Internet of things data edge computing processing method for ship shore-based collaborative management

Abstract

This invention belongs to the field of data processing technology, specifically relating to an IoT data edge computing processing method for ship-shore collaborative management. The method includes: acquiring real-time operating data of the ship's engine room equipment; loading a physical association weight matrix, calculating the local spatiotemporal coupling strength index of each sensor by calculating the cross-correlation coefficient and time lag between each sensor within the current window; extracting the transient change characteristics of the data at the current moment, calculating the prediction residual of the current observation using a prediction model, and calculating the adaptive transmission priority weight of each sensor's data based on the local spatiotemporal coupling strength index; and dividing the data stream into different levels according to the adaptive transmission priority weight, and executing a dynamic queue scheduling strategy for each level. This invention can improve the accuracy of fault diagnosis while reducing redundant data uploaded by the system.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to an IoT data edge computing processing method for ship-shore collaborative management. Background Technology

[0002] With the rapid development of intelligent ship technology, ship-shore collaborative management has become an important means to improve shipping safety and efficiency. Modern ships typically deploy multiple Internet of Things (IoT) sensors in their engine rooms to monitor and acquire operating status data such as temperature, pressure, and speed of the main engine core equipment and auxiliary engine core equipment. The operating status data of the main engine core equipment and auxiliary engine core equipment needs to be transmitted back to the shore-based data center via satellite communication link so that the shore-based expert system can perform remote fault diagnosis and health management.

[0003] Currently, most existing shipboard data acquisition gateways employ either timed polling or simple dead-zone threshold reporting strategies. Timed polling refers to a communication method where the shipboard data acquisition gateway sends query commands to each IoT sensor in the engine room at preset fixed time intervals, forcing each IoT sensor to return its current real-time operating status data. Dead-zone threshold reporting means that the operating status data collected by the IoT sensor will only be reported when the difference between the new value collected by the IoT sensor and the previously reported value exceeds a preset threshold range.

[0004] However, the operating status data such as temperature, pressure and speed in the ship's engine room are highly correlated in terms of physical correlation. Periodic polling or simple dead zone threshold reporting strategies result in the repeated transmission of a large amount of trend information, which seriously wastes satellite communication bandwidth resources. At the same time, the low sampling rate set in order to save satellite communication bandwidth resources often leads to the loss of key waveform details at the moment of equipment failure, making it difficult for shore-based operators to perform accurate root cause analysis. Summary of the Invention

[0005] To address the problems of severe communication bandwidth waste and difficulty in diagnosis due to the loss of key waveform details, this invention provides an IoT data edge computing processing method for ship-shore collaborative management, which can improve the accuracy of fault diagnosis while reducing redundant data uploaded by the system.

[0006] This invention provides an IoT data edge computing processing method for ship-shore collaborative management, comprising: acquiring real-time operating data of ship engine room equipment, constructing a data acquisition vector, calculating the statistical characteristics of the data within the current window using a sliding time window, and standardizing the original physical measurement values ​​to obtain standardized observation values; loading a physical correlation weight matrix, calculating the cross-correlation coefficient and time lag between each sensor within the current window, and combining the physical correlation weight matrix and lag attenuation factor to calculate the local spatiotemporal coupling strength index of each sensor to assess the degree of co-occurrence of current sensor data with other sensor data; extracting the transient change characteristics of the data at the current moment, calculating the prediction residual of the current observation value using a prediction model, and calculating the adaptive transmission priority weight of each sensor data in combination with the local spatiotemporal coupling strength index, wherein the adaptive transmission priority weight is positively correlated with the data change amplitude and prediction residual, and negatively correlated with the degree of coupling redundancy; and dividing the data stream into different levels according to the adaptive transmission priority weight, and executing dynamic queue scheduling strategies of direct transmission, intelligent compression, or silent suppression respectively to complete the transmission of data to the shore-based data center.

[0007] By adopting the above technical solution, real-time operational data of ship engine room equipment is acquired and a data acquisition vector is constructed. Statistical characteristics of the data are calculated using a sliding time window, and the original physical measurements are standardized. This effectively solves the problem of the large variety of IoT sensors in the ship's engine room and the significant differences between their dimensions. By mapping data of different dimensions, such as main engine speed and cylinder exhaust temperature, to a dimensionless space of the same scale, the calculation bias caused by different physical dimensions is eliminated, laying a unified data foundation for subsequent multidimensional data fusion analysis. Simultaneously, by loading a physical correlation weight matrix and combining it with a hysteresis attenuation factor to calculate the local spatiotemporal coupling strength index, the temporal and spatial correlation of different sensor data can be dynamically identified, accurately distinguishing independent information from redundant information. This provides a scientific basis for subsequent execution of differentiated dynamic queue scheduling strategies, reducing the generation of massive amounts of low-value redundant data at the source.

[0008] Preferably, the standardized observation values Satisfying the relation:

[0009] ;

[0010] in, Representing the One sensor in The original physical measurements at that time and Sensors The mean and standard deviation within the current sliding time window To prevent constants with a denominator of 0.

[0011] By adopting the above technical solution, the processing of subtracting the mean and dividing by the standard deviation enables the standardized observations to objectively reflect the degree to which the current sensor values ​​deviate from historical levels. The standardization of the original physical measurement values ​​not only eliminates the influence of physical dimensions on subsequent calculations, but also highlights the fluctuation characteristics of the data through standardization, enabling the edge computing gateway to more sensitively capture the slight deviations in the operating status of the equipment, thus providing a standardized and stable data source for achieving high-precision ship-shore collaborative management.

[0012] Preferably, the first The local spatiotemporal coupling strength index of the sensor Satisfying the relation:

[0013] ;

[0014] in, This represents the total number of associated sensor sets. For the sensors in the pre-stored physical correlation weight matrix and Physical association weights, Sensors within the current window and The absolute value of the Pearson correlation coefficient corresponding to the peak value of the cross-correlation. This represents the time lag corresponding to the peak value of the cross-correlation. As the normalization factor, This is the hysteresis decay constant.

[0015] By adopting the above technical solution, the pre-stored physical correlation weight matrix, the absolute value of the Pearson correlation coefficient corresponding to the cross-correlation peak, and the time lag are comprehensively considered. The introduction of the exponential decay term significantly enhances the scientific rigor of the algorithm, effectively increasing the weight of signals that change synchronously in real time and reducing the weight of signals with excessively long lag times. In this way, the system can dynamically identify whether exhaust temperature data has independent transmission value when facing correlated operating conditions such as increased exhaust temperature due to increased host engine speed. When high data redundancy is detected, the system reduces the amount of repeated backhaul via the satellite communication link, thereby significantly reducing the total number of data transmission packets while ensuring information integrity.

[0016] Preferably, the step of extracting the transient change characteristics of the data at the current moment and calculating the prediction residual of the current observation using a prediction model specifically includes: calculating the absolute value of the first difference of the data at the current moment relative to the previous moment as the change magnitude; running a linear regression prediction model, using the associated principal parameters to predict the theoretical value of the current parameter, and calculating the absolute value of the difference between the theoretical value and the current standardized observation as the prediction residual.

[0017] By employing the aforementioned technical solution, the absolute value of the first-order difference between the current data and the previous data is calculated to capture high-frequency fluctuation characteristics. This is combined with a linear regression prediction model to calculate the prediction residual, providing a dual verification mechanism for assessing data importance. This model-based residual calculation method can keenly identify decoupling anomalies in equipment, thus transforming abstract data fluctuations into physically meaningful fault symptom indicators. This method not only captures transient changes in data at high frequencies but also identifies potential performance degradation or sudden faults through residual analysis, providing crucial input parameters for subsequent calculations of adaptive transmission priority weights and ensuring the accurate extraction of fault characteristic information.

[0018] Preferably, the adaptive transmission priority weight Satisfying the relation:

[0019] ;

[0020] in, The rate of change weighting coefficient. Based on the weighting coefficient, It is an index of local spatiotemporal coupling strength. As a redundancy sensitivity factor, For anomaly detection weighting coefficients, For the range of change, To predict residuals.

[0021] By adopting the above technical solution and introducing the rate of change weight coefficient, basic weight coefficient, redundancy sensitivity factor and anomaly capture weight coefficient, the system can automatically reduce the weight when the data is stable and the redundancy is high, and forcibly increase the weight when the prediction residual is detected to be abnormal or the data fluctuates drastically. This provides a logical basis for the subsequent screening of high-value data, enabling the edge computing gateway to automatically identify key information with fault signs and suppress low-value duplicate information.

[0022] Preferably, dividing the data stream into different levels according to the adaptive transmission priority weight specifically includes: setting a high-priority threshold. and medium priority threshold If adaptive transmission priority weight Greater than If the data stream level is determined to be direct transmission level, then the adaptive transmission priority weight is used. Greater than and less than or equal to If the data stream level is determined to be intelligent compression level, then the adaptive transmission priority weight is used. Less than or equal to If so, the data stream level is determined to be the silent suppression level.

[0023] By adopting the above technical solution and setting high-priority and medium-priority thresholds, a rigorous data flow classification and judgment logic was established, laying the foundation for implementing differentiated dynamic queue scheduling strategies. This classification approach enables edge computing gateways to adjust processing modes in real time based on the actual value of the data, ensuring that data of different importance is matched with the most suitable transmission resources. This not only improves the automation level of data processing but also provides clear triggering conditions for flexibly adjusting sampling frequency, data packet transmission, or local storage under different operating conditions, demonstrating the real-time optimization and economic benefits of data transmission strategies.

[0024] Preferably, when the data stream level is direct transmission level, the following strategy is executed: maintain the original data acquisition frequency and do not downsample the data; mark the current data frame as a key frame and immediately package and send it to the shore-based data center.

[0025] Preferably, when the data stream level is intelligent compression level, the following strategy is executed: the data is downsampled; the downsampled data is marked with an association tag so that the shore-based system can use the master parameter waveform for interpolation and restoration.

[0026] Preferably, when the data stream level is the silent suppression level, the following strategy is executed: real-time data is stored in local storage medium and not sent to the shore-based data center in real time; every preset time interval, a statistical packet containing statistical mean and extreme values ​​is generated and sent to the shore-based data center for the shore-based system to generate virtual data.

[0027] Preferably, the length of the sliding time window is set to 5 minutes, and the data from each sensor are statistically analyzed in real time through the time window to calculate the mean and the standard deviation.

[0028] The beneficial effects of this invention are:

[0029] This invention utilizes the physical coupling characteristics of ship engine room equipment to construct a local correlation model at the edge, and solves the problem of redundant evaluation of time-series signals by introducing hysteresis correlation calculation and exponential decay mechanism. At the same time, for asynchronous but related physical quantities, the overlap of physical quantity information is accurately calculated through time lag compensation, so that the system can significantly reduce duplicate data packets without losing correlation information.

[0030] Furthermore, by suppressing redundant data generated by physical correlations, traffic flow is significantly reduced and operating costs are lowered during steady-state ship cruising. At the same time, the physical model residual triggering mechanism is used to capture minute abnormal signs and automatically switch to high-frequency sampling to completely preserve fault waveforms, thereby providing detailed evidence for shore-based diagnosis. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the IoT data edge computing processing method for ship-shore collaborative management in this invention.

[0032] Figure 2 This is a schematic view illustrating the physical coupling strength analysis of multiple sensors in a ship's engine room according to an embodiment of the present invention;

[0033] Figure 3 This is a schematic diagram illustrating the original waveform of the sensor according to an embodiment of the present invention;

[0034] Figure 4 This is a schematic diagram illustrating the adaptive transmission frequency response relationship according to an embodiment of the present invention;

[0035] Figure 5 This is a schematic diagram showing a comparison of data traffic consumption between existing technologies and the present invention. Detailed Implementation

[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0037] This invention discloses an IoT data edge computing processing method for ship-shore collaborative management.

[0038] Reference Figure 1 An IoT data edge computing processing method for ship-shore collaborative management includes steps S1-S4:

[0039] S1. Obtain real-time operating data of the ship's engine room equipment, construct a data acquisition vector, calculate the statistical characteristics of the data within the current window using a sliding time window, and standardize the original physical measurement values ​​to obtain standardized observation values.

[0040] In an optional embodiment, the edge computing gateway periodically collects operational status data of the main engine core equipment and auxiliary engine core equipment in the ship's engine room at a preset frequency via a shipboard industrial bus or OPCUA interface. The shipboard industrial bus is typically a CAN bus or Modbus-RTU protocol, and the preset frequency for the edge computing gateway to collect operational status data is 10 Hz in this embodiment. Subsequently, a data acquisition vector is defined by the system. , specifically, ,in Representing the One sensor in The original physical measurement value at that moment.

[0041] It is important to note that there are many types of IoT sensors in a ship's engine room, and the dimensions of the operational status data collected by different IoT sensors vary significantly. For example, the main engine speed may be 100-120 rpm, while the cylinder exhaust temperature may be as high as 400-500℃, and the lubricating oil pressure may only be 0.5-0.8 MPa. If the original data is used directly for calculation, the larger operational status data will have a higher weight, while the changes in smaller but still important operational status data will be masked. Therefore, this embodiment of the invention sets a sliding time window, which uses each IoT sensor... Mean within the current window and standard deviation Characterize the rate of change of the data collected by each IoT sensor and calculate the standardized observation value. Standardized observations The calculation method is as follows:

[0042] ;

[0043] in, To prevent extremely small constants with a denominator of 0, in this embodiment of the invention, it is set as follows: .

[0044] To more clearly illustrate the role and calculation process of standardized observations, the embodiments of this invention provide examples:

[0045] Suppose at a certain moment Sensor at cylinder number 1 Measured exhaust temperature ℃, the average value calculated over the past 5-minute sliding window ℃; standard deviation ℃; for Then standardized observations This means that the sensor at cylinder number 1 The current value is one standard deviation higher than the historical average. Through this processing, the data from all sensors are mapped to a dimensionless space of the same scale.

[0046] In this way, by standardizing the dataset, the calculation bias caused by different physical dimensions is eliminated, laying a unified data foundation for subsequent multidimensional data fusion analysis.

[0047] S2. Load the physical correlation weight matrix. By calculating the cross-correlation coefficient and time lag between each sensor in the current window, and combining the physical correlation weight matrix and lag attenuation factor, calculate the local spatiotemporal coupling strength index of each sensor to evaluate the degree of co-occurrence of current sensor data with other sensor data.

[0048] In an optional embodiment, by identifying the relationships between different operational state data, redundant information between sensors can be discovered. When the value of a certain sensor can be derived from the data of other sensors through logical operations, the normal steady-state data generated by that sensor does not need to be repeatedly transmitted back via the satellite communication link. This processing method can significantly reduce the total amount of data packets that need to be transmitted. Therefore, the present invention calculates the first... Local spatiotemporal coupling strength index of individual sensors This is used to characterize the redundancy information between different sensors.

[0049] Specifically, the edge computing gateway first pre-stores a physical association weight matrix issued by the shore-side expert database. Matrix elements Indicates sensor and The correlation strength lies in the physical mechanism; for example, an increase in engine speed leads to an increase in exhaust temperature, indicating a high correlation between engine speed and exhaust temperature; while the correlation between engine speed and cabin temperature is low. The edge computing gateway then calculates the sensor data within the sliding window. and Find the absolute value of the Pearson correlation coefficient corresponding to the peak value of the cross-correlation function. and the time lag when the peak occurs. Finally, calculate the first... Local spatiotemporal coupling strength index of individual sensors :

[0050] ;

[0051] in, This represents the total number of associated sensor sets. For the sensors in the pre-stored physical correlation weight matrix and Physical association weights, Sensors within the current window and The absolute value of the Pearson correlation coefficient corresponding to the peak value of the cross-correlation. This represents the time lag corresponding to the peak value of the cross-correlation. Normalization factor; The hysteresis decay constant; the exponential decay term This is used to increase the weight of signals that change synchronously in real time, while decreasing the weight of signals with excessively long lag times.

[0052] To explain the first more clearly Local spatiotemporal coupling strength index of individual sensors The function and calculation process of [the method / mechanism] are illustrated by examples in this embodiment of the invention:

[0053] Suppose we need to calculate the local spatiotemporal coupling strength index of the cylinder exhaust temperature measured by sensor A at cylinder 1. The main sensors associated with the cylinder exhaust temperature measured by sensor A are... To simplify calculations, the measured host speed data only considers the sensor data. The measured host speed data is associated with this object, and settings are configured. The value is 2.

[0054] Simultaneously set the parameter as: physical association weight. The value is 0.9, provided by the shore end, indicating a close physical relationship between cylinder exhaust temperature and main engine speed data; the cross-correlation coefficient calculated by the sliding window. A value of 0.95 indicates a high degree of similarity between the waveforms of cylinder exhaust temperature and main engine speed data; time lag. The 2-second delay indicates that the cylinder exhaust temperature lags behind the engine speed data by 2 seconds; the hysteresis decay constant. If it is 0.5; then .

[0055] The above calculations show that although the cylinder exhaust temperature and engine speed data have high physical weight and correlation, due to a 2-second lag between their changes, the instantaneous redundancy after attenuation is only 0.35. This indicates that simply transmitting engine speed data cannot fully represent the current exhaust temperature data in real time; the exhaust temperature data still has some independent transmission value. If the lag is 0 seconds, then... The redundancy will be as high as 0.95, at which point the exhaust temperature data will be almost completely redundant.

[0056] Thus, by introducing the spatiotemporal coupling strength index, we can dynamically identify the correlation between data in time and space, accurately distinguish between independent and redundant information, and avoid misjudgment.

[0057] S3. Extract the transient change characteristics of the data at the current moment, use the prediction model to calculate the prediction residual of the current observation, and combine the local spatiotemporal coupling strength index to calculate the adaptive transmission priority weight of each sensor data. The adaptive transmission priority weight is positively correlated with the data change amplitude and prediction residual, and negatively correlated with the degree of coupling redundancy.

[0058] In an optional embodiment, the variation range of the data measured by each sensor is taken into account. Local spatiotemporal coupling strength index and predicted residuals This transforms abstract data features into an adaptive transmission priority weight. This provides a logical basis for subsequent screening of high-value data, enabling the identification of key information with signs of failure and low-value duplicate information.

[0059] When it is necessary to calculate adaptive transmission priority weights First, calculate the magnitude of change in the data at the current time relative to the previous time. This variation amplitude is used to capture high-frequency fluctuations; subsequently, a lightweight linear regression prediction model is run through data processing hardware nodes such as ship-side data acquisition gateways or edge computing nodes deployed in the ship's engine room to calculate the predicted value. Thus, the predicted residuals are obtained. If the predicted residual value is large, it indicates a decoupling anomaly, which usually means an internal fault in the equipment. Calculate the adaptive transmission priority weight. The method is as follows:

[0060] ;

[0061] in, The weighting coefficient for the rate of change; The weighting coefficients for the basic items; The weighting coefficient for outliers; This is the redundancy sensitivity factor.

[0062] To more clearly illustrate the adaptive transmission priority weight The function and calculation process of [the method / mechanism] are illustrated by examples in this embodiment of the invention:

[0063] First, set: It is 0.1; It is 0.9; It is 0.05; =1; =1; It is 3; It is 5;

[0064] If this is under normal operating conditions, then Therefore, it has a lower priority at this time, making it suitable for low-frequency transmission or silence;

[0065] If this is under abnormal operating conditions, for example: although It is 0.1, and The value was 0.9, but the actual value deviated from the predicted value. For example, pipe blockage caused abnormal pressure, making the actual value 0.9. If it is 2, then Therefore, the priority spikes significantly, far exceeding normal values, requiring the immediate uploading of signal data.

[0066] Thus, through this multi-factor coupling priority formula, the system can automatically reduce weights when the data is stable and redundant, and forcibly increase weights when abnormal model residuals are detected, ensuring keen detection of fault signs.

[0067] S4. Based on the adaptive transmission priority weight, the data stream is divided into different levels, and dynamic queue scheduling strategies such as direct transmission, intelligent compression, or silent suppression are executed respectively to complete the transmission of data to the shore-based data center.

[0068] In an optional embodiment, a high-priority threshold is first set at the edge. and medium priority threshold The high-priority threshold is set to 8 and the medium-priority threshold is set to 2.

[0069] When the calculated adaptive transmission priority weight is greater than or equal to the high priority threshold of 8, the data stream level is determined to be the direct transmission level. In this mode, the edge end maintains the original data acquisition frequency, does not perform downsampling processing on the data, marks the current data frame as a key frame, and immediately packages and sends it to the shore-based data center, thereby ensuring that the shore end can obtain key waveform details of the transient state of the equipment failure.

[0070] When the calculated adaptive transmission priority weight is between the medium priority threshold 2 and the high priority threshold 8, the data stream level is determined to be the intelligent compression level. In this mode, the edge performs 4:1 downsampling on the data and adds an association mark to the downsampled data so that the shore-based system can use the associated master parameter waveform for interpolation and restoration, thereby saving satellite communication bandwidth resources while retaining the necessary trend characteristics.

[0071] When the calculated adaptive transmission priority weight is less than or equal to the medium priority threshold of 2, the data flow level is determined to be the silent suppression level. In this mode, real-time data is only stored in the local storage medium and is not sent in real time. The edge end generates a statistical packet containing the statistical mean and extreme values ​​every preset time interval and sends it to the shore-based data center. The shore-based system generates virtual data based on the local spatiotemporal coupling strength index and historical model to maintain the logical continuity of the remote monitoring system with extremely low bandwidth consumption.

[0072] In this way, through hierarchical dynamic scheduling, the data transmission strategy can be optimized in real time, which significantly reduces communication costs while ensuring the availability and integrity of data at critical moments.

[0073] Reference Figure 2 It intuitively displays the matrix The calculated coupling relationships are shown, with the darker areas at the intersection of engine speed and exhaust temperature indicating high correlation, which are the primary targets for redundancy reduction in this method. (Refer to...) Figure 3 and Figure 4 When the original waveform is stable, the transmission frequency remains low; however, when the original waveform experiences a sudden spike, the transmission frequency jumps instantaneously. (Refer to...) Figure 5 The solid line representing the coupling redundancy analysis of this invention significantly reduces the cumulative traffic compared to the dashed line representing the fixed-frequency full-volume transmission of the prior art. It only experiences a slight increase during abnormal times, demonstrating the economic benefits of this invention.

[0074] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

Claims

1. An IoT data edge computing processing method for ship-shore collaborative management, characterized in that, include: Acquire real-time operating data of ship engine room equipment, construct data acquisition vectors, calculate statistical characteristics of data within the current window using a sliding time window, and standardize the raw physical measurement values ​​to obtain standardized observation values; Load the physical correlation weight matrix, calculate the cross-correlation coefficient and time lag between each sensor within the current window, and combine the physical correlation weight matrix and lag attenuation factor to calculate the local spatiotemporal coupling strength index of each sensor. ;in, This represents the total number of associated sensor sets. For the sensors in the pre-stored physical correlation weight matrix and Physical association weights, Sensors within the current window and The absolute value of the Pearson correlation coefficient corresponding to the peak value of the cross-correlation. This represents the time lag corresponding to the peak value of the cross-correlation. As the normalization factor, It is the hysteresis decay constant, used to assess the degree of co-occurrence of current sensor data with other sensor data; Extract the transient change characteristics of the data at the current moment, use the prediction model to calculate the prediction residual of the current observation, and combine the local spatiotemporal coupling strength index to calculate the adaptive transmission priority weight of each sensor data. The adaptive transmission priority weight is positively correlated with the data change amplitude and prediction residual, and negatively correlated with the degree of coupling redundancy. Based on the adaptive transmission priority weight, the data stream is divided into different levels, and dynamic queue scheduling strategies such as direct transmission, intelligent compression, or silent suppression are executed respectively to complete the transmission of data to the shore-based data center.

2. The IoT data edge computing processing method for ship-shore collaborative management according to claim 1, characterized in that, The standardized observations Satisfying the relation: ； in, Representing the One sensor in The original physical measurements at that time and Sensors The mean and standard deviation within the current sliding time window To prevent constants with a denominator of 0.

3. The IoT data edge computing processing method for ship-shore collaborative management according to claim 1, characterized in that, The step of extracting the transient change features of the current data and calculating the prediction residual of the current observation using a prediction model specifically includes: Calculate the absolute value of the first-order difference between the current time and the previous time, as the magnitude of change; Run a linear regression prediction model, use the associated principal parameters to predict the theoretical value of the current parameter, and calculate the absolute value of the difference between the theoretical value and the current standardized observation value as the prediction residual.

4. The IoT data edge computing processing method for ship-shore collaborative management according to claim 1, characterized in that, The adaptive transmission priority weight Satisfying the relation: ； in, The rate of change weighting coefficient. Based on the weighting coefficient, It is an index of local spatiotemporal coupling strength. As a redundancy sensitivity factor, For anomaly detection weighting coefficients, For the range of change, To predict residuals.

5. The IoT data edge computing processing method for ship-shore collaborative management according to claim 2, characterized in that, The step of dividing the data stream into different levels based on the adaptive transmission priority weight specifically includes: Set high priority threshold and medium priority threshold ; If adaptive transmission priority weight Greater than If so, the data stream level is determined to be the direct transmission level; If adaptive transmission priority weight Greater than and less than or equal to If so, the data stream level is determined to be the intelligent compression level; If adaptive transmission priority weight Less than or equal to If so, the data stream level is determined to be the silent suppression level.

6. The IoT data edge computing processing method for ship-shore collaborative management according to claim 5, characterized in that, When the data stream level is direct transmission level, the following strategy is executed: Maintain the original data acquisition frequency and do not downsample the data; Mark the current data frame as a key frame and immediately package and send it to the shore-based data center.

7. The IoT data edge computing processing method for ship-shore collaborative management according to claim 5, characterized in that, When the data stream level is set to intelligent compression, the following strategy is executed: Downsampling is performed on the data; The downsampled data is labeled with an association marker so that the shore-based system can use the master parameter waveform for interpolation and reconstruction.

8. The IoT data edge computing processing method for ship-shore collaborative management according to claim 5, characterized in that, When the data flow level is the silent suppression level, the following strategy is executed: Real-time data is stored on local storage media and is not sent to the shore-based data center in real time. At preset time intervals, a statistical packet containing the statistical mean and extreme values ​​is generated and sent to the shore-based data center for the shore-based system to generate virtual data.

9. The IoT data edge computing processing method for ship-shore collaborative management according to claim 2, characterized in that, The length of the sliding time window is set to 5 minutes. The data from each sensor are statistically analyzed in real time through the time window to calculate the mean and the standard deviation.

Images