeneous monitoring protocol adaptive analysis engine and a data value density real-time sensing model, the multi-source heterogeneous monitoring protocol adaptive analysis engine realizes adaptive analysis of multi-vendor device protocols through dynamic protocol feature self-learning and cross-protocol conflict resolution technology, and outputs standardized data; The data value density real-time sensing model integrates deformation rate, geological vulnerability coefficient, historical risk weight, real-time environmental factors, combines dynamic weight coefficient and data time effectiveness decay coefficient, and generates quantified data value density; The VPN transmission module based on value priority first evaluates link delay, packet loss rate and bandwidth in real time through tools, divides link quality levels, divides data into emergency alarm type, key deformation trend type and routine scanning data type according to the data value density output by the intelligent data processing module, establishes a matching mechanism of data value level and link quality level, and makes high-value data preferentially occupy high-quality VPN links for transmission; The B / S cloud early warning platform realizes hierarchical storage based on data value density, and triggers multi-dimensional hierarchical early warning combined with AI trend prediction results and data value density, to ensure preferential treatment of high-value risk data.

2. The side slope radar data VPN private network transmission and remote early warning system according to claim 1, characterized in that, The multi-source heterogeneous monitoring protocol adaptive analysis engine comprises a dynamic protocol feature self-learning mechanism and a cross-protocol conflict resolution technology; The dynamic protocol feature self-learning mechanism realizes adaptive adaptation of unknown protocols by capturing device data frames, extracting fixed identifiers and variable field rules, classifying and inferring semantics of variable fields, updating and auditing feature libraries; The cross-protocol conflict resolution technology resolves protocol conflicts by constructing a field semantic ontology library, automatically identifying and precision preserving and converting data formats, adapting verification rules and verifying data integrity.

3. The side slope radar data VPN private network transmission and remote early warning system according to claim 2, characterized in that, The dynamic analysis execution process of the multi-source heterogeneous monitoring protocol adaptive analysis engine is as follows: Step 1: After receiving binary data, the starting segment of the data is extracted and matched with a preset protocol header feature library, if the matching fails, the dynamic protocol feature self-learning mechanism is triggered to analyze the protocol field structure; Step 2: Based on the information entropy algorithm, the data frame is segmented by calculating the information entropy of each byte position, combining the preset sliding window parameters to generate a continuous entropy value sequence, and synchronously calculating the entropy value mutation amplitude and mutation rate; Step 3: Set the preset entropy value mutation amplitude threshold and mutation rate threshold, when a byte position meets: the attenuation ratio of the entropy value compared with the entropy value of the previous byte position meets the standard, and the entropy value mutation amplitude and the entropy value mutation rate both reach the corresponding preset threshold, it is determined that the byte position is the field boundary, and the field segmentation of the data frame is completed; Step 4: For each independent field after segmentation, associate the system standard field based on the field mapping relationship obtained by the dynamic protocol feature self-learning mechanism, unify the data format according to the preset data format conversion rule in the cross-protocol conflict resolution technology, and resolve the field naming conflict through the field semantic ontology library.

4. The side slope radar data VPN private network transmission and remote early warning system according to claim 3, characterized in that, The entropy value mutation amplitude and the entropy value mutation rate are specifically as follows: The whole data frame is traversed to obtain a continuous entropy value sequence m is the data frame byte length; and the entropy value mutation amplitude is defined as ； Wherein, k is the byte position index, 2≤k≤m; Define the entropy value mutation rate as follows: ； wherein is a denominator compensation value.

5. The side slope radar data VPN private network transmission and remote early warning system according to claim 1, characterized in that, The data value density real-time perception model is specifically as follows: ； wherein, denotes the normalized deformation rate, with a value range of [0, 1]; denotes the geological vulnerability coefficient, with a value range of [0, 1]; denotes the historical risk weight, with a value range of [0, 1]; denotes the real-time environmental factor, with a value range of [0, 1]; denotes the dynamic weight coefficient, satisfying ; is a data timeliness decay coefficient, with a value range of [0.1, 1].

6. The side slope radar data VPN private network transmission and remote early warning system according to claim 5, characterized in that, The dynamic weight coefficient is dynamically adjusted by combining the random forest algorithm with the real-time environment, and is specifically as follows: Training data set construction: collect landslide event data in the past 5 years, add environment and deformation correlation data, and form a four-dimensional training data set including deformation rate, geological vulnerability, historical risk and environmental factors; Random forest model training: set the number of trees, maximum tree depth and feature selection method, and train for 50 iterations to make the model validation set accuracy meet the expectation; Real-time environment driven adjustment: based on the weight coefficient output by the random forest algorithm, the adjustment rule is preset according to expert experience for secondary optimization to adapt to the weight of extreme environment or sudden scene, and the adjustment rule is preset according to expert experience.

7. The side slope radar data VPN private network transmission and remote early warning system according to claim 1, characterized in that, The depth binding mechanism of data type, data value density and link quality based on value priority VPN transmission module is specifically as follows: Emergency alarm data: business data scenarios include slope landslide early warning and monitoring equipment failure, data value density V≥0.7, and transmission is required to be completed within 1s, and the priority of VPN link selection is to preferentially occupy VPN links with high link quality; Key deformation trend data: business data scenarios include slope deformation rate data of geological vulnerable zone and slope monitoring data after rainfall, data value density 0.3≤V<0.7, and transmission is required to be completed within 5s, and the priority of VPN link selection is to preferentially occupy VPN links with good link quality, and secondly to occupy VPN links with medium link quality; Routine scanning data: business data scenarios include slope point cloud data of large stable area and periodic slope inspection data, data value density V<0.3, transmission time is relaxed to 30s, and the priority of VPN link selection is to preferentially occupy VPN links with medium link quality, and secondly to occupy VPN links with poor link quality.

Images