Edge computing compression transmission method and system for diving monitoring data

Abstract

The invention discloses an edge computing compression transmission method and system for diving monitoring data, and relates to the field of data transmission, and the method comprises the steps: obtaining diving environment parameters; mapping the current pressure data, the first-order derivative and the second-order derivative to a three-dimensional state space; when the coordinate point of the three-dimensional state space is located in the target working condition area, outputting a current working condition ID; according to the current working condition ID, defining a pressure sensor, a temperature sensor and a gas partial pressure sensor as grid nodes in the physical correlation topological graph; calculating to obtain theoretical expected coordinates of each sensor node; acquiring actual measurement node coordinates of each sensor node, and mapping the actual measurement node coordinates to the grid nodes; calculating a deformation distance, and determining a grid deformation tensor; determining a characteristic frequency coefficient and a phase spectrum; and compressing and packaging the current working condition ID, the characteristic frequency coefficient and the phase spectrum into a data frame, and sending the data frame to a water surface monitoring master station. The safety of the saturated diving operation can be effectively guaranteed.

Description

Edge computing compression transmission method and system for underwater monitoring data Technical Field

[0001] This application relates to the field of data transmission, and in particular to an edge computing compression transmission method and system for underwater monitoring data. Background Technology

[0002] In deep-sea saturation diving operations, divers need to remain in hyperbaric chambers for extended periods, where pressure can reach tens of atmospheres. To ensure diver safety, saturation diving monitoring systems need to collect and transmit a large amount of environmental parameter data in real time, including key indicators such as chamber pressure, temperature, oxygen partial pressure, and carbon dioxide partial pressure. This monitoring data is transmitted from edge PLC slave stations to the surface monitoring master station via an industrial fieldbus for real-time monitoring of the divers' living environment.

[0003] However, saturation diving operations exhibit significant phased characteristics. During the pressurization phase, the pressure inside the chamber rises rapidly according to a pre-set pressurization curve; during the pressure holding phase, the pressure is maintained at a constant value corresponding to the target depth for an extended period; and during the decompression phase, the pressure decreases slowly according to a strict decompression gauge. Under different operational conditions, there are clear physical relationships between various environmental parameters. For example, according to the ideal gas law, there is a deterministic mathematical relationship between pressure, temperature, and gas partial pressure within a confined chamber; simultaneously, the diver's physiological metabolism consumes oxygen and produces carbon dioxide, causing changes in gas partial pressure to follow a predictable metabolic model.

[0004] In existing technologies, monitoring systems typically use a fixed sampling frequency to periodically transmit raw data from all sensors, or use general data compression algorithms to compress the data stream. However, during periods of stable data, such as pressure holding, the changes in sensor readings are minimal and conform to physical expectations. The system still needs to continuously transmit a large number of redundant heartbeat data packets, resulting in wasted bandwidth resources. Furthermore, during periods of drastic changes in operating conditions, such as pressure increases or decreases, multiple sensor data change rapidly simultaneously, causing a surge in instantaneous data volume. This can easily lead to congestion on the industrial bus network, affecting the real-time transmission of critical data and potentially delaying the timely detection and handling of abnormal situations.

[0005] In summary, the existing data transmission methods have low stability and cannot guarantee the safety of saturation diving operations. Summary of the Invention

[0006] This application provides an edge computing compression transmission method and system for diving monitoring data, which effectively ensures the stability of data transmission and thus effectively protects the safety of saturation diving operations.

[0007] To achieve the above objectives, the embodiments of this application adopt the following technical solutions: Firstly, an edge computing compression transmission method for diving monitoring data is provided, applied to a saturated diving monitoring system. The saturated diving monitoring system includes at least one edge PLC slave station and a surface monitoring master station. The method includes: acquiring diving environment parameters through the edge PLC slave station, wherein the diving environment parameters include current pressure data, pressure data sequence, temperature, and gas partial pressure; calculating the first and second derivatives of the pressure data sequence in real time, and mapping the current pressure data, first and second derivatives to a three-dimensional state space; outputting the current operating condition ID when the coordinate point in the three-dimensional state space is located in a preset target operating condition area; retrieving the corresponding physical association topology map according to the current operating condition ID, and mapping the pressure sensor, temperature sensor, and surface monitoring master station in the physical association topology map. Sensors and gas partial pressure sensors are defined as grid nodes. The theoretical rigid displacement of the physical correlation topology is calculated to obtain the theoretical expected coordinates of each sensor node. The measured node coordinates of each sensor node are obtained and mapped to the grid nodes. The deformation distance between the measured node coordinates and the theoretical expected coordinates is calculated, and the grid deformation tensor is determined based on the deformation distance. Based on the grid deformation tensor, the characteristic frequency coefficients and phase spectrum are determined. The current operating condition ID, characteristic frequency coefficients, and phase spectrum are compressed and encapsulated into a data frame, and the data frame is sent to the water surface monitoring master station. The water surface monitoring master station is used to invert and restore the grid deformation tensor based on the characteristic frequency coefficients and phase spectrum, generate the theoretical rigid displacement based on the current operating condition ID, and superimpose the grid deformation tensor onto the theoretical rigid displacement to restore the original time series data of each sensor.

[0008] In one possible implementation of the first aspect, the first and second derivatives of the pressure data sequence are calculated in real time, and the current pressure data, the first and second derivatives are mapped to a three-dimensional state space, including: calculating the first derivative of the pressure data sequence using the finite difference method to obtain the pressure change rate; calculating the second derivative of the pressure data sequence using the quadratic difference method to obtain the pressure change acceleration; and constructing real-time state points in the three-dimensional state space with the current pressure data as the X-axis coordinate, the pressure change rate as the Y-axis coordinate, and the pressure change acceleration as the Z-axis coordinate.

[0009] In another possible implementation of the first aspect, when the coordinates of the three-dimensional state space are located in a preset target working condition area, the current working condition ID is output, including: obtaining the coordinates of the real-time state point in the three-dimensional state space; calculating the distance between the real-time state point and the center point of each target working condition area based on the coordinates of the real-time state point; selecting the target working condition area with the smallest distance as the matching working condition area; determining whether the real-time state point is located within the boundary range of the matching working condition area; if the real-time state point is located within the boundary range of the matching working condition area, the working condition ID corresponding to the matching working condition area is output as the current working condition ID; if the real-time state point is not located within the boundary range of any target working condition area, it is marked as an abnormal working condition, and a default working condition ID is output.

[0010] In another possible implementation of the first aspect, calculating the theoretical rigid displacement of the physical correlation topology to obtain the theoretical expected coordinates of each sensor node includes: determining the main pressure sensor in the physical correlation topology and obtaining the current measured pressure value of the main pressure sensor; calculating the theoretical expected temperature value of the temperature sensor node at the current measured pressure value according to a preset gas state equation; calculating the theoretical expected partial pressure value of each gas partial pressure sensor node at the current measured pressure value according to a preset partial pressure formula; using the theoretical expected temperature value and the theoretical expected partial pressure value as the theoretical expected coordinates of the corresponding sensor node in the physical correlation topology; and defining the set of theoretical expected coordinates of all sensor nodes as the theoretical rigid displacement of the physical correlation topology.

[0011] In another possible implementation of the first aspect, the deformation distance between the measured node coordinates and the theoretical expected coordinates is calculated, and the mesh deformation tensor is determined based on the deformation distance. This includes: calculating the Euclidean distance between the measured node coordinates and the corresponding theoretical expected coordinates of each sensor node to obtain the deformation distance; retaining sensor nodes whose deformation distance is greater than a preset sensor allowable error threshold and removing sensor nodes whose deformation distance is less than or equal to the sensor allowable error threshold; for the retained sensor nodes, calculating their deformation direction vector and deformation amplitude based on the deformation distance, and constructing the mesh deformation tensor based on the deformation direction vector and deformation amplitude.

[0012] In another possible implementation of the first aspect, the deformation direction vector and deformation amplitude are calculated based on the deformation distance, and a mesh deformation tensor is constructed based on the deformation direction vector and deformation amplitude. This includes: for each retained sensor node, calculating the difference vector between the measured node coordinates and the theoretically expected coordinates; normalizing the difference vector to obtain a unit deformation direction vector; using the deformation distance as the deformation amplitude; multiplying each component of the deformation direction vector by the deformation amplitude to obtain a deformation displacement vector; arranging the deformation displacement vectors of all retained sensor nodes according to their node indices to construct a deformation displacement matrix; extracting the eigenvalues ​​and eigenvectors corresponding to the main deformation modes in the deformation displacement matrix; and reconstructing the mesh deformation tensor based on the eigenvalues ​​and eigenvectors.

[0013] In another possible implementation of the first aspect, the characteristic frequency coefficients and phase spectrum are determined based on the mesh deformation tensor, including: performing a Fourier transform on the mesh deformation tensor to convert the deformation information in the spatial domain to the frequency domain; calculating the amplitude of each frequency component in the frequency domain to obtain the amplitude spectrum; sorting the amplitude spectrum in descending order and selecting the top M frequency components with the largest amplitudes as characteristic frequencies, where M is a positive integer; extracting the amplitude corresponding to each characteristic frequency and normalizing the amplitude to obtain the characteristic frequency coefficients; and extracting the phase angle corresponding to each characteristic frequency to construct the phase spectrum, wherein the characteristic frequency coefficients are encoded using floating-point compression and the phase spectrum is encoded using angle discretization.

[0014] In another possible implementation of the first aspect, the current operating condition ID, characteristic frequency coefficients, and phase spectrum are compressed and encapsulated into a data frame, including: constructing a data frame header, writing a frame synchronization identifier, data frame length, and timestamp into the data frame header; adding a current operating condition ID field after the data frame header and using fixed-length encoding; performing Huffman coding on the characteristic frequency coefficients to generate a compressed data block of characteristic frequency coefficients; performing differential encoding and run-length encoding on the phase spectrum to generate a compressed data block of phase spectrum; concatenating the data frame header, current operating condition ID field, compressed data block of characteristic frequency coefficients, and compressed data block of phase spectrum in that order; and adding a cyclic redundancy check code to the end of the data frame to complete the encapsulation of the data frame.

[0015] Secondly, this application provides a machine-readable storage medium storing instructions that cause a machine to execute the aforementioned edge computing compression transmission method for diving monitoring data.

[0016] Thirdly, this application provides a saturation diving monitoring system, comprising: at least one edge PLC slave station; and a surface monitoring master station connected to each edge PLC slave station.

[0017] The above technical solution achieves automatic identification of operating conditions through three-dimensional state space mapping. Pressure data, first derivatives, and second derivatives are mapped to a three-dimensional space, where different operating condition stages exhibit distinct regional distribution characteristics. The pressure holding stage clusters in regions where the derivative is close to zero, while the pressurization and depressurization stages correspond to regions with positive and negative derivatives, respectively. Identifying the operating condition ID lays the foundation for subsequent differentiated processing. Based on the current operating condition ID, the corresponding topology map is retrieved, and various sensors are defined as grid nodes. The connections between nodes reflect the constraints of physical laws such as the ideal gas law and physiological metabolic patterns. Theoretical rigid displacement can be calculated to obtain the theoretically expected coordinates of each sensor under the current operating condition, derived from physical laws. By comparing the measured node coordinates with the theoretically expected coordinates, the deformation distance is calculated and the deformation tensor is determined. This allows us to ascertain the deviation between the actual measured value and the theoretically expected value. Under normal operating conditions, if all sensor data conform to the constraints of physical laws, the deformation tensor is close to zero. Significant deviations only occur under abnormal conditions. Therefore, the system no longer transmits all original data but only the deviation information, i.e., the deformation tensor. During periods of stable data, such as pressure holding, the measured and theoretical values ​​are highly consistent, resulting in a very small deformation tensor and a significant reduction in the amount of data transmitted, thus resolving the redundant heartbeat packet problem. To further improve compression efficiency, the mesh deformation tensor is converted to the frequency domain to extract characteristic frequency coefficients and the phase spectrum. Utilizing the sparsity of the deformation tensor in the frequency domain and the ability of the phase spectrum to preserve temporal structure, only three key pieces of information—operating condition ID, characteristic frequency coefficients, and the phase spectrum—need to be transmitted. The data frame size is much smaller than the complete original data. Although the deformation tensor may increase during periods of drastic changes in operating conditions, frequency domain compression can still effectively control the amount of data and avoid network congestion. After receiving compressed data frames, the surface monitoring master station generates theoretical rigid displacement based on the working condition ID. It then inverts and restores the mesh deformation tensor based on the characteristic frequency coefficient and phase spectrum. The deformation tensor is superimposed on the theoretical rigid displacement to completely restore the original time-series data of each sensor. This ensures both data integrity and accuracy while achieving high compression ratio transmission. It realizes intelligent edge computing compression, ensures the stability of data transmission, and thus effectively guarantees the safety of saturation diving operations.

[0018] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0019] Figure 1 is a flowchart illustrating an edge computing compression transmission method for diving monitoring data provided in an embodiment of this application; Figure 2 is a structural diagram illustrating a saturated diving monitoring system provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0021] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0022] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0023] Figure 1 schematically illustrates a flowchart of an edge computing compression transmission method for diving monitoring data according to an embodiment of this application. As shown in Figure 1, this application embodiment provides an edge computing compression transmission method for diving monitoring data, applied to a saturated diving monitoring system. The saturated diving monitoring system includes at least one edge PLC slave station and a surface monitoring master station. The method may include the following steps.

[0024] S110. Obtain diving environment parameters from the edge PLC slave station, including current pressure data, pressure data sequence, temperature, and gas partial pressure; S120. Calculate the first and second derivatives of the pressure data sequence in real time and map the current pressure data, first and second derivatives to a three-dimensional state space; S130. When the coordinates of the three-dimensional state space are located in the preset target operating condition area, output the current operating condition ID; S14. Retrieve the corresponding physical association topology diagram based on the current operating condition ID, and define the pressure sensor, temperature sensor, and gas partial pressure sensor as grid nodes in the physical association topology diagram; S150. Calculate the theoretical rigid displacement of the physical association topology diagram to obtain the theoretical displacement of each sensor node. S160: Obtain the measured node coordinates of each sensor node and map the measured node coordinates to the grid node; S170: Calculate the deformation distance between the measured node coordinates and the theoretical expected coordinates, and determine the grid deformation tensor based on the deformation distance; S180: Determine the characteristic frequency coefficients and phase spectrum based on the grid deformation tensor; S190: Compress and encapsulate the current operating condition ID, characteristic frequency coefficients, and phase spectrum into a data frame, and send the data frame to the water surface monitoring master station. The water surface monitoring master station is used to invert and restore the grid deformation tensor based on the characteristic frequency coefficients and phase spectrum, generate the theoretical rigid displacement based on the current operating condition ID, and superimpose the grid deformation tensor onto the theoretical rigid displacement to restore the original time series data of each sensor.

[0025] In this embodiment, the surface monitoring master station reconstructs the grid deformation tensor based on the characteristic frequency coefficients and phase spectrum inversion. Based on the current operating condition ID, it generates a theoretical rigid displacement and superimposes the grid deformation tensor onto the theoretical rigid displacement to restore the original time-series data of each sensor. This includes: the surface monitoring master station receiving data frames and performing cyclic redundancy check to verify the integrity of the data frames; after successful verification, parsing the data frame header to extract the timestamp and data frame length; extracting the current operating condition ID from the data frame and retrieving pre-stored physical correlation topology diagrams and gas state equation parameters based on the current operating condition ID; and performing Huffman decoding on the characteristic frequency coefficient compressed data block. The process involves: restoring the characteristic frequency coefficients; performing run-length decoding and differential decoding on the compressed phase spectrum data block to restore the phase spectrum; reconstructing the complete frequency domain representation based on the characteristic frequency coefficients and phase spectrum using inverse Fourier transform; converting the frequency domain representation back to the spatial domain to inversely restore the mesh deformation tensor; calculating the theoretical rigid displacement based on the current working condition ID and physical association topology to obtain the theoretical expected coordinates of each sensor node; superimposing the deformation displacement vector in the mesh deformation tensor onto the theoretical expected coordinates of the corresponding sensor node to obtain the restored coordinate values ​​of each sensor; and arranging the restored coordinate values ​​according to the timestamp to generate the original time-series data for each sensor.

[0026] In the actual implementation of saturation diving operations, the edge PLC slave station first acquires diving environment parameters through various sensors deployed in the hyperbaric chamber. Pressure sensors monitor real-time pressure changes within the chamber. Temperature sensors monitor temperature fluctuations. Gas partial pressure sensors include oxygen and carbon dioxide partial pressure sensors. The edge PLC slave station cyclically reads all sensor values ​​at a fixed sampling period. Current pressure data refers to the pressure reading at the current moment, while the pressure data sequence contains historical pressure data from the current moment and several previous moments (e.g., the most recent 10 sampling points), forming a pressure change trend within a time window. Temperature and gas partial pressure data are also acquired at the current moment. This raw data is temporarily stored in the edge PLC slave station's memory buffer, providing a data foundation for subsequent real-time calculations and analysis. Data acquisition at the edge avoids directly transmitting all raw data to the surface master station, laying the foundation for edge computing compression.

[0027] After acquiring the pressure data sequence, the edge PLC slave station immediately performs real-time differential calculations to extract the dynamic characteristics of pressure changes. The first derivative reflects the rate of pressure change and is calculated using the backward difference method. This involves subtracting the previous pressure value from the current pressure value and then dividing by the sampling time interval to obtain the rate of pressure change, expressed in atmospheres per second. For example, if the current pressure is 20.5 atmospheres, the pressure one second ago was 20.3 atmospheres, and the sampling interval is 1 second, then the first derivative is 0.2 atmospheres per second, indicating that the pressure is rising at a rate of 0.2 atmospheres per second.

[0028] The second derivative reflects the acceleration of pressure change and is calculated using the finite difference method. This involves subtracting the first derivative from the first derivative at the previous moment from the current first derivative, then dividing by the sampling time interval to obtain the pressure acceleration, expressed in atmospheres per second squared. If the current first derivative is 0.2 and the first derivative at the previous second was 0.15, then the second derivative is 0.05, indicating that the pressure increase rate itself is still accelerating.

[0029] After calculation, a real-time state point is determined in a three-dimensional Cartesian coordinate system using the current pressure data as the X-axis coordinate, the pressure change rate as the Y-axis coordinate, and the pressure change acceleration as the Z-axis coordinate. This three-dimensional state space can simultaneously characterize the absolute value of pressure, its change trend, and the change of that trend. Different operating conditions exhibit obvious clustering characteristics in this space. During the pressure holding stage, the pressure is basically constant, and both the first and second derivatives are close to zero, with state points clustered near the X-axis. During the pressurization stage, the pressure continues to rise, the first derivative is positive and relatively stable, and the second derivative is close to zero, with state points distributed in the positive half-axis region of the Y-axis. During the depressurization stage, the pressure decreases slowly, the first derivative is negative, and state points are located in the negative half-axis region of the Y-axis. Through this three-dimensional mapping, complex time-series pressure data is transformed into spatial geometric features, providing an intuitive basis for identifying operating conditions.

[0030] Once the real-time state points in the 3D state space are calculated, it is necessary to determine their corresponding operating condition type. Before implementation, several target operating condition regions have been predefined, each corresponding to a typical operating condition, such as pressurization, pressure holding, and pressure reduction. Each region is described by its center point coordinates and boundary range. The edge PLC slave first calculates the Euclidean distance between the real-time state point and the center point of each target operating condition region. The distance formula is the straight-line distance between two points in 3D space. The target operating condition region with the smallest distance is selected as the candidate matching operating condition region. Then, it is further determined whether the real-time state point is truly located within the boundary range of this region. The boundary range can be a spherical region, defined by the center point and radius; or it can be a cuboid region, defined by the upper and lower limits of each coordinate axis. If the coordinates of the real-time state point satisfy the boundary constraints, the matching is confirmed as successful, and the operating condition ID corresponding to the operating condition region is output. The operating condition ID is an integer code, for example, 1 represents pressurization, 2 represents pressure holding, and 3 represents pressure reduction. If the real-time status point is not located within any predefined target operating condition area, it indicates that the current status is abnormal or in a transition period of operating condition change. In this case, it is marked as an abnormal operating condition, and a default operating condition ID (e.g., 0) is output, triggering an alarm mechanism to notify the operator. This method, based on geometric distance and boundary determination, achieves automatic identification of operating conditions, enabling real-time tracking of the stages of diving operations without manual intervention, and providing crucial contextual information for subsequent differentiated data processing.

[0031] After determining the current operating condition ID, the edge PLC slave station retrieves the corresponding physical association topology from the pre-stored configuration library based on the operating condition ID. The physical association topology is an abstract network structure that abstracts various sensors into mesh nodes. The connections between nodes reflect the constraints of physical laws and physiological metabolic patterns. For example, there is an ideal gas law relationship between pressure sensor nodes and temperature sensor nodes; a Dalton's law of partial pressure relationship between pressure sensor nodes and oxygen partial pressure sensor nodes; and a relationship between oxygen partial pressure and carbon dioxide partial pressure based on the respiratory metabolism of divers.

[0032] The strength and form of these physical relationships may differ under different operating conditions. Under pressurization conditions, the pressure rises rapidly, the temperature increases due to the work done by gas compression, and the gas partial pressure increases proportionally, resulting in a larger connection weight between nodes in the topology diagram. Under pressure-holding conditions, the pressure is constant, the temperature slowly dissipates and tends to stabilize, the partial pressure of oxygen decreases due to metabolic consumption, and the partial pressure of carbon dioxide increases due to metabolic production; the topology diagram reflects a metabolically dominant relationship pattern. By pre-constructing dedicated physical relationship topology diagrams for different operating conditions, the theoretical relationships between parameters under the current state can be described more accurately.

[0033] In the topology diagram, pressure sensors are typically defined as master nodes or anchor points because pressure is the most critical control variable in saturation diving operations; the theoretical values ​​of other sensors can be derived from pressure values ​​using physical formulas. Temperature sensors and various gas partial pressure sensors act as slave nodes, their theoretical values ​​constrained by the pressure nodes. This topology provides a clear computational path for subsequent calculations of theoretical rigid displacements.

[0034] After retrieving the physical correlation topology map, the edge PLC starts calculating the theoretical rigid displacement from the slave station. This is the theoretically expected coordinates that each sensor node should exhibit based on physical laws under the current pressure conditions. First, the main pressure sensor node is determined, and its current measured pressure value is obtained, for example, 25 atmospheres. Then, the ideal gas law is applied... Under the conditions of a constant volume V in a sealed chamber and a known number of gas moles n, a linear relationship between pressure P and temperature T can be derived. If the initial pressure is 1 atmosphere and the temperature is 20℃ (293K), the theoretically expected temperature should be [temperature value missing] when the current pressure rises to 25 atmospheres. However, the actual cabin has a heat exchange system, so the temperature will not rise to such a high level. Therefore, a correction factor is introduced in the actual calculation to take into account the heat conduction and heat dissipation effects, so as to obtain a more realistic theoretical expected temperature value, such as 30°C.

[0035] According to Dalton's law of partial pressures, the total pressure of a gas mixture is equal to the sum of the partial pressures of its components. If the gas inside the chamber is a helium-oxygen mixture with an oxygen mole fraction of 0.21, then at a total pressure of 25 atmospheres, the theoretically expected partial pressure of oxygen is: One atmosphere. The partial pressure of carbon dioxide needs to consider the diver's metabolic output. Based on the metabolic rate model, a certain amount of carbon dioxide is produced per minute. Combining this with the chamber volume and ventilation rate, the theoretically expected partial pressure of carbon dioxide can be calculated. These theoretically expected temperature and partial pressure values ​​are used as the theoretically expected coordinates of the corresponding sensor nodes in the topology diagram. The set of theoretically expected coordinates of all nodes constitutes the theoretical rigid displacement of the physical correlation topology diagram. This rigid displacement represents the state that each sensor reading should present under ideal physical conditions, providing a benchmark reference for subsequent deviation analysis.

[0036] After calculating the theoretical rigid displacement, the edge PLC obtains the measured node coordinates of each sensor node from the slave station, which are the actual measured values ​​of the sensors. The measured node coordinates are directly derived from the current readings of the sensors; for example, the measured value of the temperature sensor is 31°C, the measured value of the oxygen partial pressure sensor is 5.3 atmospheres, and the measured value of the carbon dioxide partial pressure sensor is 0.05 atmospheres. These measured values ​​are mapped to the grid nodes of the physical correlation topology graph, corresponding one-to-one with the theoretically expected coordinates.

[0037] The mapping process essentially converts physical measurements into node attributes in a topological graph, enabling subsequent comparisons and calculations within a unified topological framework. For example, the measured coordinates of a temperature sensor node are 31°C, while the theoretical expected coordinates are 30°C; the measured coordinates of an oxygen partial pressure sensor node are 5.3 atmospheres, while the theoretical expected coordinates are 5.25 atmospheres. This mapping integrates scattered sensor data into a topological mesh structure, providing a unified data representation for deformation analysis. The difference between the measured node coordinates and the theoretical expected coordinates reflects the degree of deviation of the actual environment from the ideal physical model. This deviation may stem from sensor errors, environmental disturbances, diver activities, or may indicate the occurrence of anomalies.

[0038] After obtaining the measured node coordinates and the theoretical expected coordinates, the edge PLC slave station calculates the deformation distance between them. The deformation distance is measured using Euclidean distance, which is the straight-line distance between the measured coordinates and the theoretical expected coordinates in numerical space. For the temperature node, the deformation distance is |31-30|=1℃; for the oxygen partial pressure node, the deformation distance is |5.3-5.25|=0.05 atmospheres. After calculating the deformation distance for each node, it is compared with a preset sensor allowable error threshold. The allowable error threshold is set according to the sensor accuracy and normal fluctuation range. For example, the allowable error for the temperature sensor is 0.5℃, and the allowable error for the oxygen partial pressure sensor is 0.02 atmospheres. If the deformation distance is less than or equal to the allowable error threshold, it means that the measured value of the node is within the normal range and highly consistent with the theoretical expectation. The node can be removed and not included in the construction of the deformation tensor, thereby reducing data redundancy. If the deformation distance is greater than the allowable error threshold, it means that the node has a significant deviation and needs to be retained for further analysis.

[0039] In the above example, the temperature node's deformation distance of 1℃ is greater than the allowable error of 0.5℃ and needs to be retained; the oxygen partial pressure node's deformation distance of 0.05 atmospheres is greater than the allowable error of 0.02 atmospheres and also needs to be retained. For the retained nodes, their deformation direction vector is calculated, which is the difference vector obtained by subtracting the theoretical expected coordinates from the measured coordinates. Then, the difference vector is normalized to obtain the unit deformation direction vector, indicating the direction of the deviation. The deformation amplitude is the deformation distance itself. Multiplying each component of the deformation direction vector by the deformation amplitude yields the deformation displacement vector, which fully describes the magnitude and direction of the deviation at that node. Arranging the deformation displacement vectors of all retained nodes by node index, a deformation displacement matrix is ​​constructed.

[0040] Furthermore, principal component analysis or singular value decomposition is performed on the deformation displacement matrix to extract eigenvalues ​​and eigenvectors corresponding to the main deformation modes. Based on these eigenvalues ​​and eigenvectors, the mesh deformation tensor is reconstructed. The mesh deformation tensor is a low-rank matrix that retains the main characteristics of the deformation and filters out secondary noise. This reduces the amount of data while preserving key information, laying the foundation for subsequent frequency domain conversion and data transmission.

[0041] After obtaining the mesh deformation tensor, the edge PLC slave station transforms it from the spatial domain to the frequency domain to further extract sparse features. A Fourier transform is performed on the mesh deformation tensor, decomposing the deformation information in the spatial domain into a series of sine and cosine components of different frequencies. The result of the Fourier transform is a complex matrix, where each element corresponds to a frequency component, containing amplitude and phase information. The amplitude of each frequency component, i.e., the modulus of the complex number, is calculated to obtain the amplitude spectrum. The amplitude spectrum reflects the contribution of each frequency component to the deformation tensor; typically, most of the energy is concentrated in a few low-frequency components, while the amplitudes of high-frequency components are very small and can be ignored.

[0042] The amplitude spectrum is sorted in descending order, and the top M frequency components with the largest amplitudes are selected as characteristic frequencies. M is a preset positive integer, such as 5 or 10, determined according to the required compression ratio and reconstruction accuracy. The amplitude corresponding to each characteristic frequency is extracted and normalized so that the sum of all amplitudes is 1, yielding characteristic frequency coefficients. These coefficients represent the relative importance of each characteristic frequency. Simultaneously, the phase angle corresponding to each characteristic frequency is extracted to construct a phase spectrum. The phase spectrum preserves the structural information of the deformation tensor in time or space, which is crucial for accurately reconstructing the original data.

[0043] To further compress the data, the feature frequency coefficients are encoded using floating-point compression, for example, compressing 64-bit double-precision floating-point numbers into 32-bit single-precision floating-point numbers, or using fixed-point representation, sacrificing a small amount of precision for a smaller data volume. The phase spectrum uses angle discretization encoding, quantizing continuous phase angles into a finite number of discrete values. For example, the phase range from 0 to 360 degrees is divided into 256 levels, with each phase angle represented by an 8-bit integer, significantly reducing the number of bits. Through frequency domain transformation and sparse feature extraction, the mesh deformation tensor is efficiently compressed into a small number of feature frequency coefficients and a phase spectrum, reducing the data volume by more than an order of magnitude compared to the original deformation tensor, while retaining the core information required for reconstruction.

[0044] After determining the characteristic frequency coefficients and phase spectrum, the edge PLC slave stations encapsulate this compressed data along with the current operating condition ID into standardized data frames for transmission to the surface monitoring master station via the industrial fieldbus. The data frame encapsulation follows a predefined protocol format. First, a data frame header is constructed, containing a frame synchronization identifier (typically a specific byte sequence like 0xAA55) to identify the start of the data frame at the receiving end. Next, a data frame length field is written, indicating the total number of bytes in the frame for correct parsing by the receiving end. Finally, a timestamp is written to record the precise time of data acquisition for timing alignment and traceability.

[0045] Following the data frame header, a current operating condition ID field is added, using fixed-length encoding, such as one byte, supporting up to 256 different operating condition types. Next is the characteristic frequency coefficient compressed data block, which uses Huffman coding. Huffman coding is a lossless compression algorithm that assigns different length codes based on the frequency of data occurrence—short codes for high-frequency data and long codes for low-frequency data—thus achieving overall compression. Next is the phase spectrum compressed data block. First, the phase spectrum is differentially encoded, representing the difference between adjacent phase angles. Since phase angles typically change gradually, the difference values ​​are small and easier to compress. Then, run-length encoding is used to represent consecutive identical difference values ​​as a single value plus a repetition count, further reducing the data size. The data frame is then concatenated in the following order: data frame header, current operating condition ID field, characteristic frequency coefficient compressed data block, and phase spectrum compressed data block, forming the main body of the data frame. Finally, a cyclic redundancy check (CRC) code is added to the end of the data frame. The entire data frame is checked, and the resulting check code is appended to the end. This code is used by the receiving end to verify the integrity and correctness of the data transmission. If the check fails, the data frame is discarded and a retransmission is requested.

[0046] After encapsulation, the data frame is sent to the water surface monitoring master station via the industrial fieldbus. The data frame size is usually only tens to hundreds of bytes, which is much smaller than the amount of data required to transmit all the original sensor data. This effectively reduces the bus load, avoids network congestion, and ensures the real-time performance and reliability of data transmission.

[0047] After receiving a data frame, the water surface monitoring master station first performs a cyclic redundancy check, recalculating the checksum of the data frame and comparing it with the received checksum. If they match, the verification is successful, indicating that the data transmission is error-free; if they do not match, the verification fails, the data frame is discarded, and a retransmission request is sent to the edge PLC slave station. After successful verification, the data frame header is parsed to extract the timestamp and data frame length, confirming the data's time attributes and integrity. The current operating condition ID is extracted from the data frame, and based on this ID, the physical association topology map and corresponding gas state equation parameters pre-stored in the master station database are retrieved. These parameters are completely consistent with the parameters used by the edge PLC slave station, ensuring calculation consistency.

[0048] Next, Huffman decoding is performed on the compressed data block of characteristic frequency coefficients to restore the original floating-point values ​​of the characteristic frequency coefficients. Run-length decoding is performed on the compressed data block of the phase spectrum to restore the difference value sequence, and then differential decoding is performed to accumulate the difference values ​​and restore the complete phase spectrum. Based on the restored characteristic frequency coefficients and phase spectrum, the complete frequency domain representation is reconstructed through inverse Fourier transform, that is, the amplitude and phase information of the M characteristic frequency components are combined to generate a complex matrix in the frequency domain. Then, the frequency domain representation is converted back to the spatial domain, and the mesh deformation tensor is restored through inverse Fourier transform. The restored mesh deformation tensor is highly consistent with the deformation tensor calculated by the edge PLC slave station in terms of value. The error mainly comes from the quantization error introduced by floating-point compression and phase discretization, but these errors are within an acceptable range and do not affect the validity of the data.

[0049] Based on the current operating condition ID and physical association topology, the surface monitoring master station recalculates the theoretical rigid displacement to obtain the theoretical expected coordinates of each sensor node. The calculation method is exactly the same as that of the edge PLC slave station, using the same physical formulas and parameters. The deformation displacement vector in the mesh deformation tensor is superimposed on the theoretical expected coordinates of the corresponding sensor node, i.e., the theoretical expected coordinates are added to the deformation displacement vector to obtain the restored coordinate value of each sensor. This restored coordinate value is the actual measured value of the sensor. The restored coordinate values ​​are arranged according to timestamps to generate the original time-series data of each sensor, completely reproducing all sensor readings collected by the edge PLC slave station. Through this data reconstruction method based on physical model and deformation superposition, the surface monitoring master station can completely recover the original monitoring data from highly compressed data frames, achieving both high efficiency in data transmission and ensuring data integrity and accuracy, providing reliable data support for the safety monitoring of diving operations.

[0050] This embodiment realizes intelligent edge computing compression transmission of saturated diving monitoring data. Three-dimensional state space mapping and automatic operating condition identification mechanisms enable adaptive adjustments to data processing strategies at different operational stages, avoiding the blindness of traditional fixed sampling methods. The introduction of physical correlation topology diagrams and theoretical rigid displacements fully utilizes the physical constraints between parameters in the saturated diving environment, transforming data transmission from complete raw data to deviation information transmission. During stable data periods such as pressure holding, the measured values ​​closely match theoretical values, resulting in extremely small deformation tensors and a significant reduction in transmitted data volume, effectively solving the redundant heartbeat packet problem and reducing bus load. The frequency domain sparse representation and characteristic frequency extraction of the mesh deformation tensor further improve compression efficiency. Even during periods of drastic operating condition changes such as depressurization and acceleration, frequency domain compression effectively controls data volume, avoiding network congestion and ensuring real-time transmission of critical data. The surface monitoring master station, based on a physical model and deformation superposition data reconstruction method, can losslessly recover the original monitoring data, ensuring data integrity and accuracy. The overall solution achieves high compression ratio, low latency, and high reliability data transmission, significantly improving the stability and safety of saturation diving monitoring and providing solid technical support for deep-sea operations.

[0051] In one embodiment of this invention, the first and second derivatives of the pressure data sequence are calculated in real time, and the current pressure data, the first and second derivatives are mapped to a three-dimensional state space, including the following steps: S210, the first derivative of the pressure data sequence is calculated using the finite difference method to obtain the pressure change rate; S220, the second derivative of the pressure data sequence is calculated using the quadratic difference method to obtain the pressure change acceleration; S230, real-time state points in the three-dimensional state space are constructed using the current pressure data as the X-axis coordinate, the pressure change rate as the Y-axis coordinate, and the pressure change acceleration as the Z-axis coordinate.

[0052] In practical implementation, edge PLC slave stations use the finite difference method to perform real-time differential calculations on the pressure data sequence to extract the dynamic characteristics of pressure changes. The finite difference method is a basic method of numerical differentiation and is suitable for calculating derivatives of discrete sampled data.

[0053] Specifically, the backward difference method is used to calculate the first derivative, which involves subtracting the pressure value from the previous pressure value from the current pressure value, and then dividing by the sampling time interval. For example, if the pressure at the current time t is P(t) = 20.5 atmospheres, and the pressure at the previous time t-1 is P(t-1) = 20.3 atmospheres, the sampling time interval... If the pressure changes in seconds, then the first derivative, i.e., the rate of pressure change, is... One atmosphere per second.

[0054] This positive value indicates that the pressure is rising at a rate of 0.2 atmospheres per second, consistent with the characteristics of a pressurization condition. A negative result indicates that the pressure is decreasing, corresponding to a depressurization condition; a result close to zero indicates that the pressure is basically stable, corresponding to a pressure-maintaining condition. By using the differential method, the discrete pressure data sequence is transformed into a continuous rate of pressure change, providing the first key feature dimension for condition identification.

[0055] After obtaining the pressure change rate, the edge PLC slave station further calculates the second derivative of the pressure data sequence using the quadratic difference method to obtain the acceleration information of the pressure change. The quadratic difference method is essentially a differential calculation of the first derivative, reflecting the change in the pressure change trend itself.

[0056] In the specific calculation, the first derivative at the current moment is subtracted from the first derivative at the previous moment, and then divided by the sampling time interval. For example, if the first derivative at the current moment is... Atm per second, the first derivative of the previous moment is If the pressure is at one atmosphere per second, then the second derivative, i.e., the acceleration due to pressure change, is: One atmosphere per second squared.

[0057] A positive value indicates that the pressure rise rate is still accelerating, possibly corresponding to the initial stage of pressurization. A negative second derivative indicates a slowing pressure rise rate, possibly corresponding to the later stage of pressurization, where the pressurization rate needs to be reduced as the target pressure approaches. A second derivative close to zero indicates a constant pressure change rate, corresponding to a uniform pressurization or depressurization stage, or a pressure holding stage where the pressure remains almost constant. The introduction of the second derivative enhances the finer characterization of pressure change patterns, making the dynamic characteristics of different operating conditions clearer, especially distinguishing between the transition and stabilization periods of operating condition shifts, providing richer information for accurate operating condition identification.

[0058] After calculating the three key parameters—current pressure data, pressure change rate, and pressure change acceleration—the edge PLC slave station maps them to a three-dimensional Cartesian coordinate system, constructing real-time state points in the three-dimensional state space.

[0059] Specifically, the current pressure data is used as the X-axis coordinate, reflecting the absolute value of the pressure; the pressure change rate is used as the Y-axis coordinate, reflecting the pressure change trend; and the pressure change acceleration is used as the Z-axis coordinate, reflecting the change in the pressure change trend. For example, if the current pressure is 20.5 atmospheres, the pressure change rate is 0.2 atmospheres per second, and the pressure change acceleration is 0.05 atmospheres per second squared, then the corresponding real-time state point coordinates in the three-dimensional state space are (20.5, 0.2, 0.05).

[0060] This three-dimensional state space can simultaneously characterize three levels of pressure information: position, velocity, and acceleration, similar to the phase space in classical mechanics that describes the motion of an object. Different operating conditions exhibit distinct clustering distribution characteristics in this three-dimensional space.

[0061] During the pressure holding phase, the pressure remains relatively constant near the target value. Both the rate of pressure change and acceleration are close to zero. Real-time status points cluster on the X-axis at a specific pressure value, forming a point cluster close to the origin. During the pressurization phase, the pressure continues to rise. The rate of pressure change is positive and relatively stable. The acceleration of pressure change is positive initially, close to zero in the middle stage, and negative in the later stage. Real-time status points move along a specific trajectory in three-dimensional space, progressing from the low-pressure region to the high-pressure region. The Y-axis coordinate remains positive, while the Z-axis coordinate is initially positive and then negative.

[0062] During the decompression phase, the pressure decreases slowly at a negative rate. The real-time state point is located in the region corresponding to the negative half of the Y-axis, moving along another trajectory from the high-pressure region to the low-pressure region. This three-dimensional mapping transforms complex time-series pressure data into intuitive spatial geometric features. Different operating conditions occupy different regions in space with clear boundaries, providing a solid mathematical foundation for subsequent operating condition identification based on geometric distance. This significantly simplifies the complexity of operating condition judgment and improves the real-time performance and accuracy of identification.

[0063] This implementation method expands a single pressure data sequence into a three-dimensional feature vector containing absolute pressure value, rate of change, and acceleration of change through stepwise calculations using the finite difference method and the quadratic difference method, fully mining the dynamic information contained in the pressure data. The construction of the three-dimensional state space transforms the time series analysis problem into a spatial geometric classification problem. Different operating conditions form clear clusters in the three-dimensional space, allowing operating condition identification to be achieved through simple geometric distance calculation and boundary determination, avoiding complex time series pattern matching algorithms, significantly reducing the computational burden on edge PLC slave stations, and improving real-time response speed. This method has a certain robustness to sampling noise; through smooth distribution in the three-dimensional space, random errors at a single sampling point will not lead to misjudgment of the operating condition.

[0064] In one embodiment of this example, when the coordinates of a point in the three-dimensional state space are located within a preset target working condition area, the current working condition ID is output, including the following steps: S310, obtaining the coordinates of a real-time state point in the three-dimensional state space; S320, calculating the distance between the real-time state point and the center point of each target working condition area based on the coordinates of the real-time state point; S330, selecting the target working condition area with the smallest distance as the matching working condition area; S340, determining whether the real-time state point is located within the boundary range of the matching working condition area; S350, if the real-time state point is located within the boundary range of the matching working condition area, outputting the working condition ID corresponding to the matching working condition area as the current working condition ID; S360, if the real-time state point is not located within the boundary range of any target working condition area, marking it as an abnormal working condition and outputting a default working condition ID.

[0065] In practical implementation, the edge PLC slave station first acquires the complete coordinate information of the real-time state point in the three-dimensional state space. This coordinate is calculated by the aforementioned steps and contains three components: the X-axis coordinate is the current pressure data, the Y-axis coordinate is the pressure change rate, and the Z-axis coordinate is the pressure change acceleration.

[0066] For example, at a certain moment, the coordinates of the real-time state point are: This indicates that the current pressure is 22.3 atmospheres, and the pressure is rising at a rate of 0.18 atmospheres per second, but the rate of increase is slowing down, with an acceleration of -0.02 atmospheres per second squared. This coordinate system fully describes the instantaneous state of the pressure system at the current moment.

[0067] The edge PLC slave station stores these three values ​​in memory variables as input data for subsequent operating condition identification calculations. The coordinate acquisition process needs to ensure data synchronization; that is, the three components must correspond to the state at the same moment to avoid positional deviations of state points due to inconsistent sampling timing. In engineering implementation, pressure acquisition, first derivative calculation, second derivative calculation, and coordinate construction are typically completed sequentially within one sampling cycle to ensure data temporal consistency.

[0068] After acquiring the real-time status point coordinates, the edge PLC slave station calculates the distance between the status point and the center point of each predefined target operating condition area. During the system initialization phase, several target operating condition areas have been predefined based on historical data and expert knowledge. Each area corresponds to a typical operating condition, such as pressurization, pressure holding, and pressure reduction.

[0069] Each target operating condition region is described by its center point coordinates and boundary parameters. The center point coordinates represent the typical state of that operating condition. For example, the center point of the pressure holding condition might be (25.0, 0.0, 0.0), indicating that the pressure is stable at 25 atmospheres, with both the rate of change and acceleration being zero; the center point of the pressurization condition might be (15.0, 0.3, 0.0), indicating that the pressure is around 15 atmospheres and is rising at a constant rate of 0.3 atmospheres per second.

[0070] The edge PLC slave traverses all predefined target operating areas. For each area, it calculates the Euclidean distance between the real-time status point and the center point of that area. The Euclidean distance formula is the straight-line distance between two points in three-dimensional space, i.e., ... ,in The coordinates are the center point coordinates of the target working area.

[0071] For example, if the real-time status point coordinates are (22.3, 0.18, -0.02), and the center point of a certain pressurized operating area is (20.0, 0.2, 0.0), then the distance is: .

[0072] The distances between the real-time status point and the center point of all target working condition areas are calculated sequentially to obtain a set of distance values. Each distance value reflects the similarity between the real-time status and the corresponding working condition, and the smaller the distance, the more similar they are.

[0073] After calculating the distances between the real-time status point and the center points of all target operating conditions, the edge PLC slave compares these distance values ​​and selects the target operating condition area with the smallest distance as the candidate matching operating condition area. The smallest distance means that the real-time status point is geometrically closest to the typical state of that operating condition area, and has the highest matching probability.

[0074] For example, if the calculated distance between the real-time status point and the pressurization condition area is 2.30, the distance to the pressure holding condition area is 5.80, and the distance to the depressurization condition area is 8.50, then the pressurization condition area is selected as the matching condition area. The selection process is achieved through simple numerical comparison, traversing all distance values ​​and recording the minimum value and its corresponding condition area index.

[0075] After selecting a matching working condition area, the ID of that working condition is not output immediately. Instead, it is necessary to further verify whether the real-time status point is truly within the effective boundary range of that working condition area. This is because the minimum distance does not guarantee that the status point belongs to that working condition. All working condition areas may be far apart, and the status point may be in the transition period of working condition change or in an abnormal state.

[0076] After selecting the matching operating condition region, the edge PLC slave station needs to determine whether the real-time status point is actually located within the boundary range of the matching operating condition region. The boundary range defines the effective coverage area of ​​the operating condition region in the three-dimensional state space, and only status points located within the boundary can be confirmed as belonging to that operating condition.

[0077] There are several ways to define boundary ranges, the most common being spherical boundaries, ellipsoidal boundaries, and cuboid boundaries. Spherical boundaries are the simplest, defined by a center point and a radius. The determination method is to calculate the distance between the state point and the center point; if the distance is less than or equal to the radius, the state point is within the boundary. For example, if the center point of the pressurized operating condition region is (20.0, 0.2, 0.0), and the radius is 3.0, and the distance between the real-time state point and the center point is 2.30, which is less than the radius 3.0, then the state point is determined to be within the boundary of this operating condition region.

[0078] An ellipsoidal boundary can more accurately characterize the shape of the working area, taking into account the differences in characteristic scales across different dimensions. The ellipsoidal boundary is defined by the covariance matrix, and the determination method is to calculate the Mahalanobis distance, i.e. ,in Let the coordinate vector of the state point be... The center point coordinate vector, The covariance matrix is ​​such that if the Mahalanobis distance is less than the threshold, it lies within the boundary.

[0079] The boundary of the cuboid is defined by the upper and lower limits of each coordinate axis. The determination method is to check whether each coordinate component of the state point is within the corresponding upper and lower limit range. For example, if the boundary of the pressurized working condition region is defined as the pressure range [18.0, 22.0], the velocity range [0.1, 0.3], and the acceleration range [-0.05, 0.05], and the real-time state point coordinates are (22.3, 0.18, -0.02), where the pressure of 22.3 exceeds the upper limit of 22.0, then it is determined that it is not within the boundary.

[0080] In practical applications, the appropriate boundary type is selected based on the characteristics of the operating conditions. For pressure-holding conditions, spherical or ellipsoidal boundaries are typically used because the state points are clustered near the center. For pressurization and depressurization conditions, cuboid boundaries may be used because the state points are distributed along a specific direction. The accuracy of boundary determination directly affects the reliability of operating condition identification. Boundaries that are too large can lead to misjudgments, confusing different operating conditions; boundaries that are too small can lead to missed judgments, marking normal operating conditions as abnormal.

[0081] If, after boundary determination, it is confirmed that the real-time status point is within the boundary range of the matching operating condition area, the edge PLC slave station outputs the operating condition ID corresponding to the matching operating condition area as the current operating condition ID. The operating condition ID is an integer code used to uniquely identify the operating condition type. For example, 1 represents rapid pressurization, 2 represents slow pressurization, 3 represents pressure holding, 4 represents slow depressurization, and 5 represents rapid depressurization, etc.

[0082] The output of the condition ID signifies the completion of condition identification at the current moment. This ID will serve as crucial contextual information, passing it to subsequent data processing modules to guide operations such as selecting the physical correlation topology and calculating theoretical rigid displacement. The update frequency of the condition ID is consistent with the sampling frequency of the pressure data, ensuring that condition identification can track the phase changes of the diving operation in real time.

[0083] During stable operating periods, the operating condition ID remains unchanged across multiple consecutive sampling cycles. During operating condition transitions, the operating condition ID changes, for example, from pressurization to pressure holding; this transition reflects the progress of the diving operation. To avoid frequent fluctuations in the operating condition ID due to random noise, a time lag mechanism can be introduced. This means that the switch only occurs after the new operating condition ID has been maintained for several consecutive sampling cycles, thus improving the stability of operating condition identification.

[0084] The output operating condition ID can also trigger corresponding monitoring strategy adjustments. For example, under rapid pressurization conditions, the data sampling frequency and transmission frequency can be encrypted to enhance the monitoring of pressure changes; under pressure holding conditions, the sampling frequency can be reduced to decrease the amount of data transmitted and optimize resource utilization.

[0085] If, after boundary determination, the real-time status point is found to be outside the boundary range of the matched operating condition area, or even outside the boundary range of any predefined target operating condition area, the edge PLC slave will mark the current status as an abnormal operating condition and output a default operating condition ID. The default operating condition ID is usually set to 0 or a special identifier value to represent an unknown or abnormal state.

[0086] There may be several reasons for this situation: First, the diving operation is in the transition period of the working condition change, and the real-time status point is located in the blank area between the two working condition areas; second, there is equipment failure or operation error, which causes the pressure change mode to deviate from the normal range; third, there is a special working mode that is not covered by the predefined working condition library.

[0087] The marking of abnormal operating conditions is of great significance for safety monitoring. Once an abnormal operating condition is detected, the edge PLC slave station will immediately trigger the alarm mechanism, reminding the on-site operators through the audible and visual alarm device. At the same time, it will send a high-priority abnormality report to the water surface monitoring master station, which includes the current complete status point coordinates, timestamp, and distance information from each operating condition area, so that the operators can quickly determine the cause of the abnormality and take countermeasures.

[0088] While outputting the default operating condition ID, the data processing strategy will switch to conservative mode, no longer relying on the physical model for data compression, but directly transmitting the raw sensor data to ensure data integrity under abnormal conditions and avoid loss of key information due to the failure of the compression algorithm's assumptions.

[0089] Abnormal operating condition data can also be used for continuous learning and updating of the operating condition database. By analyzing the distribution patterns of abnormal state points, new operating condition types can be discovered or the boundary parameters of existing operating condition areas can be adjusted to achieve adaptive optimization of the system.

[0090] This implementation achieves rapid and accurate automatic identification of operating conditions through a two-step decision-making strategy based on geometric distance. First, the distance between the real-time state point and the center points of all target operating condition regions is calculated, quickly narrowing down the candidate range and selecting the most likely matching operating condition region. This avoids the computational overhead of complex boundary determination for all operating condition regions, significantly improving identification efficiency. Then, precise boundary verification is performed on the candidate operating condition regions to ensure the reliability of the identification results and avoid potential misjudgments based solely on nearest neighbor distance. This method fully utilizes the clustering distribution characteristics of different operating conditions in the three-dimensional state space, achieving complex pattern recognition tasks through simple geometric calculations. The algorithm has low complexity and good real-time performance. The abnormal operating condition detection and marking mechanism enhances system security, enabling timely detection of equipment failures and operational anomalies, triggering alarms and protective measures, and effectively preventing risks associated with diving operations.

[0091] In one embodiment of this example, calculating the theoretical rigid displacement of the physical correlation topology to obtain the theoretical expected coordinates of each sensor node includes the following steps: S410, determining the main pressure sensor in the physical correlation topology and obtaining the current measured pressure value of the main pressure sensor; S420, calculating the theoretical expected temperature value of the temperature sensor node under the current measured pressure value according to a preset gas state equation; S430, calculating the theoretical expected partial pressure value of each gas partial pressure sensor node under the current measured pressure value according to a preset partial pressure formula; S440, using the theoretical expected temperature value and the theoretical expected partial pressure value as the theoretical expected coordinates of the corresponding sensor node in the physical correlation topology; S450, defining the set of theoretical expected coordinates of all sensor nodes as the theoretical rigid displacement of the physical correlation topology.

[0092] In practical implementation, after the edge PLC slave station retrieves the corresponding physical association topology diagram based on the current operating condition ID, it first needs to identify the main pressure sensor in the topology diagram and obtain its current measured pressure value. The main pressure sensor is the core sensor of the entire monitoring system, usually installed in a key location in the high-pressure chamber, and can most accurately reflect the overall pressure status inside the chamber.

[0093] In the physical correlation topology diagram, the main pressure sensor is defined as the anchor node or root node, and the theoretical expected values ​​of all other sensor nodes are derived from the measured values ​​of the main pressure sensor through physical laws. This pressure-based calculation architecture conforms to the actual situation of saturation diving operations because pressure is a direct reflection of diving depth and is the most critical control variable; all other environmental parameters have a clear physical correlation with pressure.

[0094] The edge PLC slave station reads the current measured value from the main pressure sensor via the industrial fieldbus. For example, if the pressure value reads as 24.8 atmospheres at a certain moment, this measured pressure value serves as the starting point and benchmark for all subsequent theoretical calculations, and its measurement accuracy directly affects the accuracy of the entire theoretical rigid displacement calculation. The acquired measured pressure value is temporarily stored in a memory variable as an input parameter for subsequent theoretical expected value calculations.

[0095] After acquiring the current measured pressure value from the main pressure sensor, the edge PLC slave station calculates the theoretical expected temperature value of the temperature sensor node based on a preset gas law equation. The gas law equation describes the relationship between the pressure, volume, temperature, and amount of substance of a gas. In the closed chamber environment of saturated diving, the ideal gas law equation... It provides a basic relationship between pressure and temperature.

[0096] Where P is the gas pressure, V is the chamber volume, n is the number of moles of gas, R is the ideal gas constant, and T is the absolute temperature. In a sealed chamber, the volume V is constant, and the number of moles of gas n remains essentially unchanged over a short period of time; therefore, pressure and temperature are directly proportional.

[0097] In practice, saturation diving chambers are equipped with temperature control systems that continuously dissipate heat through heat exchangers, maintaining the internal temperature within a suitable range. Therefore, in actual calculations, a thermodynamic correction model needs to be introduced to consider factors such as heat conduction through the chamber walls, the cooling capacity of the heat exchange system, and the specific heat capacity of the gas.

[0098] The revised temperature calculation formula can be ,in This is the pressure-temperature coefficient, reflecting the temperature rise caused by the work done during gas compression. The heat dissipation coefficient reflects the cooling effect of the heat exchange system. Let be the time interval from the initial state to the current moment. This modified model allows us to obtain a more realistic theoretically predicted temperature value.

[0099] The parameters of the temperature calculation model may differ under different operating conditions. For example, under rapid pressurization conditions, the pressure-induced temperature rise effect is significant. The value is relatively large; under pressure holding conditions, heat dissipation is the dominant effect. The values ​​are relatively large. These parameters are pre-stored in the physical association topology configuration corresponding to the operating condition and are automatically retrieved based on the current operating condition ID.

[0100] After calculating the theoretical expected value of the temperature sensor node, the edge PLC slave station further calculates the theoretical expected partial pressure value of each gas partial pressure sensor node according to the preset partial pressure formula. Saturated diving chambers are typically filled with a helium-oxygen mixture, mainly composed of helium and oxygen, with small amounts of carbon dioxide, nitrogen, and water vapor.

[0101] According to Dalton's law of partial pressures, the total pressure of a gas mixture is equal to the sum of the partial pressures of its components. Given the total pressure and the mole fraction of each component gas, the partial pressure of each component can be calculated.

[0102] The theoretical expected value of oxygen partial pressure needs to consider two factors: the initial oxygen mole fraction set during initial inflation and the oxygen consumption by the diver's respiratory metabolism. If the initial oxygen mole fraction is... Without considering metabolic consumption, the partial pressure of oxygen is One atmosphere.

[0103] However, in reality, as divers continuously consume oxygen, the oxygen mole fraction gradually decreases. Based on human metabolic models, the rate of decrease in oxygen partial pressure per unit time can be calculated. By accumulating the consumption from the initial moment to the current moment, the corrected theoretical expected oxygen partial pressure can be obtained.

[0104] The calculation of carbon dioxide partial pressure is the opposite. As divers breathe and metabolize, carbon dioxide is produced, and its partial pressure gradually increases. Simultaneously, the chamber is equipped with a carbon dioxide absorption system that continuously removes carbon dioxide using absorbents such as soda lime, maintaining its partial pressure within a safe range. Calculating the theoretically expected partial pressure of carbon dioxide requires comprehensively considering both the metabolic production rate and the absorption and removal rate, establishing a dynamic equilibrium model.

[0105] Helium, as an inert gas, does not participate in metabolism; its partial pressure changes are mainly caused by changes in total pressure and the partial pressure changes of other gases. This can be observed through… The calculations were performed based on physical laws and physiological metabolic models to obtain the theoretical expected partial pressure values ​​for each gas partial pressure sensor node.

[0106] After calculating the theoretical expected temperature value of the temperature sensor node and the theoretical expected partial pressure value of each gas partial pressure sensor node, the edge PLC slave station uses these theoretical expected values ​​as the theoretical expected coordinates of the corresponding sensor node in the physical association topology diagram.

[0107] The physical correlation topology is an abstract network structure where each sensor is represented as a node, and the coordinates of the node represent the sensor's measured or expected value. In the calculation of theoretical rigid displacement, the node coordinates are assigned theoretical expected values, reflecting the ideal state that each sensor should exhibit under current pressure conditions, based on physical laws and physiological models.

[0108] For example, the theoretical expected coordinates of the temperature sensor node are 28°C, the oxygen partial pressure sensor node is 5.15 atmospheres, and the carbon dioxide partial pressure sensor node is 0.04 atmospheres. These coordinate values ​​constitute an ideal configuration in the topology diagram, and the relationships between the nodes fully comply with the constraints of physical laws. The coordinates of the main pressure sensor node directly use its measured value of 24.8 atmospheres as the anchoring reference for the entire topology diagram.

[0109] In this way, the physical association topology diagram is transformed from an abstract structural template into a state representation with specific numerical values, providing a benchmark reference system for subsequent comparative analysis with measured values.

[0110] After obtaining the theoretical expected coordinates of all sensor nodes, the edge PLC slave defines the set of these coordinates as the theoretical rigid displacement of the physical association topology. The theoretical rigid displacement is a vector set containing the theoretical expected coordinate values ​​of each node in the topology, for example... .

[0111] This set of vectors fully describes the ideal environmental state derived from physical laws under the current pressure conditions. It is called "rigid displacement" because the relationship between these theoretically expected values ​​is rigid, that is, it strictly follows the constraints of physical laws and there are no arbitrary degrees of freedom.

[0112] The theoretical rigid displacement represents an idealized, undisturbed baseline state that perfectly conforms to the physical model, serving as the reference frame for subsequent deformation analysis. In real-world environments, due to factors such as sensor errors, environmental disturbances, and model simplification, measured values ​​often deviate from theoretically expected values. This deviation is known as deformation. By calculating the difference between the measured values ​​and the theoretical rigid displacement, we can extract the truly effective information that needs to be transmitted—the deviation information—rather than the complete raw data.

[0113] This implementation establishes a precise data benchmark reference system through theoretical rigid displacement calculation based on physical laws and physiological models. Using the measured value of the main pressure sensor as an anchor point, the theoretical expected values ​​of all other sensors are derived through the gas state equation and the law of partial pressures. This fully utilizes the physical correlations between parameters in a saturated diving environment, transforming redundant information from multi-dimensional sensor data into a predictable theoretical model. The introduction of theoretical rigid displacement transforms data transmission from complete raw data to deviation information transmission. Under normal operating conditions, the measured values ​​closely match the theoretical values ​​with minimal deviation, significantly reducing data transmission volume and effectively solving the problem of redundant heartbeat packets in traditional methods. This method not only reduces the data transmission burden but also enhances anomaly detection capabilities. When the measured value significantly deviates from the theoretical expected value, it indicates a possible equipment failure or environmental anomaly, triggering timely alarms.

[0114] In one embodiment of this example, the deformation distance between the measured node coordinates and the theoretical expected coordinates is calculated, and the mesh deformation tensor is determined based on the deformation distance. The steps include: S510, calculating the Euclidean distance between the measured node coordinates and the corresponding theoretical expected coordinates of each sensor node to obtain the deformation distance; S520, retaining sensor nodes whose deformation distance is greater than a preset sensor allowable error threshold, and removing sensor nodes whose deformation distance is less than or equal to the sensor allowable error threshold; S530, for the retained sensor nodes, calculating their deformation direction vector and deformation amplitude based on the deformation distance, and constructing the mesh deformation tensor based on the deformation direction vector and deformation amplitude.

[0115] In practical implementation, after obtaining the theoretical rigid displacement and measured node coordinates from the station, the edge PLC begins to calculate the deformation distance of each sensor node. The deformation distance is measured using Euclidean distance, which is the straight-line distance between the measured node coordinates and the theoretical expected coordinates in numerical space.

[0116] For each sensor node, the edge PLC reads the measured value from the station as the measured node coordinates, and simultaneously extracts the corresponding theoretical expected coordinates from the theoretical rigid displacement. Then, the absolute value of the difference between the two is calculated. For example, if the measured value of the temperature sensor is 29.2℃ and the theoretical expected value is 28℃, then the deformation distance is |29.2 - 28| = 1.2℃. If the measured value of the oxygen partial pressure sensor is 5.18 atmospheres and the theoretical expected value is 5.15 atmospheres, then the deformation distance is |5.18 - 5.15| = 0.03 atmospheres.

[0117] Deformation distance reflects the degree of deviation between actual measured values ​​and the expected values ​​of physical theory, and is a key indicator for quantifying the anomalies of sensor data. Under normal operating conditions, if all sensor data fully comply with the constraints of physical laws, the deformation distance should be close to zero, indicating only measurement errors within the sensor itself. If the deformation distance is large, it indicates that the actual environmental conditions deviate from the predictions of the theoretical model, possibly due to factors such as environmental disturbances, diver activities, or equipment malfunctions.

[0118] By calculating the deformation distance of each sensor node, the edge PLC slave station constructs a deformation distance vector, which fully describes the deviation distribution of the entire monitoring system at the current moment, providing a quantitative basis for subsequent data filtering and deformation tensor construction.

[0119] After calculating the deformation distance of all sensor nodes, the edge PLC slave station filters the data according to the preset sensor allowable error threshold, retaining sensor nodes with deformation distance greater than the threshold and removing sensor nodes with deformation distance less than or equal to the threshold.

[0120] The permissible error thresholds for sensors are set according to the accuracy specifications and normal fluctuation range of each type of sensor. These thresholds take into account the sensor's measurement accuracy, environmental noise level, and data compression requirements, aiming to avoid misinterpreting normal measurement errors as valid deformations while ensuring that true abnormal deviations can be captured.

[0121] In the above example, the deformation distance of the temperature sensor (1.2℃) is greater than the allowable error threshold of 0.5℃, so the node is retained; the deformation distance of the oxygen partial pressure sensor (0.03 atmospheres) is greater than the allowable error threshold of 0.02 atmospheres, so the node is retained; and the deformation distance of the carbon dioxide partial pressure sensor (0.001 atmospheres) is less than the allowable error threshold of 0.005 atmospheres, so the node is discarded.

[0122] Nodes that are removed indicate that their measured values ​​closely match the theoretical expected values, falling within the normal error range. Therefore, there is no need to transmit deviation information, and the data can be directly reconstructed at the receiving end using the theoretical expected values, thus achieving data compression. Nodes that are retained indicate significant deviations, requiring the deviation information to be encoded and transmitted so that the receiving end can accurately reconstruct the original data.

[0123] Through this threshold-based screening mechanism, during periods of stable data, the deformation distance of most sensor nodes is within the allowable error range, and most nodes are eliminated, resulting in a significant reduction in data transmission. During periods of drastic changes in operating conditions or under abnormal circumstances, the deformation distance of multiple sensor nodes exceeds the threshold, increasing the number of nodes retained and correspondingly increasing the data transmission volume, thus achieving an adaptive data compression strategy.

[0124] For the retained sensor nodes, the edge PLC slave station further calculates their deformation direction vector and deformation amplitude, and constructs a mesh deformation tensor based on this information. The deformation direction vector is obtained by calculating the difference vector between the measured node coordinates and the theoretically expected coordinates, i.e. ,in These are the measured coordinates. These are the theoretically expected coordinates.

[0125] Then, the difference vector is normalized to obtain the unit deformation direction vector. This indicates the direction of the deviation. The deformation amplitude is the deformation distance itself. Multiplying each component of the deformation direction vector by the deformation amplitude yields the deformation displacement vector. It fully describes the magnitude and direction of the deviation at that node.

[0126] For example, the deformation displacement vector of the temperature sensor node is [1.2]℃, and the deformation displacement vector of the oxygen partial pressure sensor node is [0.03] atmospheres. The deformation displacement vectors of all retained nodes are arranged by node index to construct a deformation displacement matrix. .

[0127] Furthermore, principal component analysis or singular value decomposition is performed on the deformation displacement matrix to extract eigenvalues ​​and eigenvectors corresponding to the main deformation modes. Principal component analysis can identify the main directions of change in the deformation data, projecting high-dimensional deformation information into a low-dimensional subspace to achieve dimensionality reduction and compression. By retaining the principal components corresponding to the first few largest eigenvalues, the entire deformation displacement matrix can be approximated with a small number of parameters, filtering out secondary noise and redundant information.

[0128] Based on the extracted feature values ​​and feature vectors, the mesh deformation tensor is reconstructed. This is a low-rank matrix that preserves the main features of the deformation while significantly compressing the data volume. The mesh deformation tensor not only contains the deviation information of each sensor node, but also implicitly contains the correlation structure between nodes, laying the foundation for subsequent frequency domain conversion and data transmission, and realizing efficient conversion from raw sensor data to compressed deformation representation.

[0129] This implementation method achieves precise quantification of sensor data deviation by calculating the deformation distance between the measured node coordinates and the theoretically expected coordinates. Based on a filtering mechanism using the sensor's allowable error threshold, it effectively distinguishes between normal measurement errors and actual environmental deviations, eliminating redundant data within the normal range and retaining only nodes that significantly deviate from theoretical expectations, thus achieving initial data compression. During periods of stable data, most sensor nodes are eliminated, significantly reducing data transmission volume and effectively solving the problem of redundant heartbeat packets. During periods of drastic changes in operating conditions, more nodes are retained, ensuring the complete transmission of key deviation information. The calculation of deformation direction vectors and deformation amplitudes provides a structured representation of deviation information, providing fundamental data for constructing the mesh deformation tensor. Principal component analysis extracts the main deformation modes, further compressing the data dimensionality and filtering out secondary noise. The generated low-rank mesh deformation tensor retains key deformation characteristics while significantly reducing the data volume.

[0130] In one embodiment of this example, the deformation direction vector and deformation amplitude are calculated based on the deformation distance, and a mesh deformation tensor is constructed based on the deformation direction vector and deformation amplitude, including the following steps: S610, for each retained sensor node, calculate the difference vector between the measured node coordinates and the theoretical expected coordinates; S620, normalize the difference vector to obtain a unit deformation direction vector; S630, use the deformation distance as the deformation amplitude; S640, multiply each component of the deformation direction vector by the deformation amplitude to obtain a deformation displacement vector; S650, arrange the deformation displacement vectors of all retained sensor nodes according to the node index to construct a deformation displacement matrix; S660, extract the eigenvalues ​​and eigenvectors corresponding to the main deformation modes in the deformation displacement matrix; S670, reconstruct the mesh deformation tensor based on the eigenvalues ​​and eigenvectors.

[0131] In practical implementation, for each selected and retained sensor node, the edge PLC slave station first calculates the difference vector between the measured node coordinates and the theoretical expected coordinates. The difference vector is a direct mathematical representation of the deviation between the measured value and the theoretical value. For a sensor with a single physical quantity, the difference vector degenerates into a scalar difference.

[0132] For example, the measured coordinates of the temperature sensor node are 29.2℃, the theoretical expected coordinates are 28℃, and the difference vector is... ℃. The measured coordinates of the oxygen partial pressure sensor node are 5.18 atmospheres, while the theoretically expected coordinates are 5.15 atmospheres. The difference vector is... One atmosphere.

[0133] For multi-dimensional sensors or combined measurement nodes, the difference vector may contain multiple components. The sign of the difference vector has important physical meaning: a positive value indicates that the measured value is higher than the theoretical expected value, and a negative value indicates that the measured value is lower than the theoretical expected value. By calculating the difference vector, the abstract concept of deviation is transformed into a concrete numerical representation, providing the basic data for subsequent vector operations and tensor construction.

[0134] After calculating the difference vector, the edge PLC slave station normalizes the difference vector to obtain the unit deformation direction vector. The normalization process divides the vector by its magnitude, making the length of the resulting vector equal to 1, thus retaining only the direction information.

[0135] For scalar differences, the normalized result is +1 or -1, indicating the positive or negative direction of the deviation. For example, the modulus of a temperature difference vector of 1.2℃ is 1.2, and after normalization, the result is a vector representing the direction of unit deformation. This indicates a positive deviation in temperature.

[0136] For multidimensional difference vectors The normalization formula is The sum of squares of the components of the resulting unit vector is 1. The introduction of the unit deformation direction vector decouples the magnitude and direction of the deviation. The direction information is represented by the unit vector, while the magnitude information is represented separately by the deformation amplitude. This separation facilitates subsequent data processing and compression encoding.

[0137] After obtaining the unit deformation direction vector, the edge PLC slave station directly uses the previously calculated deformation distance as the deformation amplitude. The deformation amplitude is the absolute size of the deviation, expressed in the units of the original physical quantity. For example, the deformation amplitude of the temperature deviation is 1.2℃, and the deformation amplitude of the oxygen partial pressure deviation is 0.03 atmospheres.

[0138] The deformation amplitude and deformation distance are numerically identical, both equal to the magnitude of the difference vector, i.e. The magnitude of deformation reflects the severity of the deviation; a larger magnitude indicates a greater deviation between the actual state and theoretical expectations, and a higher potential risk of anomalies. During data transmission and compression, the magnitude of deformation is crucial information that must be preserved, as it directly determines the accuracy of the reconstructed data.

[0139] After determining the unit deformation direction vector and deformation amplitude, the edge PLC slave multiplies each component of the deformation direction vector by the deformation amplitude to obtain the deformation displacement vector. The deformation displacement vector is a complete reconstruction of the difference vector, i.e. .

[0140] For scalar sensors, the deformation displacement vector degenerates into a scalar; for example, the deformation displacement vector of a temperature sensor is... At ℃, the deformation displacement vector of the oxygen partial pressure sensor is Atmospheric pressure. For multidimensional sensors, each component of the deformation displacement vector is equal to the corresponding unit direction vector component multiplied by the deformation amplitude, i.e. .

[0141] The deformation displacement vector is numerically equivalent to the original difference vector, but its construction process, through the separation of amplitude and direction, provides operational space for subsequent data compression. The deformation displacement vector fully describes the deviation state of each retained sensor node and is the basic element for constructing the deformation displacement matrix.

[0142] After calculating the deformation displacement vectors of all retained sensor nodes, the edge PLC slave station arranges these vectors according to the node index to construct a deformation displacement matrix. The node index is a predefined sensor number, ensuring the consistency and traceability of the matrix rows and columns.

[0143] Assuming that K sensor nodes are retained after screening, and the deformation displacement vector of each node is a scalar, then the deformation displacement matrix is ​​a K-dimensional column vector. For example, if we retain three nodes—temperature sensor, oxygen partial pressure sensor, and helium partial pressure sensor—with deformation displacement vectors of 1.2℃, 0.03 atmospheres, and 0.15 atmospheres respectively, then the deformation displacement matrix is: .

[0144] The deformation displacement matrix is ​​a set representation of the deviation information of all retained nodes, containing the complete deformation state of the monitoring system at the current moment. The sparsity of the matrix reflects the potential for data compression. During periods of stable data, the number of retained nodes K is small, the matrix dimension is low, and the data volume is small. During periods of drastic changes in operating conditions, K increases, and the matrix dimension increases, but significant compression is still achieved compared to transmitting all the original sensor data.

[0145] After constructing the deformation displacement matrix, the edge PLC slave station decomposes the matrix to extract the eigenvalues ​​and eigenvectors corresponding to the main deformation modes. This step uses singular value decomposition or principal component analysis to decompose the deformation displacement matrix into a linear combination of several orthogonal basis vectors. Each basis vector corresponds to a deformation mode, and its coefficients, i.e., eigenvalues, reflect the contribution of that mode.

[0146] Singular value decomposition transforms the matrix Decomposed into ,in and It is an orthogonal matrix. It is a diagonal matrix, and the diagonal elements are the singular values. The eigenvalues ​​are arranged in descending order, and the first few largest eigenvalues ​​correspond to the main deformation modes, contributing most of the energy of the deformation displacement matrix.

[0147] By retaining only the first L largest eigenvalues ​​and their corresponding eigenvectors, the original deformation displacement matrix can be approximated using a low-rank approximation, i.e. This approach achieves dimensionality reduction and compression of the data. The feature vectors describe the spatial distribution patterns of deformation modes. The extracted feature values ​​and feature vectors not only compress the data volume but also reveal the intrinsic structure and correlation patterns of the deformation data, providing valuable information for anomaly pattern recognition and fault diagnosis.

[0148] After extracting eigenvalues ​​and eigenvectors, the edge PLC slave station reconstructs the mesh deformation tensor based on this feature information. The mesh deformation tensor is a low-rank approximation of the deformation displacement matrix. By preserving the main deformation modes and filtering out secondary noise, efficient data compression is achieved.

[0149] The reconstruction formula is , where L is the number of principal modes retained. The reconstructed mesh deformation tensor Numerically, it is approximately equal to the original deformation displacement matrix. However, its representation is more compact, requiring only the storage of L feature values ​​and their corresponding feature vectors.

[0150] The reconstructed mesh deformation tensor retains the main characteristics of the deformation data, and the reconstruction error is controlled within an acceptable range. Typically, the reconstruction error is required to be less than the sensor's allowable error threshold to ensure that the accuracy of the reconstructed data meets monitoring requirements. The mesh deformation tensor is the core data object for subsequent frequency domain conversion and data transmission. Its completion marks the end of the spatial domain data compression stage, laying the foundation for entering the frequency domain compression stage.

[0151] This implementation method achieves the transformation from discrete sensor deviation data to a structured deformation representation through systematic deformation displacement vector calculation and matrix construction. The calculation of the difference vector quantifies the deviation between measured and theoretical values, and normalization decouples the deviation direction and amplitude, enabling subsequent separation and encoding. By extracting the main deformation modes through singular value decomposition, the low-rank characteristics and spatial correlation of deformation data are fully utilized, projecting high-dimensional deformation information onto a low-dimensional subspace, achieving significant dimensionality reduction and compression while filtering out secondary noise and improving the signal-to-noise ratio. The mesh deformation tensor reconstructed based on eigenvalues ​​and eigenvectors represents the main features of the deformation displacement matrix in a compact form, significantly reducing the data volume while maintaining reconstruction accuracy.

[0152] In one embodiment of this example, the characteristic frequency coefficients and phase spectrum are determined based on the mesh deformation tensor, including the following steps: S710, performing a Fourier transform on the mesh deformation tensor to convert the deformation information in the spatial domain to the frequency domain; S720, calculating the amplitude of each frequency component in the frequency domain to obtain the amplitude spectrum; S730, sorting the amplitude spectrum in descending order and selecting the top M frequency components with the largest amplitudes as characteristic frequencies, where M is a positive integer; S740, extracting the amplitude corresponding to each characteristic frequency and normalizing the amplitude to obtain the characteristic frequency coefficients; S750, extracting the phase angle corresponding to each characteristic frequency to construct the phase spectrum, wherein the characteristic frequency coefficients are encoded using floating-point compression and the phase spectrum is encoded using angle discretization.

[0153] In practical implementation, after constructing the mesh deformation tensor, the edge PLC slave performs a Fourier transform on it to convert the deformation information in the spatial domain to the frequency domain. The Fourier transform is a mathematical transformation method that can decompose signals in the time or spatial domain into a superposition of sine and cosine components of different frequencies, revealing the frequency structure characteristics of the signal.

[0154] For mesh deformation tensor The mathematical expression of the Fourier transform is: ,in This represents the Fourier transform operator. This is represented in the frequency domain. In the discrete case, it is implemented using the Discrete Fourier Transform or Fast Fourier Transform algorithm. For example, if the mesh deformation tensor is a K-dimensional vector... Then its discrete Fourier transform is ,in This is the frequency index, where i is the imaginary unit.

[0155] Transformation result It is a complex vector, each element Each frequency component contains amplitude and phase information. The physical meaning of the frequency domain representation is that the spatial distribution pattern in the deformation tensor can be decomposed into fluctuation components of different spatial frequencies. The low-frequency components correspond to the overall deformation trend on a large scale, while the high-frequency components correspond to local fluctuations and noise on a small scale.

[0156] In saturated submersible monitoring, deformation under normal operating conditions typically exhibits low-frequency characteristics because environmental parameters change relatively smoothly, and the deviations between sensors have spatial continuity. Under abnormal operating conditions, high-frequency components may appear, reflecting abrupt changes or local anomalies. By using Fourier transform to convert the deformation tensor from the spatial domain to the frequency domain, a mathematical basis is provided for subsequent frequency-selective compression, enabling efficient compression by utilizing the frequency domain sparsity characteristics of the deformation signal.

[0157] After completing the Fourier transform, the edge PLC slave station calculates the amplitude of each frequency component in the frequency domain, obtaining the amplitude spectrum. Frequency components It is a complex number, which can be represented as ,in For amplitude, This is the phase angle. The amplitude calculation formula is: ,in and These are the real and imaginary parts of a complex number, respectively.

[0158] Amplitude spectrum The amplitude spectrum describes the contribution of each frequency component to the deformation tensor; a larger amplitude indicates higher energy and a more significant impact on deformation. The distribution characteristics of the amplitude spectrum reflect the frequency structure of the deformation signal. Under normal operating conditions, the amplitude spectrum typically exhibits energy concentration in the low-frequency band, with rapid amplitude decay in the high-frequency band. This frequency domain sparsity is key to achieving a high compression ratio. By analyzing the amplitude spectrum, the main frequency components can be identified, providing a basis for subsequent characteristic frequency selection.

[0159] After obtaining the amplitude spectrum, the edge PLC slave station sorts the amplitude spectrum in descending order and selects the top M frequency components with the largest amplitudes as characteristic frequencies. The descending order is achieved through a sorting algorithm, which sorts all frequency components from largest to smallest amplitude to obtain a sorted frequency index sequence.

[0160] For example, if the amplitude spectrum is [0.1, 0.8, 0.05, 0.6, 0.02], the descending index sequence is [1, 3, 0, 2, 4], and the corresponding amplitude sequence is [0.8, 0.6, 0.1, 0.05, 0.02]. The first M frequency components are selected, where M is a preset positive integer determined according to the compression ratio requirements and reconstruction accuracy requirements.

[0161] The selection of characteristic frequencies is based on the principle of energy concentration, meaning that a few key frequency components contain most of the signal's energy. By retaining only these key components, the original signal can be approximately reconstructed with far less information than the original data. The choice of the value of M requires a trade-off between compression ratio and reconstruction accuracy. A larger M results in higher reconstruction accuracy but a lower compression ratio; a smaller M results in a higher compression ratio but may lose important information. Through characteristic frequency selection, the complete frequency domain representation containing K frequency components is compressed into a sparse representation containing only M characteristic frequencies, reducing the data volume to M / K of the original, achieving significant frequency domain compression.

[0162] After selecting the characteristic frequencies, the edge PLC extracts the amplitude corresponding to each characteristic frequency from the slave station and normalizes the amplitude to obtain the characteristic frequency coefficients. Normalization involves dividing the amplitude of all characteristic frequencies by their sum, such that the sum of the normalized coefficients is 1. ,in Let be the original amplitude of the i-th characteristic frequency. These are the normalized characteristic frequency coefficients.

[0163] The purpose of normalization is to convert amplitude information into a relative proportion, which facilitates subsequent encoding and transmission, and also facilitates reconstruction calculations at the receiving end.

[0164] The characteristic frequency coefficients reflect the relative importance of each characteristic frequency; a larger coefficient indicates a greater contribution of that frequency component to the deformation. To further compress the data, the characteristic frequency coefficients employ floating-point compression encoding, for example, compressing a 64-bit double-precision floating-point number into a 32-bit single-precision floating-point number, sacrificing some numerical precision for a smaller data volume. Through normalization and compression encoding, the characteristic frequency coefficients are efficiently represented, laying the foundation for constructing compact data frames.

[0165] While extracting the characteristic frequency coefficients, the edge PLC slave station extracts the phase angle corresponding to each characteristic frequency to construct a phase spectrum. It can be calculated using the arctangent function of a complex number, i.e. The result is arrive The radian value, or converted to an angle value between 0 and 360 degrees.

[0166] For example, if the complex representation of a certain characteristic frequency component is: Then the phase angle is Phase spectrum It contains phase information of M characteristic frequencies. The phase angle describes the time or spatial offset of each frequency component and is crucial for accurate signal reconstruction.

[0167] While amplitude information determines the energy level of each frequency component, phase information determines how they are superimposed. The same amplitude spectrum combined with different phase spectra can reconstruct completely different signals. Therefore, the phase spectrum must be preserved and transmitted.

[0168] To compress the data volume of the phase spectrum, angle discretization encoding is employed, which quantizes continuous phase angles into a finite number of discrete values. For example, the phase range from 0 to 360 degrees is divided into 256 levels, each level corresponding to an angle interval of 360 / 256 = 1.406°. Each phase angle is quantized to the nearest discrete level and represented by an 8-bit integer. Through angle discretization, each phase angle requires only one byte of storage, significantly reducing the data volume. The combination of phase spectrum compression encoding and characteristic frequency coefficient compression encoding achieves efficient representation of frequency domain information, providing a compact data source for the final data frame encapsulation and transmission.

[0169] This implementation transforms the mesh deformation tensor from the spatial domain to the frequency domain using Fourier transform, fully utilizing the frequency-domain sparsity of the deformation signal and providing a mathematical foundation for frequency-selective compression. The calculation of the amplitude spectrum reveals the energy distribution of each frequency component. By arranging them in descending order and selecting characteristic frequencies, only a few major frequency components are retained, achieving significant data dimensionality reduction. Normalization and floating-point compression encoding of the characteristic frequency coefficients convert the amplitude information into a compact relative weight representation, further reducing the number of data bits. Angle discretization encoding of the phase spectrum, while ensuring reconstruction accuracy, quantizes continuous phase information into discrete integer representations, requiring only 1 byte of storage for each phase angle, significantly reducing the data volume. This frequency-domain compression method not only achieves a high compression ratio but also preserves the main features and structural information of the signal, ensuring that the receiver can accurately reconstruct the original deformation tensor.

[0170] In one embodiment of this example, the current operating condition ID, characteristic frequency coefficients, and phase spectrum are compressed and encapsulated into a data frame, including the following steps: S810, constructing a data frame header, writing a frame synchronization identifier, data frame length, and timestamp into the data frame header; S820, adding a current operating condition ID field after the data frame header, and using fixed-length encoding; S830, performing Huffman coding on the characteristic frequency coefficients to generate a compressed data block of characteristic frequency coefficients; S840, performing differential coding and run-length encoding on the phase spectrum to generate a compressed data block of phase spectrum; S850, concatenating the data frame header, current operating condition ID field, compressed data block of characteristic frequency coefficients, and compressed data block of phase spectrum in that order; S860, adding a cyclic redundancy check code to the end of the data frame to complete the encapsulation of the data frame.

[0171] In practice, the edge PLC slave station first constructs the data frame header, which serves as the start identifier and carrier of metadata for the data frame. The data frame header contains three key fields: frame synchronization identifier, data frame length, and timestamp.

[0172] The frame synchronization identifier is a predefined special byte sequence used by the receiver to identify the start position of a data frame. It typically employs a specific pattern unlikely to appear in normal data, such as 0xAA55 or 0xFF00FF00. The design of the frame synchronization identifier needs to consider interference resistance to avoid misidentification due to noise or errors during data transmission.

[0173] The data frame length field indicates the total number of bytes in the entire data frame, including the header, data payload, and checksum, and is typically represented as a 16-bit or 32-bit unsigned integer. Accurate recording of the data frame length allows the receiving end to correctly parse the data frame boundaries, avoiding data packet fragmentation or splicing issues.

[0174] The timestamp field records the precise time of data acquisition, using the Unix timestamp format, typically represented by a 32-bit or 64-bit integer. The standardized header design ensures the consistency and parsability of the data frame format, providing a reliable protocol foundation for subsequent data transmission and reception processing.

[0175] After constructing the data frame header, the edge PLC adds a current condition ID field after the station header, using a fixed-length encoding. The current condition ID is an integer code output from the aforementioned condition identification step, used to identify the current condition stage of the diving operation. For example, 1 represents rapid pressurization, 2 represents slow pressurization, 3 represents pressure holding, 4 represents slow depressurization, 5 represents rapid depressurization, and 0 represents abnormal conditions, etc.

[0176] The range of values ​​for operating condition IDs is usually small, generally not exceeding 256 different operating condition types. Therefore, they can be represented by 1 byte, with a value range of 0 to 255. Fixed-length encoding means that regardless of the specific value of the operating condition ID, it occupies the same number of bytes, simplifying the encoding and decoding logic and improving processing efficiency.

[0177] The operational condition ID field follows immediately after the data frame header, located at the very beginning of the data payload. This placement allows the receiver to obtain operational condition information first, and then retrieve the corresponding physical topology map and decoding parameters based on the operational condition ID, preparing for subsequent data reconstruction. The operational condition ID is crucial contextual information for the entire data compression and reconstruction scheme, and its accurate transmission is essential. Therefore, a simple and reliable fixed-length encoding method is adopted to avoid reconstruction failures caused by encoding errors.

[0178] After adding the operating condition ID field, the edge PLC slave station performs Huffman coding on the characteristic frequency coefficients to generate compressed data blocks of the characteristic frequency coefficients. Huffman coding is a lossless compression algorithm based on the frequency of symbol occurrence, assigning short codes to high-frequency symbols and long codes to low-frequency symbols, thereby achieving overall compression.

[0179] Before applying Huffman coding, the feature frequency coefficients need to be quantized into discrete symbols. For example, the normalized feature frequency coefficients can be quantized into 256 levels, with each coefficient mapped to an integer between 0 and 255. Quantizing M feature frequency coefficients yields M integer symbols. Then, the frequency of these symbols is counted to construct a Huffman tree.

[0180] In saturated submersible monitoring applications, the distribution of characteristic frequency coefficients typically exhibits a certain regularity, resulting in some symbols appearing more frequently than others. Huffman coding, based on this frequency distribution, assigns shorter bit sequences to high-frequency symbols and longer bit sequences to low-frequency symbols. All quantized symbols of the characteristic frequency coefficients are sequentially Huffman encoded and concatenated into a bit stream, forming a compressed data block of the characteristic frequency coefficients.

[0181] After generating the characteristic frequency coefficient compressed data block, the edge PLC slave station performs differential encoding and run-length encoding on the phase spectrum to generate a phase spectrum compressed data block. The phase spectrum contains phase angles of M characteristic frequencies, each of which has been discretized and encoded into an 8-bit integer.

[0182] Differential coding is a predictive coding method that uses the correlation between adjacent data for compression. Specifically, differential coding calculates the difference between adjacent phase angles instead of directly transmitting the absolute value. The first phase angle is transmitted directly, and subsequent phase angles transmit the difference from the previous phase angle. For example, if the phase spectrum is [38, 42, 45, 44, 46], it will be [38, 4, 3, -1, 2] after differential coding.

[0183] Because the phase angle typically changes gradually and the difference between adjacent phase angles is small, the numerical range after differential coding is significantly reduced, and the number of bits required for coding is greatly decreased. Run-length coding is then applied after differential coding. Run-length coding is a compression method for consecutively repeating data, representing consecutive identical values ​​as a single value plus the number of repetitions.

[0184] For example, if the sequence after differential encoding is [38, 2, 2, 2, 3, 3, 1], the sequence after run-length encoding is [38×1, 2×3, 3×2, 1×1]. Combining differential encoding and run-length encoding fully utilizes the temporal correlation and local stationarity of the phase spectrum, achieving efficient compression.

[0185] After generating the characteristic frequency coefficient compressed data block and the phase spectrum compressed data block, the edge PLC slave station splices them together in a predefined order to construct the complete data payload. The splicing order is: data frame header, current operating condition ID field, characteristic frequency coefficient compressed data block, and phase spectrum compressed data block.

[0186] This sequential arrangement follows the principles of logical data dependencies and ease of parsing. The data frame header is located at the very beginning and contains a frame synchronization identifier, data frame length, and timestamp, enabling the receiver to first identify the start position of the data frame, determine the total length of the data frame, and obtain the time attribute of the data. The current operating condition ID field follows immediately, using fixed-length encoding, allowing the receiver to quickly read the operating condition information and prepare the corresponding parameters and models for subsequent data decoding and reconstruction.

[0187] The characteristic frequency coefficient compressed data block, located after the operating condition ID, contains frequency-domain compressed amplitude information and employs Huffman coding. It requires decoding according to a predefined or transmitted encoding table. The phase spectrum compressed data block, located last, contains frequency-domain compressed phase information and employs differential and run-length encoding. The decoding process requires gradually reconstructing the differential values ​​and repeating sequences. Standardized splicing order and structural design ensure the consistency and scalability of the data frame format.

[0188] After splicing the data payload, the edge PLC adds a cyclic redundancy check (CRC) code to the end of the data frame, completing the data frame encapsulation. Cyclic redundancy check is a widely used error detection method that generates a fixed-length check code by performing polynomial division on all bytes of the data frame, which is then appended to the end of the data frame.

[0189] The receiving end performs the same CRC calculation on the received data frame and compares the calculation result with the received check code. If they match, the verification is successful, indicating that the data transmission is error-free; if they do not match, the verification fails, indicating that an error occurred during data transmission, and the data frame needs to be discarded and a retransmission requested.

[0190] In saturation diving monitoring applications, CRC-16 is typically used to generate a 2-byte checksum, which can detect the vast majority of single-bit errors, double-bit errors, and burst errors, resulting in high reliability. Adding a CRC checksum significantly improves the reliability of data transmission, enabling timely detection and handling of transmission errors and preventing monitoring failures or security risks caused by the misuse of erroneous data.

[0191] After the CRC checksum is added, the data frame is encapsulated, forming a complete, standard-formatted data packet with error detection capabilities, ready to be sent to the surface monitoring master station via the industrial fieldbus. The encapsulated data frame is typically between tens and hundreds of bytes long. Compared to transmitting all raw sensor data, this achieves orders-of-magnitude data compression, significantly reducing bus bandwidth usage and improving the real-time performance and efficiency of data transmission.

[0192] This implementation achieves efficient organization and reliable transmission of compressed data through a standardized data frame encapsulation process. The data frame header design includes a frame synchronization identifier, data frame length, and timestamp, providing the receiving end with fundamental information for data frame identification, boundary resolution, and timing alignment, ensuring the robustness of the data transmission protocol. Huffman coding of the characteristic frequency coefficients fully utilizes the non-uniformity of symbol frequency distribution to achieve lossless compression, further reducing the data volume. The combination of differential coding and run-length coding of the phase spectrum fully utilizes the temporal correlation and local stationarity of the phase angle, achieving efficient compression while preserving the complete phase information required for reconstruction. The addition of cyclic redundancy check (CRC) significantly improves the reliability of data transmission, enabling timely detection of transmission errors, triggering retransmission mechanisms, and avoiding monitoring errors caused by erroneous data.

[0193] This application also provides a machine-readable storage medium storing instructions that cause a machine to execute the aforementioned edge computing compression transmission method for diving monitoring data.

[0194] Referring to Figure 2, this application embodiment also provides a saturation diving monitoring system, including: at least one edge PLC slave station; and a surface monitoring master station connected to each edge PLC slave station.

[0195] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0196] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0197] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0198] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0199] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0200] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0201] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0202] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0203] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. An edge computing compression transmission method for underwater monitoring data, characterized in that, This method is applied to a saturated diving monitoring system, which includes at least one edge PLC slave station and a surface monitoring master station. The method includes: acquiring diving environment parameters via the edge PLC slave station, including current pressure data, pressure data sequence, temperature, and gas partial pressure; calculating the first and second derivatives of the pressure data sequence in real time, and mapping the current pressure data, first and second derivatives to a three-dimensional state space; outputting the current operating condition ID when the coordinates of the three-dimensional state space are located in a preset target operating condition area; retrieving the corresponding physical association topology graph based on the current operating condition ID, defining the pressure sensor, temperature sensor, and gas partial pressure sensor as grid nodes in the physical association topology graph; and calculating the physical association topology... The theoretical rigid displacement of the PTZ is used to obtain the theoretical expected coordinates of each sensor node; the measured node coordinates of each sensor node are obtained and mapped to the mesh nodes; the deformation distance between the measured node coordinates and the theoretical expected coordinates is calculated, and the mesh deformation tensor is determined based on the deformation distance; based on the mesh deformation tensor, the characteristic frequency coefficients and phase spectrum are determined; the current operating condition ID, characteristic frequency coefficients, and phase spectrum are compressed and encapsulated into a data frame, and the data frame is sent to the water surface monitoring master station. The water surface monitoring master station is used to invert and restore the mesh deformation tensor based on the characteristic frequency coefficients and phase spectrum, generate the theoretical rigid displacement based on the current operating condition ID, and superimpose the mesh deformation tensor onto the theoretical rigid displacement to restore the original time series data of each sensor.

2. The method according to claim 1, characterized in that, The first and second derivatives of the pressure data sequence are calculated in real time, and the current pressure data, first and second derivatives are mapped to a three-dimensional state space. This includes: calculating the first derivative of the pressure data sequence using the finite difference method to obtain the pressure change rate; calculating the second derivative of the pressure data sequence using the quadratic difference method to obtain the pressure change acceleration; and constructing real-time state points in the three-dimensional state space with the current pressure data as the X-axis coordinate, the pressure change rate as the Y-axis coordinate, and the pressure change acceleration as the Z-axis coordinate.

3. The method according to claim 1, characterized in that, When the coordinates of a point in the 3D state space are located within a preset target working condition area, the current working condition ID is output, including: obtaining the coordinates of the real-time state point in the 3D state space; calculating the distance between the real-time state point and the center point of each target working condition area based on the coordinates of the real-time state point; selecting the target working condition area with the smallest distance as the matching working condition area; determining whether the real-time state point is within the boundary range of the matching working condition area; if the real-time state point is within the boundary range of the matching working condition area, the working condition ID corresponding to the matching working condition area is output as the current working condition ID; if the real-time state point is not within the boundary range of any target working condition area, it is marked as an abnormal working condition, and a default working condition ID is output.

4. The method according to claim 1, characterized in that, The theoretical rigid displacement of the physical correlation topology is calculated to obtain the theoretical expected coordinates of each sensor node. This includes: determining the main pressure sensor in the physical correlation topology and obtaining the current measured pressure value of the main pressure sensor; calculating the theoretical expected temperature value of the temperature sensor node at the current measured pressure value according to the preset gas state equation; calculating the theoretical expected partial pressure value of each gas partial pressure sensor node at the current measured pressure value according to the preset partial pressure formula; using the theoretical expected temperature value and the theoretical expected partial pressure value as the theoretical expected coordinates of the corresponding sensor node in the physical correlation topology; and defining the set of theoretical expected coordinates of all sensor nodes as the theoretical rigid displacement of the physical correlation topology.

5. The method according to claim 1, characterized in that, The deformation distance between the measured node coordinates and the theoretical expected coordinates is calculated, and the mesh deformation tensor is determined based on the deformation distance. This includes: calculating the Euclidean distance between the measured node coordinates and the corresponding theoretical expected coordinates of each sensor node to obtain the deformation distance; retaining sensor nodes whose deformation distance is greater than a preset sensor allowable error threshold and removing sensor nodes whose deformation distance is less than or equal to the sensor allowable error threshold; for the retained sensor nodes, calculating their deformation direction vector and deformation amplitude based on the deformation distance, and constructing the mesh deformation tensor based on the deformation direction vector and deformation amplitude.

6. The method according to claim 5, characterized in that, The deformation direction vector and deformation amplitude are calculated based on the deformation distance, and a mesh deformation tensor is constructed based on the deformation direction vector and deformation amplitude. This includes: for each retained sensor node, calculating the difference vector between the measured node coordinates and the theoretical expected coordinates; normalizing the difference vector to obtain the unit deformation direction vector; using the deformation distance as the deformation amplitude; multiplying each component of the deformation direction vector by the deformation amplitude to obtain the deformation displacement vector; arranging the deformation displacement vectors of all retained sensor nodes according to the node index to construct a deformation displacement matrix; extracting the eigenvalues ​​and eigenvectors corresponding to the main deformation modes in the deformation displacement matrix; and reconstructing the mesh deformation tensor based on the eigenvalues ​​and eigenvectors.

7. The method according to claim 1, characterized in that, Based on the mesh deformation tensor, the characteristic frequency coefficients and phase spectrum are determined, including: performing a Fourier transform on the mesh deformation tensor to convert the deformation information in the spatial domain to the frequency domain; calculating the amplitude of each frequency component in the frequency domain to obtain the amplitude spectrum; sorting the amplitude spectrum in descending order and selecting the top M frequency components with the largest amplitudes as characteristic frequencies, where M is a positive integer; extracting the amplitude corresponding to each characteristic frequency and normalizing the amplitude to obtain the characteristic frequency coefficients; and extracting the phase angle corresponding to each characteristic frequency to construct the phase spectrum. The characteristic frequency coefficients are encoded using floating-point compression, and the phase spectrum is encoded using angle discretization.

8. The method according to claim 1, characterized in that, The current operating condition ID, characteristic frequency coefficients, and phase spectrum are compressed and encapsulated into a data frame, including: constructing a data frame header, writing a frame synchronization identifier, data frame length, and timestamp into the data frame header; adding a current operating condition ID field after the data frame header and using fixed-length encoding; performing Huffman coding on the characteristic frequency coefficients to generate a compressed data block of characteristic frequency coefficients; performing differential encoding and run-length encoding on the phase spectrum to generate a compressed data block of phase spectrum; concatenating the data frame in the order of data frame header, current operating condition ID field, compressed data block of characteristic frequency coefficients, and compressed data block of phase spectrum; and adding a cyclic redundancy check code to the end of the data frame to complete the encapsulation of the data frame.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the edge computing compression transmission method for diving monitoring data according to any one of claims 1 to 8.

10. A saturation diving monitoring system, applied to the method described in any one of claims 1-8, characterized in that, include: At least one edge PLC slave station; The water surface monitoring master station is connected to each edge PLC slave station.

Images