A real-time analysis system for mine communication control and remote monitoring data

Abstract

The present application relates to the technical field of mine safety monitoring, and discloses a mine communication control and remote monitoring data real-time analysis system, which transmits a detection signal between communication nodes and calculates a channel transfer function to obtain a resonance fingerprint representing a physical environment of a mine; changes in the fingerprint are analyzed, and a hypothesis about a physical event is generated according to a physical model library that can be self-adaptively calibrated, and then an optimized detection signal most sensitive to the hypothesis is generated and controlled to be transmitted to verify the hypothesis; a sensor service signal transmitted through the channel is decoupled and processed using the real-time measured channel transfer function to restore original data; a relationship between the verified channel fingerprint change and a true value of the physical event is stored in a causal knowledge base, and state estimation is performed on an area where no physical sensor is deployed to realize virtual sensing and early warning. The present application realizes integration of communication and sensing, improves monitoring precision, data fidelity and early warning capability, and expands the monitoring coverage range.

Description

Technical Field

[0001] This invention relates to the field of mine safety monitoring technology, specifically to a mine communication control and remote monitoring data real-time analysis system. Background Technology

[0002] As a unique working environment, real-time and accurate monitoring of physical environmental parameters such as roof stability, gas composition, temperature, and humidity is a core prerequisite for ensuring production safety and preventing disasters in mines. Currently, mine safety monitoring generally relies on a monitoring network composed of a large number of dedicated sensor nodes. These sensors (such as roof segregation meters, gas sensors, and carbon monoxide sensors) are deployed at key locations in roadways and working faces. The collected data is aggregated to a ground monitoring center via wired (such as fiber optic cables or electrical cables) or wireless (such as Wi-Fi or LoRa) communication networks, thereby enabling remote monitoring of the mine's status.

[0003] However, this traditional monitoring model has inherent limitations. First, there is a significant contradiction between sensor deployment density and cost, resulting in blind spots in the vast and complex network of mine tunnels that cannot be directly covered by physical sensors. This leaves room for potential and developing risks. Second, with wireless communication becoming the mainstream transmission method underground, the complex electromagnetic environment and multipath propagation effects caused by tunnel structures in mines can severely interfere with and distort the operational data signals uploaded by sensors. This leads to discrepancies between the data received by the monitoring center and the actual measurements on site, affecting the accuracy of decision-making. Furthermore, existing systems are mostly passive data acquisition and alarm systems. They faithfully report measurements but lack in-depth insights into environmental changes and proactive verification capabilities. The systems cannot proactively detect potential, nascent risks, nor can they utilize the environmental information inherent in the ubiquitous communication signals themselves.

[0004] In summary, the separation of communication and sensing functions in existing technologies not only increases deployment and maintenance costs but also limits the valuable wireless channel resources to being used merely as "data pipelines," failing to realize their physical potential as environmental probes. This restricts the development of mine safety monitoring systems towards higher precision, wider coverage, and greater intelligence. Therefore, there is an urgent need for a new technological solution that can deeply integrate communication and environmental sensing, ensuring data transmission quality while proactively and intelligently analyzing and predicting environmental risks. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that existing mining communication and monitoring systems suffer from signal transmission distortion and insufficient data accuracy when facing the complex and ever-changing environment of mines, and it is difficult to effectively perceive the macroscopic evolution of the environmental state through changes in the communication signals themselves.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a real-time analysis system for mine communication control and remote monitoring data, comprising multiple communication nodes deployed underground and at least one processing unit, wherein the processing unit is configured to:

[0008] Control the first communication node to transmit a detection signal;

[0009] The second communication node is controlled to receive the response signal formed after the detection signal propagates in the mine environment;

[0010] Based on the detection signal and the response signal, calculate the channel transfer function that characterizes the channel characteristics between the first communication node and the second communication node;

[0011] Based on the change of the channel transfer function over time and according to the preset physical model library, the physical event hypothesis corresponding to the current environment is determined;

[0012] Based on the physical event hypothesis, signal parameters for an optimized detection signal are generated, the optimized detection signal being configured to verify the physical event hypothesis.

[0013] Control the first communication node to transmit the optimized detection signal generated based on the signal parameters.

[0014] Preferably, the processing unit is further configured to:

[0015] Receive service signals carrying service data transmitted by sensor nodes inside the mine;

[0016] Based on the channel transfer function, the service signal is decoupled to reconstruct the service data.

[0017] In one specific embodiment, the processing unit performs decoupling processing on the service signal, specifically configured as follows:

[0018] The inverse matrix of the channel transfer function is calculated using a pseudo-inverse or regularization method;

[0019] The inverse matrix is ​​applied to the service signal to reconstruct the service data.

[0020] Preferably, the physical model library contains a mapping relationship between mine physical events and the channel transfer function variation characteristics.

[0021] In one specific embodiment, the processing unit is further configured to:

[0022] Based on the response results to the optimized detection signal, the physical event hypothesis is verified or revised.

[0023] If the response result deviates from the prediction of the physical model library, the parameters of the physical model library are updated according to the deviation.

[0024] Preferably, the processing unit is further configured to:

[0025] Construct a causal knowledge base, which is used to store the correlation between the variation patterns of the channel transfer function and verified physical events;

[0026] Monitor the changing patterns of the channel transfer function of a communication link without deployed physical sensors;

[0027] Match the change pattern of the channel transfer function of the communication link with the change pattern in the causal knowledge base;

[0028] When a match is successful, virtual sensor readings associated with the area covered by the communication link are generated.

[0029] In one specific embodiment, the processing unit constructs the causal knowledge base, specifically configured as follows:

[0030] The physical events represented by the decoupled and reconstructed service data are associated with the change patterns of the channel transfer function corresponding to the occurrence of the physical event, and the association is stored in the causal knowledge base.

[0031] Preferably, the processing unit is further configured to:

[0032] Based on the virtual sensor readings, an early warning signal for the mine environment is generated.

[0033] In one specific embodiment, the detection signal is a linear frequency modulated signal.

[0034] In one specific embodiment, the optimized detection signal is a narrowband pulse signal whose center frequency is determined by the physical event hypothesis.

[0035] This invention provides a real-time analysis system for communication control and remote monitoring data in mining applications. It offers the following advantages:

[0036] 1. This invention controls communication nodes to transmit probe signals and calculate the channel transfer function. Further, based on the analyzed physical event hypotheses, it generates and transmits optimized probe signals for directional verification. This method forms a closed loop of "hypothesis-verification," enabling the judgment of environmental changes to no longer rely solely on single, passive measurements, but rather on active, directional detection for confirmation, thereby improving the accuracy and reliability of the final sensing results.

[0037] 2. This invention utilizes a channel transfer function calculated in real time, which precisely quantifies the distortion effects experienced by the signal along the propagation path. By decoupling the received service signals, the system can separate signal variations caused by environmental factors from the mixed signals, thereby reconstructing the original service data that is closer to the source, effectively reducing sensor data errors or distortions caused by the complex environment of mines.

[0038] 3. This invention constructs a causal knowledge base, associating specific channel transfer function variation patterns with verified physical events. Based on this knowledge base, the system can infer potential physical events at any communication node by monitoring channel changes and generate virtual sensor readings. This approach leverages existing communication infrastructure to reuse and extend sensing capabilities, expanding monitoring coverage without requiring extensive deployment of additional physical sensors.

[0039] 4. This invention actively detects and analyzes the channel transfer function, and adaptively adjusts the detection signal based on the analysis results for verification. This allows the acquisition of channel information related to changes in the physical environment of the mine. Using this channel information, on the one hand, sensor service data can be decoupled and its original state reconstructed; on the other hand, changes in channel information can be correlated with physical events, enabling the perception of the environmental state in areas where no sensors are deployed. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the system architecture of one embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram of the functional modules of a processing unit according to an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the environmental resonance fingerprint acquisition process according to an embodiment of the present invention;

[0043] Figure 4 This is a schematic diagram of the physical model analysis process according to an embodiment of the present invention;

[0044] Figure 5 This is a schematic diagram illustrating the process of establishing a causal knowledge base according to an embodiment of the present invention.

[0045] Among them, 100 is the processing unit; 110 is the signal transmission and reception control module; 120 is the channel function calculation module; 130 is the physical model analysis module; 131 is the physical model library; 140 is the adaptive detection decision module; 150 is the data decoupling module; 160 is the knowledge base and virtual perception module; 161 is the causal knowledge base; 200 is the communication node; 200a is the first communication node; 200b is the second communication node; and 300 is the sensor node. Detailed Implementation

[0046] To make the objectives, technical solutions, and effects of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] See attached document Figure 1 , Figure 1 This is a schematic diagram of a system architecture according to an embodiment of the present invention. The present invention provides a real-time analysis system for communication control and remote monitoring data in mining applications, which can be deployed in a mining environment.

[0048] In one embodiment, the system physically includes at least one processing unit 100, multiple communication nodes 200 deployed in mine roadways or working faces, and at least one sensor node 300.

[0049] Specifically, when the system performs a channel detection task, these communication nodes 200 are distinguished according to their functions as a first communication node 200a that transmits detection signals and a second communication node 200b that receives response signals. It should be noted that any physical communication node 200 can be defined as a first communication node or a second communication node at different times or in different tasks according to the instructions of the processing unit 100.

[0050] In one specific embodiment, the processing unit 100 may be a high-performance server deployed in a mine surface monitoring center, or a ruggedized, explosion-proof underground edge computing device. The processing unit 100 internally includes, but is not limited to: a central processing unit (CPU), a graphics processing unit (GPU) or dedicated digital signal processor (DSP) for parallel acceleration of calculations such as Fourier transforms, large-capacity memory, and non-volatile memory for storing physical model libraries and causal knowledge bases. The communication node 200 is a wireless communication device integrating a radio frequency transceiver front-end, antenna, digital-to-analog / analog-to-digital converter (ADC / DAC), and a field-programmable gate array (FPGA) or microcontroller (MCU) for performing low-level signal generation and reception control. The sensor node 300 is a standard underground sensing device integrating a specific physical quantity sensor (such as a methane sensor or temperature sensor) and a basic communication module.

[0051] The processing unit 100 may functionally include multiple logic modules for executing specific steps of the present invention. (See attached diagram.) Figure 2 , Figure 2 This is a schematic diagram of the functional modules of a processing unit according to an embodiment of the present invention. The processing unit 100 includes:

[0052] The signal transceiver control module 110 is configured to generate control commands and send the control commands to the designated communication node 200 through the communication interface to control the communication node 200 to transmit specific signals or perform signal receiving operations.

[0053] The channel function calculation module 120 is connected to the signal transceiver control module 110. This module receives the response signal data and the corresponding raw probe signal data collected by the communication node 200, and calculates the channel transfer function, which characterizes the channel characteristics between the two communication nodes, based on the two data.

[0054] The physical model analysis module 130 is connected to the channel function calculation module 120. This module internally stores a physical model library and is configured to receive channel transfer function data output by the channel function calculation module 120, analyze its time-varying characteristics, and determine the physical event hypotheses associated with these characteristics based on the physical model library.

[0055] The adaptive detection decision module 140 is connected to the physical model analysis module 130. This module is configured to receive a physical event hypothesis and generate signal parameters for an optimized detection signal to verify the hypothesis.

[0056] The data decoupling module 150 is connected to the channel function calculation module 120. This module is configured to receive service signals transmitted from the sensor node 300 via the communication node 200, and to process the service signals using the channel transfer function calculated by the channel function calculation module 120 to reconstruct the original service data.

[0057] The knowledge base and virtual sensing module 160 are connected to the physical model analysis module 130 and the data decoupling module 150. This module is configured to establish and store the correlation between channel transfer function variation patterns and physical events, and to generate virtual sensor readings based on this correlation.

[0058] The overall workflow of a system according to an embodiment of the present invention may include the following steps:

[0059] Step S1: The signal transceiver control module 110 generates and sends control commands to control a first communication node 200a to transmit an initial detection signal and to control a second communication node 200b to receive a response signal formed after propagation in the mine environment.

[0060] In step S2, the second communication node 200b transmits the received response signal data to the processing unit 100. The channel function calculation module 120 receives the response signal data and the corresponding initial probe signal data, and calculates the current channel transfer function characterizing the transmit / receive path based on the two.

[0061] Step S3: The physical model analysis module 130 receives the channel transfer function output by the channel function calculation module 120, compares it with historical data to obtain its change over time, and matches the change to one or more physical event hypotheses according to its internally stored physical model library, and then outputs the physical event hypothesis.

[0062] Step S4: The adaptive detection decision module 140 receives the physical event hypothesis output by the physical model analysis module 130 and generates a set of signal parameters for optimized detection signals based on the hypothesis. These optimized detection signals are configured for directional verification of the physical event hypothesis.

[0063] In step S5, the signal transceiver control module 110 receives the signal parameters generated by the adaptive detection decision module 140 and controls the first communication node 200a to transmit optimized detection signals accordingly to execute the verification process. The response results of this process can be used to correct physical event assumptions or update the physical model library.

[0064] Step S6: While the above process is being executed, if the sensor node 300 sends a service signal through the communication node 200, the data decoupling module 150 will obtain the service signal and call the current channel transfer function output by the channel function calculation module 120 to perform decoupling processing on the service signal, so as to output the reconstructed service data.

[0065] Step S7: The knowledge base and virtual perception module 160 receives and associates two inputs: one is the physical event verified in step S5, and the other is the physical event represented by the service data reconstructed in step S6. This module associates the physical event with its corresponding channel transfer function change pattern to update its internal causal knowledge base, and uses this knowledge base to realize the environmental status perception of the coverage area of ​​other communication nodes.

[0066] See attached document Figure 3 , Figure 3 This is a schematic diagram of an environmental resonance fingerprint acquisition process according to an embodiment of the present invention. In this embodiment, the environmental resonance fingerprint acquisition process is described in detail. This process is initiated by the signal transceiver control module 110, and the core calculation is completed by the channel function calculation module 120.

[0067] To obtain channel characteristics, the signal transceiver control module 110 controls a first communication node 200a to transmit a preset detection signal s(t).

[0068] In one specific embodiment, the detection signal s(t) is a linearly frequency modulated (LFM) signal. This type of signal has a large time-bandwidth product, and its autocorrelation function has a sharp peak, which allows the receiver to distinguish multiple response signal components with small time delays propagating from different paths. The LFM signal can be expressed as follows:

[0069]

[0070] Where A is the amplitude of the signal, f start k is the starting frequency of the signal. chirp For linear frequency modulation, T probe The duration of the signal. These parameters are pre-configured by the processing unit 100.

[0071] In another alternative embodiment, the probe signal s(t) can also be other broadband signals with good autocorrelation characteristics. For example, a signal modulated with a pseudo-random noise (PN) sequence (such as an m-sequence) can be used. Such signals exhibit broadband white noise in the spectrum, and their autocorrelation function is close to an impulse function, which is also beneficial for high-precision resolution of multipath delays. Alternatively, the probe signal can also be an orthogonal frequency division multiplexing (OFDM) signal. By loading a known pilot sequence onto each subcarrier, the channel transfer function at each subcarrier frequency point can be directly obtained at the receiver through a single FFT operation, without the need for division operations in the frequency domain, and higher measurement accuracy can be achieved under certain signal-to-noise ratio conditions.

[0072] The detection signal s(t) propagates along multiple paths in the mine environment, undergoing reflection, scattering, and diffraction from tunnel walls, equipment surfaces, rock interfaces, etc., before being received at the second communication node 200b, forming a response signal r(t). This response signal t(t) is the vector sum of the signal components propagating along all paths, carrying rich information about the physical structure and medium properties of the propagation environment.

[0073] The second communication node 200b digitizes the received response signal r(t) and transmits it to the processing unit 100. The channel function calculation module 120 receives the response signal data and calls the pre-stored data of the original detection signal s(t) corresponding to the response signal. This module performs Fourier transforms on s(t) and r(t) respectively, converting them from the time domain to the frequency domain to obtain S(f) and R(f).

[0074] Subsequently, the channel function calculation module 120 calculates the channel transfer function H(t,f), which characterizes the channel characteristics of the transmit / receive path, by performing a division operation in the frequency domain. This function is the environmental resonance fingerprint of the present invention. Its calculation formula is as follows:

[0075]

[0076] Where R(f) is the Fourier transform of the response signal r(t), S(f) is the Fourier transform of the original probe signal s(t), t is the measurement time, and f is the frequency. The calculated H(t,f) is a complex function whose amplitude |H(t,f)| represents the attenuation or gain of the signal amplitude at frequency f, and its phase ∠H(t,f) represents the phase shift of the signal at frequency f. This function's series of values ​​across the entire frequency band together constitute a precise quantitative characterization of the propagation characteristics of the mine's physical environment at the current moment.

[0077] See attached document Figure 4 , Figure 4 This is a schematic diagram of a physical model analysis process according to an embodiment of the present invention. In this embodiment, the functions of the physical model analysis module 130 and its internal physical model library 131 are described in detail.

[0078] The physical model analysis module 130 is configured to receive the time-varying channel transfer function H(t,f) output by the channel function calculation module 120. The core component of this module is a pre-defined physical model library 131. This library stores mapping relationships between various typical physical events within a mine and specific variation characteristics of the channel transfer function H(t,f). These mapping relationships are pre-calculated based on electromagnetic wave propagation theory or sound wave propagation theory, or established through experimental measurement and calibration.

[0079] In one specific embodiment, the physical model library 131 contains mappings for roadway roof settlement events. When a small settlement occurs in the roadway roof, with a distance change of Δd, this will cause a corresponding change in the length of a primary reflection path between the transmitting and receiving communication nodes. This change in path length will cause a frequency-dependent phase shift Δφ(f) in the signal components of that path. This phase shift can be determined by the following equation:

[0080]

[0081] Where c is the propagation speed of electromagnetic waves in air. This phase shift is superimposed on the phase ∠H(t,f) of the overall channel transfer function H(t,f). Therefore, the mapping relationship stored in the physical model library 131 is as follows: if the phase ∠H(t,f) of the channel transfer function exhibits a linear change across the entire operating frequency band, then this change pattern is mapped to the physical event hypothesis of "tunnel roof settlement". The distance change Δd of the settlement can be calculated based on the slope of this linear phase change.

[0082] In another specific embodiment, the physical model library 131 contains mappings for specific gas concentration changes. When the concentration of a gas (such as methane, primarily CH4) in the mine changes, it alters the complex permittivity of the air medium. This change affects the channel transfer function H(t,f) in two ways.

[0083] On the one hand, changes in the real part of the dielectric constant will cause changes in the signal propagation speed, resulting in a global phase shift.

[0084] On the other hand, at a specific frequency point (i.e., the position of the absorption spectral line of the gas molecule), the imaginary part of the dielectric constant will increase significantly, causing the signal energy to be rapidly absorbed at that frequency point, which manifests as one or more narrowband attenuation notches on the amplitude of the channel transfer function |H(t,f)|.

[0085] Therefore, the mapping relationship stored in the physical model library 131 is as follows: if a decay of |H(t,f)| is detected at a specific frequency corresponding to the gas absorption spectral line, then this change pattern is mapped to the physical event hypothesis of "increased concentration of a specific gas". The magnitude of the decay is related to the gas concentration level.

[0086] The physical model library 131 may contain mapping relationships for various other physical events, such as the movement of large conductive equipment (e.g., coal mining machines, hydraulic supports), deformation or fracture of roadway support structures, etc. Each event corresponds to one or a set of distinguishable variation characteristics of the channel transfer function in amplitude, phase, or time delay. The physical model analysis module 130 outputs one or more physical event hypotheses with confidence scores by matching the real-time monitored H(t,f) variation characteristics with the mapping relationships in the library.

[0087] In this embodiment, the process by which the physical model analysis module 130 generates physical event hypotheses is described in detail. This process is executed within the physical model analysis module 130, and its input is the channel transfer function H(t,f) continuously provided by the channel function calculation module 120.

[0088] First, the physical model analysis module 130 compares the current channel transfer function H(t,f) with a reference channel transfer function to calculate its change ΔH(t,f). The reference channel transfer function can be the baseline fingerprint H of the steady state measured during system initialization. base (f), or the channel transfer function H(t-Δt,f) of the previous measurement period. The change ΔH(t,f) can be determined by the following formula:

[0089] ΔH(t,f)=H(t,f)-H base (f);

[0090] This change ΔH(t,f) comprehensively reflects the changes in channel characteristics since the reference time.

[0091] Subsequently, the physical model analysis module 130 does not directly use the original change ΔH(t,f) for matching, but instead extracts a set of predefined technical features to form a feature vector. This step aims to transform the raw data into a more quantifiable indicator that is more sensitive to specific physical events. For example, for the models in the aforementioned physical model library 131:

[0092] For roadway roof settlement events, the module calculates the linear slope k of the phase ∠H(t,f) of ΔH(t,f) as a function of frequency f. phase This slope can be obtained by performing linear regression analysis on the phase data point set.

[0093] For a specific gas concentration change event, the module measures the amplitude of ΔH(t,f)|ΔH(t,f)| at the specific gas absorption spectral line frequency f. abs Attenuation depth M atten .

[0094] The physical model analysis module 130 combines all extracted technical features into a feature vector V. feat =[K phase M atten Meanwhile, the physical model library 131 pre-stores the corresponding ideal feature vectors for each physical event hypothesis (e.g., E1: roof settlement, E2: gas concentration change). ).

[0095] This module uses a pre-defined matching algorithm to process the feature vector V calculated in real time. feat With each ideal feature vector in the model library The comparisons are made, and a similarity score S is calculated. i The similarity score can be calculated using cosine similarity:

[0096]

[0097] Where · represents the vector dot product, and ||·|| represents the Euclidean norm of the vector.

[0098] Finally, the physical model analysis module 130 will calculate each similarity score S. i Matching threshold θ preset for this physical event i Compare them. If S i >θ i If the physical event that has occurred matches the event described by the i-th model, then the physical event E is set as follows. i Assuming E as a physical eventhypo Output. This output assumes the physical event E. hypo The data is then transmitted to the adaptive detection decision module 140 for subsequent verification. If the similarity scores of multiple models exceed their respective thresholds, the model with the highest score can be used as the primary hypothesis, or all hypotheses that meet the criteria can be output together.

[0099] In another alternative embodiment, the matching algorithm may employ a trained machine learning classifier. For example, a large number of channel feature vectors V corresponding to known physical events may be pre-collected. feat The data, along with its event labels, constitutes a training dataset. This dataset is then used to train a Support Vector Machine (SVM) model, a decision tree model, or a small neural network model. During the real-time analysis phase, the physical model analysis module 130 extracts the channel feature vector V... feat The data is fed into the trained classifier, which directly outputs the most probable physical event hypothesis category and its confidence level. This approach can learn more complex non-linear relationships between features and events, potentially improving matching accuracy.

[0100] In this embodiment, the process by which the adaptive detection decision module 140 generates parameters for optimizing the detection signal and performs verification is described in detail. The input to this process is the physical event hypothesis E output by the physical model analysis module 130. hypo .

[0101] The adaptive detection and decision module 140 receives the physical event hypothesis E. hypo The goal then is to generate a new set of optimized detection signal parameters. This is used to generate an optimized probe signal s that maximizes the sensitivity to changes in the hypothesized physical event. opt (t). This module assumes a sensitivity function for each physical event in the physics model library 131. And solve for the optimal detection signal parameters that maximize the function value.

[0102]

[0103] in, A set of parameters representing the detected signal, such as {center frequency, bandwidth, signal duration}.

[0104] In one specific embodiment, if the physical event output by the physical model analysis module 130 is assumed to be "tunnel roof settlement," its physical essence is a change in the length of the signal propagation path. According to the principle of electromagnetic wave propagation, the change in signal phase due to the change in path length is proportional to the signal frequency. Therefore, to maximize the measurement of this path length change, a signal with the highest possible frequency should be used. Accordingly, the adaptive detection decision module 140 generates optimized detection signal parameters. The signal type is a narrowband pulse signal, its center frequency is set to the highest operating frequency allowed by the system, and its bandwidth is set to a small value to concentrate energy.

[0105] In another specific embodiment, if the physical event output by the physical model analysis module 130 is assumed to be "an increase in the concentration of a specific gas", its physical essence is that the signal energy is at the absorption spectral frequency f of the gas molecule. abs The energy is absorbed at that frequency. To maximize the measurement of this energy absorption, the signal energy should be concentrated at that specific frequency. Based on this, the adaptive detection decision module 140 generates optimized detection signal parameters. The signal type is a narrowband pulse signal, and its center frequency is precisely set to the absorption spectral line frequency f of the gas being measured. abs .

[0106] The adaptive detection decision module 140 calculates the optimal detection signal parameters. Then, the optimized detection signal parameters are sent to the signal transceiver control module 110.

[0107] Upon receiving the optimized detection signal parameters, the signal transceiver control module 110 immediately generates a control command to control the first communication node 200a to transmit signals based on the optimized detection signal parameters. The generated optimized detection signal s opt (t). Simultaneously, the second communication node 200b receives its response signal r. opt (t) is then transmitted to the processing unit 100. The channel function calculation module 120 and the physical model analysis module 130 analyze the response signal. For example, in the roof settlement verification, r is analyzed. opt (t) Phase shift at high frequencies; in gas concentration verification, the analysis of r opt (t) Amplitude decay at the absorption spectral line frequency. If a specific characteristic quantity in the analysis result (such as phase shift value, amplitude decay value) exceeds the preset verification threshold, the physical event is assumed to be confirmed as a verified physical event.

[0108] See attached document Figure 4In this embodiment, the mechanism of the system's adaptive calibration of the physical model library 131 is described in detail. This mechanism is configured to be activated under specific conditions, namely, a quantifiable deviation between the result of the adaptive detection verification process and the prediction made by the physical model analysis module 130 based on the initial detection signal.

[0109] When the optimized detection signal s is as described above opt After (t) is launched and verified, the system obtains a verification value for the physical event parameters. Simultaneously, the physical model analysis module 130 has obtained a predicted value based on the feature vector extracted from the broadband probe signal in the initial stage. If this predicted value differs from the verification value, the system initiates the model calibration process.

[0110] In one specific embodiment, the calibration procedure is performed on model parameters for a "tunnel roof settlement" event. This model uses the linear slope k of the channel phase with frequency. phase The mapping is represented by the roof settlement Δd, and its initial mapping relationship can be expressed as Δd = C. old ·k phase C old These are the coefficients stored in the model library. The slope calculated by the physical model analysis module 130 based on the initial probe signal response. A settlement was predicted However, subsequent verification using optimized high-frequency narrowband pulse detection yielded a more precise settlement value of Δd. verified .

[0111] At this point, the system calculates the deviation between the predicted value and the verified value, and updates the model coefficients C based on this deviation. The updated coefficients C... new It can be calculated using the following formula:

[0112]

[0113] Here, α is a preset learning rate parameter with a value between 0 and 1, used to control the extent to which the original model parameters are modified during this calibration.

[0114] In another specific embodiment, the calibration procedure is performed on model parameters for a "specific gas concentration change" event. The model measures the signal attenuation depth M at a specific absorption spectral line frequency. atten The mapping to gas concentration Conc can be represented by a nonlinear function Conc = F(M). atten ;P old ), where P old Here is a set of model parameters. The concentration predicted based on the response to the initial detection signal is... The more precise concentration obtained through optimized detection verification is Conc. verified.

[0115] The system will (Conc) pred -Conc verified ) 2 As the loss function, gradient descent and other optimization algorithms are used to iteratively adjust the model parameters P to achieve the desired result. The output value approaches Conc verified .

[0116] The new model parameters (such as C) calculated using any of the above methods new or P new The parameters will be written back to the physical model library 131, replacing the old parameters. Thereafter, when making new physical event assumptions, the physical model analysis module 130 will use the updated model parameters, thereby adaptively adjusting the system's perception model of a specific mine environment over time to maintain the accuracy of its analysis results.

[0117] In this embodiment, the basic principle of the data decoupling module 150 in processing service signals is explained in detail.

[0118] A sensor node 300 deployed within the mine, such as a gas sensor, modulates the collected operational data into a raw operational signal x(t). This signal is transmitted via its antenna and propagates through the same physical mine channel as the aforementioned detection signal. During propagation, the signal is affected by channel multipath effects, attenuation, and phase distortion.

[0119] In signal processing theory, for a bandwidth-constrained signal, a wireless channel over a short time period can be modeled as a linear time-invariant (LTI) system. The characteristics of this system can be completely described by its channel impulse response h(t). Therefore, the traffic signal y(t) received by a communication node 200 can be expressed as the convolution of the original traffic signal x(t) and the channel impulse response h(t), superimposed with additive noise n(t). This relationship can be expressed in the time domain as:

[0120] y(t) = x(t) * h(t) + n(t);

[0121] Where * represents convolution operation, x(t) is the original service signal transmitted by sensor node 300, h(t) is the impulse response of the channel, n(t) is the additive noise in the channel, and y(t) is the service signal actually received by communication node 200 and transmitted to processing unit 100.

[0122] According to the convolution theorem of Fourier transform, convolution in the time domain is equivalent to multiplication in the frequency domain. Therefore, transforming the above equation to the frequency domain, we get:

[0123] Y(f) = X(f)·H(f) + N(f);

[0124] Where Y(f), X(f), H(f), and N(f) are the Fourier transforms of y(t), x(t), h(t), and n(t), respectively. H(f) is the channel transfer function, which is the same physical quantity as the environmental resonance fingerprint obtained by active detection in the aforementioned steps of this invention.

[0125] The fundamental goal of data decoupling is to solve for or estimate the frequency domain representation X(f) of the original service signal, given the received service signal Y(f) and the precisely measured channel transfer function H(f). Through inverse Fourier transform, the original time-domain service data can be reconstructed from the estimated X(f). Essentially, this process involves solving the aforementioned equations to recover the original signal X(f) from the signal Y(f) that has been "coupled" or "contaminated" by the channel H(f).

[0126] In this embodiment, the specific algorithm for implementing service data decoupling by the data decoupling module 150 is described in detail. The goal of this module is to solve for the original service signal X(f) based on the received service signal Y(f) and the known channel transfer function H(f).

[0127] In practical applications, due to frequency-selective fading in the channel, the amplitude of the channel transfer function H(f) will be very small at certain frequency points. If the solution is obtained directly by frequency domain division X(f) = Y(f) / H(f), the noise N(f) component will be greatly amplified at these frequency points, resulting in severe distortion of the final solution X(f) and making the method numerically unstable.

[0128] To address this issue, one embodiment of the present invention employs a regularization method for solution. The data decoupling module 150 constructs the process of solving X(f) as an optimization problem, the goal of which is to find an estimate X. est (f) This estimate, while fitting the received signal, also constrains the norm of the solution itself to suppress noise amplification. This process specifically employs the Tikhonov regularization method.

[0129] The data decoupling module 150 performs the following calculation steps to obtain the estimated original service signal X. est (f). Given the frequency domain representation Y(f) of the received traffic signal and the known channel transfer function H(f), the estimated original traffic signal X... est (f) is determined by the following formula:

[0130] X est (f)=(H(f)*H(f)+λI) -1 H(f)*Y(f);

[0131] in:

[0132] X est (f) is the estimated value of the original service signal in the frequency domain after processing.

[0133] Y(f) is the frequency domain representation of the service signal received by communication node 200.

[0134] H(f) is the instantaneous channel transfer function output by the channel function calculation module 120 and measured at the corresponding time. In a multiple-input multiple-output (MIMO) system, H(f) is a matrix; in a single-input single-output (SISO) system, it is a complex number.

[0135] It is important to understand that, due to the dynamic changes in the mine environment, the channel transfer function is physically a two-dimensional function of time t and frequency f, and its complete form can be represented as a time-varying channel transfer function H(t,f). The notation H(f) used here in this invention is a "snapshot" of this time-varying function at a specific measurement moment, used to describe the channel state at that instant for decoupling calculations. Therefore, in the context of this invention, H(f) is used whenever the channel state is involved in instantaneous calculations, while its dynamic changes refer to its property as a time-varying function H(t,f).

[0136] H(f) * It is the conjugate transpose of H(f) (or the complex conjugate in the SISO system).

[0137] I is an identity matrix with dimensions equal to H(f). * They have the same dimensions.

[0138] λ is a preset, positive real number called the regularization coefficient. This coefficient is used to balance the fidelity and smoothness of the solution. If the value of λ is too small, the solution tends to be an unstable inverse solution; if the value of λ is too large, the solution will be over-smoothed, potentially losing signal details. Its specific value can be preset or adaptively adjusted according to conditions such as the channel's signal-to-noise ratio (SNR).

[0139] Conceptually, the above calculation process can be understood as the following two steps:

[0140] First, the inverse matrix of the channel transfer function H(f) is calculated using a regularization method. Specifically, this inverse matrix is ​​(H(f)*H(f)+λI) in the above equation. -1 H(f) * The part, mathematically, is the Tikhonov regularized pseudoinverse of the channel transfer function H(f).

[0141] Then, the inverse matrix is ​​applied to the frequency domain representation Y(f) of the service signal, i.e., matrix multiplication is performed, to obtain an estimate X of the original service signal's frequency domain representation. est(f).

[0142] By performing the above calculations, the data decoupling module 150 calculates an estimate X for the frequency domain representation of the original service signal for each frequency point f. est The value of (f). In obtaining the estimated frequency domain traffic signal X for the entire frequency band. est (f) After that, the module performs an inverse Fourier transform on it to obtain the reconstructed time-domain service signal x. est (t). The signal x est (t) represents the high-fidelity service data signal that has been freed from the effects of channel distortion.

[0143] See attached document Figure 5 , Figure 5 This is a schematic diagram illustrating the process of establishing a causal knowledge base according to an embodiment of the present invention. In this embodiment, the process of establishing and expanding the internal causal knowledge base 161 of the knowledge base and the virtual perception module 160 is described in detail.

[0144] The causal knowledge base 161 is configured as a structured database to store deterministic associations between specific variation patterns of a time-varying channel transfer function H(t,f) and a confirmed physical event. The generation of a knowledge base record requires two synchronized data inputs: channel characteristic data as the "cause" and truth data of the physical event as the "result."

[0145] The knowledge base and virtual sensing module 160 are configured to obtain truth data for this physical event from two sources:

[0146] The first source is the output of the adaptive detection and verification process. When a physical event hypothesis (such as "tunnel roof settlement") is confirmed as an actual physical event through verification of optimized detection signals, the confirmed event and its quantified parameters (such as settlement Δd) are... verified This is sent as one input to the knowledge base and virtual perception module 160.

[0147] The second source is the output of the data decoupling module 150. When a sensor node 300 is deployed near a communication link, the data decoupling module 150 outputs high-fidelity service data (such as gas concentration values ​​Conc). est This data can be directly used as the truth value of the physical state at that location. This truth value data is then sent as another input to the knowledge base and the virtual sensing module 160.

[0148] Each record stored in the causal knowledge base 161 can have the following data structure:

[0149] K i ={LinkID,V feat E type ,Pvalue ,T stamp};

[0150] in:

[0151] LinkID is a unique identifier for a communication link where a channel change occurs (such as an ID pair between the first and second communication nodes).

[0152] V feat This refers to the feature vector directly related to the event extracted from the channel transfer function change ΔH(t,f) of the link, for example, containing the phase slope k. phase and specific frequency attenuation depth M atten The value.

[0153] E type For the type of confirmed physical event, for example, an enumeration value representing "roof settlement" or "gas concentration change".

[0154] P value The true value of the quantization parameter associated with the physical event, for example, a structure containing a value and a unit {value:5,unit:"mm"} or {value:1.5,unit:"%"}.

[0155] T stamp This is the timestamp when the associated record was generated.

[0156] The knowledge base and virtual perception module 160 establish a new knowledge base record through the following steps:

[0157] First, this module monitors the two truth data sources mentioned above. When at time T... stamp Receive a new physical event truth data P from any source. value Upon that, the module immediately retrieves the location (LinkID) and time (T) of the event. stamp The corresponding channel feature vector V feat .

[0158] Then, the module will linkID, V feat Event Type E type Event parameter truth value P value and timestamp T stamp Assemble into a new record K new .

[0159] Finally, this new record K new The data is stored in the data storage area of ​​the causal knowledge base 161. By continuously executing this process, the causal knowledge base 161 continuously accumulates a high-reliability mapping relationship between channel changes and physical reality under a specific mining environment.

[0160] See attached document Figure 5 In this embodiment, the functions of virtual perception and early warning implemented by the knowledge base and virtual perception module 160 are described in detail. This function aims to quantitatively estimate the environmental state of areas where sensor nodes 300 are not directly deployed.

[0161] It should be noted that, in this invention, "the area covered by the communication link" mainly refers to the main signal propagation path between the two communication nodes that form the communication link, as well as the surrounding space that can significantly affect the signal propagating along that path.

[0162] In a specific physical model, this region can be understood as the first Fresnel zone of the communication path. Any physical event occurring within this region that alters the propagation characteristics of electromagnetic waves (such as roof settlement causing changes in path length, or gas accumulation causing changes in dielectric constant) will be reflected in the changes in the channel transfer function of the link.

[0163] When a communication link LinkID X When no physical sensors are deployed on the link, but its channel transfer function changes, the system initiates a virtual sensing process. First, the physical model analysis module 130 extracts a real-time feature vector from the change in the channel transfer function ΔH(t,f) of the link, using the same method as when generating the physical event hypothesis.

[0164] The knowledge base and virtual sensing module 160 receive the channel feature vector. It then performs a retrieval operation within its internal causal knowledge base 161. The goal of this retrieval operation is to find, among all records stored in the causal knowledge base 161, a match with... The most similar feature vector. In one specific embodiment, this similarity measure is calculated by... With each record K in the knowledge base i eigenvectors The Euclidean distance between them is used to determine:

[0165]

[0166] Where, d i The distance between two feature vectors.

[0167] Module selection makes distance d i The smallest one or more knowledge base records are selected as the best match. Let K be the best matching record. j The knowledge base and virtual perception module 160 are derived from this record K. j Extract the associated physical event types. Quantization parameters of physical events

[0168] The extracted physical event quantization parameters That is, it is defined by the system in the communication link LinkID X Virtual sensor readings generated for the covered area. For example, if the best matching record K... j If the physical event stored in the system is "tunnel roof settlement" and its quantization parameter is {value:5,unit:"mm"}, then the system will output a virtual sensor reading indicating that a 5 mm roof settlement has occurred at that location.

[0169] Finally, the system compares the virtual sensor reading with a preset warning threshold table. This table sets a safety threshold for each type of physical event. If the virtual sensor reading exceeds the warning threshold for its corresponding event type, the system generates a warning signal. This warning signal can be sent to the terminal equipment at the wellhead monitoring center to alert operators to potential risks.

[0170] In a preferred embodiment, when a region to be sensed is simultaneously covered by signals from multiple different communication links (e.g., link AB and link CD), the knowledge base and virtual sensing module 160 can be configured to fuse virtual sensing results from these multiple links to improve the accuracy and robustness of the virtual sensing results. Specifically, the system can calculate an independent virtual sensor reading for each link and then perform a weighted average of these readings. The weights can be determined based on the signal quality (e.g., signal-to-noise ratio), historical sensing accuracy, or similarity score with the best-matching knowledge base record for each link. By fusing information from multiple links, misjudgments caused by accidental interference or model bias in a single link can be effectively suppressed, thereby generating a more reliable regional status assessment and early warning signal.

[0171] In summary, this invention obtains a resonant fingerprint characterizing the physical environment of a mine by transmitting probe signals between communication nodes and calculating the channel transfer function. The system analyzes changes in this fingerprint and generates hypotheses about physical events such as roof subsidence and gas concentration changes based on an adaptively calibrated physical model library. Subsequently, it generates and controls the transmission of optimized probe signals most sensitive to these hypotheses for verification. Simultaneously, the system uses the real-time measured channel transfer function and a regularized decoupling algorithm to process sensor traffic signals transmitted through the channel to eliminate channel distortion and recover the original data. Furthermore, the system stores the verified relationship between channel fingerprint changes and the truth values ​​of physical events in a causal knowledge base and uses this knowledge base to perform state estimation in areas without deployed physical sensors, thereby achieving virtual sensing and early warning.

[0172] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time analysis system for mine communication control and remote monitoring data, comprising a plurality of communication nodes deployed in a mine and at least one processing unit, characterized in that, The processing unit is configured to: control a first communication node to transmit a probe signal; control a second communication node to receive a response signal formed after the probe signal propagates in the mine environment; based on the probe signal and the response signal, calculate a channel transfer function representing a channel characteristic between the first communication node and the second communication node; based on a change of the channel transfer function over time and according to a preset physical model library, determine a physical event hypothesis corresponding to a current environment; the physical model library contains a mapping relationship between mine physical events and change characteristics of the channel transfer function; based on the physical event hypothesis, generate signal parameters of an optimized probe signal, the optimized probe signal being configured to verify the physical event hypothesis; control the first communication node to transmit the optimized probe signal generated based on the signal parameters.

2. The real-time analysis system for communication control and remote monitoring data of mine according to claim 1, characterized in that, The processing unit is further configured to: receive a service signal carrying service data transmitted by a sensor node in the mine; based on the channel transfer function, perform decoupling processing on the service signal to reconstruct the service data.

3. The real-time analysis system for mine communication control and remote monitoring data according to claim 2, characterized in that, The processing unit performs decoupling processing on the service signal, and is specifically configured to: calculate an inverse matrix of the channel transfer function using a pseudo-inverse or regularization method; apply the inverse matrix to the service signal to reconstruct the service data.

4. The real-time data analysis system for mine communication control and remote monitoring according to claim 1, characterized in that, The processing unit is further configured to: based on a response result of the optimized probe signal, verify or correct the physical event hypothesis; if the response result deviates from a prediction of the physical model library, update parameters of the physical model library according to the deviation.

5. The real-time data analysis system for mine communication control and remote monitoring according to claim 2, characterized in that, The processing unit is further configured to: construct a causal knowledge base, the causal knowledge base being used to store an association relationship between a change pattern of a channel transfer function and a verified physical event; monitor a change pattern of a channel transfer function of a communication link without deploying a physical sensor; match the change pattern of the channel transfer function of the communication link with a change pattern in the causal knowledge base; when the matching is successful, generate a virtual sensor reading associated with an area covered by the communication link.

6. The real-time analysis system for mine communication control and remote monitoring data according to claim 5, characterized in that, The processing unit constructs the causal knowledge base, and is specifically configured to: associate a physical event represented by the service data reconstructed after decoupling processing with a change pattern of a channel transfer function corresponding to a time when the physical event occurs, and store the association relationship in the causal knowledge base.

7. The real-time analysis system for mine communication control and remote monitoring data according to claim 5, characterized in that, The processing unit is further configured to: based on the virtual sensor reading, generate a warning signal of the mine environment.

8. The real-time data analysis system for mine communication control and remote monitoring according to claim 1, characterized in that, The probe signal is a linear frequency modulation signal.

9. The real-time data analysis system for mine communication control and remote monitoring according to claim 1, characterized in that, The optimized probe signal is a narrowband pulse signal, and a center frequency thereof is determined by the physical event hypothesis.

Images