Sensor state monitoring method, system and device based on industrial internet of things

Abstract

The application discloses a kind of sensor state monitoring method, system and equipment based on industrial internet of things, it is related to industrial internet of things technical field, method includes: obtaining to be measured node, collaborative node and backup source, calibration and extract noise floor extreme value;Valued space residual difference is formed by acquisition, and deviation sequence is extracted, and disturbance energy and impact extreme value are extracted;When initial threshold is over limit, activate delay inspection, compare response time difference with delay boundary, and record the intrinsic fault of mutation as primary abnormality;Deterioration consensus is calculated to non-exceptional node, and exponential damping dynamic correction threshold based on noise floor extreme value is used, and secondary comparison is executed and recorded as secondary abnormality;Finally, according to the number of exceptions, obtain confirmation duration to switch backup source, and conditionally replace back to be measured node.The application has the advantages of accurate fault identification, threshold adjustment resistance collapse and monitoring stable and reliable.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) technology, and specifically to a sensor status monitoring method, system, and device based on industrial IoT. Background Technology

[0002] In current industrial IoT monitoring scenarios, to ensure the continuity and safety of the production process, real-time status monitoring and redundancy switching of key nodes are typically required. However, existing sensor status monitoring mechanisms still face many limitations in complex real-world operating conditions.

[0003] On the one hand, sensors within the same sensing area often exhibit inherent static background biases due to limitations in installation elevation, hydrodynamic characteristics, and manufacturing tolerances. Directly comparing the original absolute values ​​to the system can easily lead to misjudgments due to misalignment of physical dimensions. On the other hand, existing evaluation models often rely on a single numerical fluctuation threshold, failing to fully incorporate the spatiotemporal propagation patterns of industrial media. When monitoring data undergoes abrupt changes, the system often struggles to distinguish whether the change stems from genuine process variations (such as fluid impact or valve actuation) or is attributable to intrinsic electrical faults caused by sensor hardware aging or short circuits. This can easily lead to the erroneous isolation of normally functioning sensors under severe operating conditions.

[0004] Furthermore, some systems, when introducing dynamic threshold mechanisms, often employ simple linear reduction logic. When a large-scale warning is triggered in the sensing area, the judgment threshold is excessively compressed or even approaches zero, causing the system to lose its resistance to inherent thermal noise from the equipment, leading to judgment boundary failure. Moreover, after triggering a switchover of redundant backup sources, conventional mechanisms often employ rigid timed recovery strategies, lacking conditional safety constraints based on real-time operating conditions. If a forced switchback occurs before physical interference factors are eliminated, it can easily cause repeated oscillations in the control loop, severely impacting the overall reliability of the industrial control system. Summary of the Invention

[0005] To address the technical problems in the background art, the present invention provides a sensor status monitoring method, system, and device based on the Industrial Internet of Things.

[0006] A sensor status monitoring method based on the Industrial Internet of Things (IIoT) includes: acquiring the target node, collaborating nodes, and backup sources; performing background calibration and extracting the noise floor extreme value of the target node; collecting the output values ​​of the target node and collaborating nodes, calculating the spatial residual based on the background calibration results, and arranging the spatial residuals within the analysis period in time sequence to form a deviation sequence; extracting the i-th disturbance energy and the i-th impact extreme value based on the deviation sequence of the i-th collaborating node; activating a delay test if the i-th disturbance energy is greater than a first threshold or the i-th impact extreme value is greater than a second threshold; and comparing the response time difference of abnormal signals. When a sudden change is determined to be an intrinsic fault, the i-th cooperating node is recorded as a Level 1 anomaly based on the delay boundary. If it is not recorded as a Level 1 anomaly, the total number of current Level 1 anomalies is obtained and the degraded consensus is calculated. The first and second thresholds are corrected using exponential damping based on the noise floor extreme value to generate a first corrected threshold and a second corrected threshold. If the energy of the i-th disturbance is greater than the first corrected threshold or the extreme value of the i-th impact is greater than the second corrected threshold, it is recorded as a Level 2 anomaly. The confirmation time is obtained based on the number of Level 1 and Level 2 anomalies, the data source is replaced with a backup source, and after the confirmation time has elapsed, it is conditionally replaced back with the node under test.

[0007] Optionally, the process of extracting the noise floor extreme value includes: synchronous sampling within a stable time window; calculating the difference between the arithmetic mean of the node under test and the i-th cooperating node, which is defined as the static deviation; calculating the standard deviation of the sampling sequence of the node under test, and defining the product of the standard deviation and a preset constant as the noise floor extreme value.

[0008] Optionally, the calculation process of the spatial residual includes: at the sampling time, subtracting the output value of the i-th cooperative node from the output value of the node to be measured, and then subtracting the corresponding static deviation, and defining the calculation result as the spatial residual.

[0009] Optionally, the extraction process of perturbation energy and shock extreme value includes: calculating the square of the difference between two adjacent spatial residuals in the deviation sequence, taking the arithmetic mean of all square values, and then performing a square root operation on the arithmetic mean to obtain the perturbation energy; scanning all spatial residuals in the deviation sequence to calculate the absolute value, and selecting the largest absolute value as the shock extreme value.

[0010] Optionally, the process of obtaining the delay boundary includes: obtaining the spatial coordinates of the node under test and the i-th cooperative node, calculating the Euclidean distance between them, which is defined as the straight-line distance; dividing the straight-line distance by the theoretical wave velocity to obtain the basic delay; subtracting the compensation coefficient from the constant to obtain the compensation ratio, and multiplying the basic delay by the compensation ratio to obtain the delay boundary.

[0011] Optionally, the process of obtaining the response time difference and determining the intrinsic fault includes: extracting the target time when the output value of the node under test reaches the maximum jump rate, and extracting the reference time when the value collected by the i-th cooperating node reaches the maximum jump rate; defining the absolute value of the difference between the target time and the reference time as the response time difference; and determining the sudden change as an intrinsic fault when the response time difference is less than the delay boundary.

[0012] Optionally, the generation process of the first correction threshold and the second correction threshold includes: dividing the total number of first-level anomalies by the total number of cooperating nodes to define the degraded consensus; subtracting the noise floor extreme value from the initial first threshold to obtain the threshold margin, multiplying the threshold margin by the attenuation factor, and adding the noise floor extreme value to generate the first correction threshold; wherein, the attenuation factor is an exponent formed by using the natural constant as the base and the negative number of the product of the convergence coefficient and the degraded consensus as the exponent; using the same attenuation factor, converting the initial second threshold and the noise floor extreme value to generate the second correction threshold.

[0013] Optionally, the process of conditionally replacing the node under test includes: dividing the sum of the number of first-level anomalies and second-level anomalies by the total number of cooperating nodes to obtain the anomaly ratio; multiplying the anomaly ratio by the initial time to obtain the confirmation duration; starting a timer with a running time equal to the confirmation duration; when the countdown ends, acquiring the disturbance energy and impact extreme value of the node under test in the latest analysis period; and replacing the data source back to the node under test when the disturbance energy is less than the first threshold and the impact extreme value is less than the second threshold.

[0014] A sensor status monitoring system based on the Industrial Internet of Things (IIoT) is also provided. The system comprises a management platform, a sensor network platform, and an object platform connected in sequence. The management platform includes: an acquisition module for acquiring the node under test, cooperating nodes, and backup sources; performing background calibration and extracting the noise floor extreme value; a processing module for collecting output values, calculating spatial residuals based on the background calibration results, and forming a deviation sequence; extracting the i-th disturbance energy and the i-th impact extreme value based on the deviation sequence; a first judgment module for activating delay verification when a threshold exceeds the limit, comparing the response time difference of the abnormal signal with the delay boundary, and recording it as a first-level anomaly when the sudden change is determined to be an intrinsic fault; a second judgment module for calculating degradation consensus when not recorded as a first-level anomaly; generating a first correction threshold and a second correction threshold using an exponentially damped correction threshold based on the noise floor extreme value; and recording it as a second-level anomaly based on the corrected thresholds; and a configuration module for obtaining the confirmation time based on the number of first-level and second-level anomalies, replacing the data source with a backup source, and conditionally replacing it back with the node under test after the confirmation time has elapsed.

[0015] An electronic device is also provided, comprising: a memory having a computer program stored thereon; and a processor for executing the computer program in the memory to implement a sensor condition monitoring method based on the Industrial Internet of Things.

[0016] The beneficial effects of this invention are reflected in: This scheme effectively eliminates measurement benchmark interference caused by differences in spatial installation location by introducing static background deviation calibration and inherent noise extraction mechanisms, and establishes an objective physical lower limit for thermal noise for the system. The spatial residual sequence extracted based on this benchmark can accurately quantify the energy of high-frequency transient disturbances and the extreme values ​​of physical impacts, achieving efficient capture of weak degradation characteristics and instantaneous pulses.

[0017] To address the issue of false alarms caused by sudden changes in operating conditions, this invention innovatively constructs a verification mechanism based on spatiotemporal propagation delay. By combining the theoretical wave velocity of the physical medium with the spatial coordinates of the nodes to generate delay boundaries, the system can accurately distinguish between real process fluctuations with spatial propagation delays and intrinsic electrical faults occurring instantaneously in the sensor body, significantly improving the accuracy of fault isolation under complex operating conditions.

[0018] To address the spread of regional faults, this solution employs an exponentially damped threshold evolution model constrained by inherent noise levels. As the consensus on degradation increases within the sensing space, the judgment threshold can be smoothly and adaptively lowered with a lower limit. This ensures sensitivity in capturing secondary anomalies while avoiding the noise reduction failure issues associated with traditional linear threshold descent. Furthermore, the system dynamically plans the isolation confirmation time based on the global anomaly percentage and incorporates conditional back-switch constraints based on real-time characteristic indicators. This effectively avoids blind back-switchovers when faults are not fully resolved, ensuring the continuity of core business operations while minimizing the risk of control system oscillations. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0020] Figure 1 This is a schematic diagram illustrating the steps of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 2 This is a schematic diagram of some steps in S1 of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 3 This is a schematic diagram of some steps in S2 of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 4 This is a schematic diagram of some steps in S3 of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 5This is a schematic diagram of some steps in S4 of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 6 This is a schematic diagram of some steps in S5 of the sensor status monitoring method based on the Industrial Internet of Things of the present invention; Figure 7 This is a schematic diagram of the sensor status monitoring system based on the Industrial Internet of Things of the present invention; Figure 8 This is a schematic diagram of the optimized industrial Internet of Things (IoT) involved in this invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] This invention provides a sensor status monitoring method based on the Industrial Internet of Things, such as... Figure 1 As shown, in one embodiment, the method includes: S1. Obtain the node under test, the cooperating node, and the backup source; perform background calibration and extract the extreme value of the noise floor of the node under test.

[0025] S2. Collect the output values ​​of the node under test and the cooperating node, calculate the spatial residuals in combination with the background calibration results, and arrange the spatial residuals in the analysis period in time sequence to form a deviation sequence; extract the i-th disturbance energy and the i-th impact extreme value based on the deviation sequence of the i-th cooperating node.

[0026] S3. If the energy of the i-th disturbance is greater than the first threshold, or the extreme value of the i-th impact is greater than the second threshold, then the delay test is activated. By comparing the response time difference and delay boundary of the abnormal signal, when the sudden change is determined to be an intrinsic fault, the i-th cooperative node is recorded as a first-level anomaly.

[0027] S4. If not recorded as a Level 1 anomaly, obtain the total number of current Level 1 anomalies and calculate the degradation consensus; use exponential damping based on the noise floor extreme value to correct the first threshold and the second threshold, and generate the first corrected threshold and the second corrected threshold; if the i-th disturbance energy is greater than the first corrected threshold or the i-th impact extreme value is greater than the second corrected threshold, then it is recorded as a Level 2 anomaly.

[0028] S5. Based on the number of first-level and second-level anomalies, obtain the confirmation time, replace the data source with the backup source, and conditionally replace it back to the node under test after the confirmation time has elapsed.

[0029] In this embodiment, it should be noted that in S1, the target node, cooperating node, and backup source are first acquired, and background calibration and noise floor extreme value extraction are performed. This process aims to address the measurement reference deviation caused by different spatial installation positions and to determine the lower limit of the sensor's physical noise. In specific applications, the coordinates of the target node are set as (0,0,0), and the coordinates of the first cooperating node are set as (10,0,0), in meters. Within a stable time window with a pipeline pressure of 100.0 kPa and no fluid fluctuations, the arithmetic mean of the sampling of the target node is 100.0 kPa, and the arithmetic mean of the first cooperating node is 99.5 kPa. Subtracting the two yields a static deviation of 0.5 kPa for the cooperating node. Simultaneously, the standard deviation of the sampling sequence of the target node within this time window is calculated to be 0.05 kPa. Multiplying this by a preset constant 3 yields a noise floor extreme value of 0.15 kPa. By calculating the static deviation, the misalignment of physical dimensions caused by different fluid dynamic characteristics or installation elevations is eliminated, avoiding errors during subsequent comparisons. Extracting the extreme value of the noise floor quantifies the physical limit of the inherent thermal noise of the equipment under the current ambient temperature, providing a reliable objective basis for subsequent dynamic adjustment of the threshold, preventing the judgment boundary from being excessively reduced and failing under harsh working conditions, and improving the reliability of the monitoring benchmark.

[0030] In S2, the spatial residuals are calculated using the static deviation of S1, and the high-frequency transient disturbance energy and impact extreme values ​​are extracted. This solves the problem that existing monitoring methods, which rely solely on simple instantaneous values, are susceptible to slow drift interference. The sampling interval is set to 10 milliseconds, with five consecutive intervals constituting the analysis period. After synchronously acquiring data and subtracting a static deviation of 0.5 kPa, the spatial residual sequence of the first coordinating node is obtained as follows: 0.2, 1.2, -0.8, 1.5, and 0.4 kPa. The absolute values ​​of this sequence are scanned, and the maximum value of 1.5 kPa is selected as the impact extreme value to characterize the maximum deviation amplitude. Simultaneously, the disturbance energy is calculated based on root mean square logic, using the formula: the square root of the mean of the sum of squares of the differences between adjacent residuals. In the specific calculation, the adjacent differences are 1.0, -2.0, 2.3, and -1.1, respectively. Squaring these values ​​yields 1.0, 4.0, 5.29, and 1.21, with a sum of 11.5. Dividing by the degrees of freedom (4) gives a mean of 2.875, and taking the square root yields a perturbation energy of 1.696 kPa. The root mean square algorithm using the squared differences can eliminate the directionality of smooth signal fluctuations and amplify high-frequency transient oscillations, enabling the differentiation between normal slow adjustments in the manufacturing process and abnormal fluctuations in the sensor's internal circuitry, thus objectively quantifying the dynamic stability of the signal.

[0031] In S3, when the monitored index exceeds the set threshold, the delay verification mechanism is activated. By comparing the response time difference with the delay boundary, the problem of distinguishing between actual process fluctuations and intrinsic electrical faults of sensors is solved. This mechanism is triggered because the disturbance energy of the first coordinating node (1.696 kPa) exceeds the preset first threshold of 1.0 kPa. The straight-line distance between the two nodes is calculated to be 10 meters, the theoretical wave velocity of the pipeline fluid is set to 1000 m / s, and the compensation coefficient is 0.1. The calculation logic for the delay boundary is: straight-line distance divided by wave velocity, multiplied by a constant minus the compensation coefficient. Substituting the data, we get 10 divided by 1000 multiplied by 0.9, yielding a delay boundary of 0.009 seconds. Subsequently, the target transition time of the node under test (0.050 seconds) and the reference transition time of the first coordinating node (0.052 seconds) are extracted, and the absolute value of the subtraction is used to obtain the response time difference of 0.002 seconds. Because 0.002 seconds is less than 0.009 seconds, this sudden change is determined to be inconsistent with the propagation time pattern of fluid in space, belonging to an intrinsic fault occurring instantaneously within the node. Therefore, the node is recorded as a Level 1 anomaly. This mechanism uses the finite propagation speed of the physical space medium as the basis for judgment, eliminating the risk of false isolation during drastic changes in operating conditions and improving the accuracy of fault identification.

[0032] In S4, for nodes not classified as Level 1 anomalies, a nonlinear exponential damping model is used to correct the threshold based on the confirmed proportion of Level 1 anomalies. This addresses the drawback of existing linear reduction methods, which may cause the threshold to drop to zero, leading to false alarms. Assuming there are four cooperating nodes in the network, and two of them (including the first node) are classified as Level 1 anomalies, the degradation consensus is 2 divided by 4, which equals 0.5. The first correction threshold is calculated using the formula: the difference between the initial threshold and the noise floor extreme value is multiplied by an exponential decay factor, and then added back to the noise floor extreme value. Substituting the first threshold of 1.0 kPa, the noise floor extreme value of 0.15 kPa, the preset convergence coefficient of 1.386, and the degradation consensus of 0.5, the exponential part -1.386 multiplied by 0.5 equals -0.693. The natural constant e raised to the power of -0.693 is approximately equal to 0.5, resulting in a first correction threshold of 0.575 kPa. This exponential damping algorithm allows the threshold to decrease reasonably during anomaly propagation to improve detection sensitivity, but it is strictly constrained by the noise floor extreme value. At this point, the disturbance energy of the third cooperating node is 0.8 kPa, which is lower than the initial 1.0 kPa but greater than 0.575 kPa. It is accurately recorded as a secondary anomaly, taking into account both the ability to capture secondary anomalies and the noise tolerance of physical hardware.

[0033] In S5, based on the global anomaly results output by S4 and S3, the confirmation duration is dynamically calculated and a conditional recovery with safety constraints is executed. This solves the problem that existing timed switchback strategies may cause repeated oscillations in the control loop when the fault is not eliminated. Statistics show two Level 1 anomalies and one Level 2 anomaly, totaling three. Dividing this by the total number of four coordinating nodes yields an anomaly percentage of 0.75. A preset initial time of 120 seconds is obtained, multiplied by the anomaly percentage of 0.75, and a confirmation duration of 90 seconds is calculated. The data source is then immediately replaced with a backup source. This logic of calculating the confirmation duration proportionally to the scale of fault propagation ensures a longer isolation observation period for more widespread problems, reflecting an objective risk hedging strategy. After starting the 90-second timer, continuous monitoring is performed in the background. When the countdown ends, recovery is not performed directly. Instead, the disturbance energy and impact extreme values ​​of the node under test in the latest cycle are reacquired. Only when all the above dynamic indicators have stably fallen back and are less than the first threshold of 1.0 kPa and the second threshold of 2.0 kPa are the data source replaced back with the node under test. This conditional constraint ensures that the recovery action is performed only after the physical interference has been truly eliminated, thus improving the smoothness and reliability of production scheduling control.

[0034] In summary, the entire sensor condition monitoring method based on the Industrial Internet of Things (IIoT) firstly eliminates measurement benchmark errors caused by installations at different spatial locations through static background deviation calibration and inherent noise extraction, and establishes a physical lower limit boundary for thermal noise. Secondly, spatial residuals are extracted based on this physical benchmark, quantifying the energy of high-frequency transient disturbances and the extreme values ​​of physical impacts, enabling effective feature extraction of weak degradation and instantaneous pulses. Simultaneously, a verification mechanism based on spatiotemporal propagation delay is introduced, using the theoretical wave velocity of the physical medium and spatial coordinates to construct a delay boundary, distinguishing between intrinsic electrical faults of the sensor and real process fluid changes with propagation delays, thus avoiding false isolation caused by changes in operating conditions. Furthermore, an exponentially damped threshold evolution model constrained by the extreme value of inherent noise is established. As the consensus on perceived spatial degradation increases, the modified threshold smoothly decreases to capture secondary abnormal nodes, and is always constrained by the physical noise, avoiding the judgment failure and false alarms caused by the existing linear threshold dropping to zero. Ultimately, not only was the isolation confirmation time dynamically calculated based on the global proportion of anomalies, but a conditional switchback constraint based on the recovery of real-time characteristic indicators was also established. This prevented premature switchback in the state of unrepaired faults, thus ensuring the continuity of core business while reducing the risk of oscillations in industrial control.

[0035] like Figure 2 As shown, in one embodiment, S1 includes: S11, acquiring the topology and spatial coordinates of the device within the physical sensing area. This involves acquiring the node to be tested and deploying a total of N cooperative nodes in the surrounding physical medium, where N is a positive integer greater than or equal to 3. The three-dimensional coordinates of the node to be tested and the three-dimensional coordinates of each cooperative node are entered into the system. An independently running model or redundant node is acquired as a backup source. Here, spatial coordinates refer to the three-dimensional geometric position data of the device in the physical environment.

[0036] S12. Execute time synchronization command. Send time synchronization command to the node under test and all cooperating nodes to ensure that the sampling clocks inside all nodes are consistent and the timestamp difference is controlled within the set error range. The method for determining the error range set in the time synchronization command is as follows: A limited number of historical network communication delay and clock jitter test data are obtained (e.g., 50 cross-node PTP time synchronization protocol point tests are performed under the existing industrial IoT physical architecture). The residual clock offset after clock alignment of each cooperating node in each test is recorded. The probability distribution characteristics of these 50 residual clock offsets are calculated. The maximum deviation tolerance limit that can cover more than 99% of the synchronization accuracy is selected, and the lower limit anti-aliasing verification is performed in conjunction with the propagation time of the theoretical wave speed of the physical medium at the shortest spatial node distance in the network. For example, through statistical processing of 50 historical network synchronization test data, it is found that 99% of the node clock residual offsets converge and are distributed within 0.8 milliseconds. At the same time, since the theoretical shortest propagation time of physical fluid between nodes (e.g., 0.01 seconds) is much greater than this offset, after this series of network delay statistics and physical tolerance verification data processing processes, the set error range for timestamp difference control is finally calibrated to 1 millisecond.

[0037] S13. Perform background calibration and noise floor extreme value extraction. Within a stable time window where the equipment is free from process parameter variations, control all nodes to synchronously perform a fixed number of data samplings. Calculate the arithmetic mean of the sampled values ​​of the node under test and the i-th cooperating node within the time window. Subtract the arithmetic mean of the i-th cooperating node from the arithmetic mean of the node under test; the difference is defined as the static deviation. Background calibration refers to the process of determining the inherent measurement difference under stable environmental conditions without external interference. Simultaneously, within the aforementioned stable time window, calculate the standard deviation of the sampling sequence of the node under test; the product of this standard deviation and a preset constant is defined as the noise floor extreme value. The noise floor extreme value characterizes the lower limit of the inherent physical thermal noise level of the sensor components under the current environment.

[0038] In this embodiment, it should be noted that in S11, the topology and spatial coordinates of the devices within the physical sensing area are first acquired. This step aims to establish an accurate geometric reference system for subsequent multi-node cross-validation and spatiotemporal propagation delay calculation. In specific applications, a total of N equal to 4 collaborative nodes are deployed around the fluid pipeline, and the three-dimensional coordinates of the node to be tested are set as (0,0,0), and the three-dimensional coordinates of the first collaborative node are set as (10,0,0), in meters. Simultaneously, an independently operating digital twin model is connected as a backup source. This operation solves the problem in existing industrial monitoring networks where each sensor is only considered an abstract data source and lacks physical location awareness. By accurately recording the three-dimensional spatial coordinates of the devices, the abstract data flow is bound to the real industrial physical environment, enabling the network to possess spatial dimension resolution capabilities. This structured topology data input provides fundamental data support for calculating the physical straight-line distance between nodes and assessing the propagation time of industrial fluid disturbances in space, allowing the entire monitoring architecture to move beyond the purely data-driven level and deeply integrate with industrial physical process parameters.

[0039] In S12, a time synchronization command is sent to the node under test and all collaborating nodes to ensure that the sampling clocks within all nodes are highly consistent. This is to ensure that the high-frequency dynamic data extracted subsequently has strict comparability on the time axis. In a distributed industrial IoT environment, the local clocks of each sensor node often have slight drifts. Without high-precision synchronization, timestamp differences will mask the actual physical signal propagation delay. By executing the time synchronization command, the timestamp error of each node is strictly controlled within 1 millisecond. This high-precision time alignment mechanism solves the data corruption problem caused by clock misalignment during multi-node collaborative monitoring, ensuring the accuracy of spatial residual calculation and transition moment extraction. Time synchronization is a prerequisite for building a spatiotemporal propagation delay verification mechanism. It enables the precise capture of the tiny time difference between the arrival of abnormal signals at different spatial nodes at the millisecond level, thus providing a reliable time scale for determining whether the abnormal signal is propagated by physical media or caused by an electrical short circuit.

[0040] In step S13, background calibration and noise floor extreme value extraction are performed within a stable time window with a pipeline pressure of 100.0 kPa and no fluid fluctuations. This is used to quantify the fixed deviations caused by the installation environment and the noise lower limit of the equipment hardware. The arithmetic mean of the samples from the tested node is calculated to be 100.0 kPa, and the arithmetic mean of the first cooperating node is 99.5 kPa. Subtracting the two yields a static deviation of 0.5 kPa. Simultaneously, the standard deviation of the sampled sequence from the tested node is calculated to be 0.05 kPa. Multiplying this by a constant 3 yields a noise floor extreme value of 0.15 kPa. The logic of multiplying by a constant 3 here is based on statistical principles and can cover most normal thermal noise fluctuations. This step solves the problem of physical dimension misalignment caused by differences in sensor installation elevation and local fluid dynamics characteristics of the pipeline, eliminating the initial measurement reference error. Meanwhile, the extracted 0.15 kPa noise floor extreme value represents the inherent physical thermal noise level of the equipment components under the current ambient temperature, setting a reliable objective physical baseline for subsequent dynamic threshold decay and preventing the failure of the judgment boundary caused by unlimited threshold decline. The method for determining the preset constant is as follows: obtain historical sampling sequences of the node under test under stable operating conditions over a limited number of times (e.g., nearly 30 fault-free natural days), calculate the local standard deviation of each day's sampling sequence, then summarize these 30 local standard deviations and calculate the upper limit coefficient of their confidence interval. For example, extracting the standard deviation of each stable period of a pipeline pressure sensor over the past 30 days reveals that a coefficient of 3 can cover 99.7% of normal physical thermal noise fluctuations without exceeding the limit. After this series of data statistical and normal distribution extraction data processing steps, the preset constant is finally calibrated to a value of 3.

[0041] like Figure 3As shown, in one embodiment, S2 includes: S21, calculating the spatial residual. A data sampling time period is set, and at a specific sampling time, the output values ​​of the node under test and the output values ​​of the i-th cooperating node are synchronously collected. The output value of the node under test is subtracted from the output value of the i-th cooperating node, and then the corresponding static deviation obtained in S13 is subtracted. The final calculation result is defined as the spatial residual. Here, the spatial residual refers to the pure dynamic physical fluctuation amount after deducting the static deviation caused by spatial distribution. The sampling interval is determined as follows: High-frequency raw analog waveform data from sensor outputs under a limited number of historical normal operating conditions and typical minor fault conditions are acquired (e.g., continuous waveform recordings from 10 typical stable periods and 10 typical minor leakage fluctuation periods are collected). These historical waveforms are then subjected to Fast Fourier Transform (FFT) spectral analysis to extract the upper limit of the main frequency band containing the true physical fluctuation characteristics. Based on the Nyquist-Shannon sampling theorem, the period corresponding to a frequency more than twice the highest historical effective characteristic frequency is selected as the sampling reference time. For example, by analyzing 20 historical waveform recordings and processing the spectrum, it is found that the highest effective frequency of the true physical fluctuation and initial degradation characteristics is concentrated around 40 Hz. According to the sampling theorem, the theoretical minimum distortion-free sampling frequency should be greater than 80 Hz (i.e., the sampling period must be less than 12.5 milliseconds). After this series of spectral extraction and theorem conversion numerical processing steps, the sampling interval for the data sampling period is ultimately preset and calibrated to 10 milliseconds.

[0042] S22. Forming a bias sequence. An analysis period is defined, consisting of K consecutive sampling intervals, where K is a positive integer greater than or equal to 2. The K spatial residuals calculated for the i-th cooperative node within one analysis period are stored and arranged in chronological order to form a bias sequence. The method for determining the number of consecutive sampling intervals K within the analysis period is as follows: Extract a finite number of historical real transient random interference pulse events (e.g., 30 historical brief and harmless spike pulses caused by known electromagnetic interference or non-process minor mechanical vibrations), count the duration of these historical random pulses on the time axis, divide the average of this duration by the set sampling interval, and round the result up before adding a certain number of redundant sample points to ensure that a complete analysis period can span and the algorithm can smooth out such single random short interference. For example, analyzing 30 historical sporadic interference pulse data reveals that the average duration of a single random pulse is approximately 35 milliseconds. Dividing 35 milliseconds by a 10-millisecond sampling interval yields 3.5 sampling points. After rounding up to 4 and adding an additional redundant observation point, the final value of the number of sampling points K constituting an analysis period is determined to be 5.

[0043] S23. Extract the perturbation energy and impact extreme value. Based on the following first formula, calculate the perturbation energy corresponding to the i-th cooperative node:

[0044] In the first formula above, The output is the i-th disturbance energy; where disturbance energy refers to the effective intensity index for measuring the high-frequency transient fluctuations of a signal within a certain time window. This is the numerical identifier of the cooperating node, with a value being a positive integer between 1 and N; The total number of spatial residuals included within an analysis period; This is the position index of each value in the deviation sequence, and its value is a positive integer between 1 and K minus 1; This represents the spatial residual at the j-th position in the deviation sequence; This refers to the spatial residual at the j+1 position within the same deviation sequence; The symbol for the summation operation indicates that a loop accumulation operation is performed.

[0045] Simultaneously, the absolute values ​​of all spatial residuals in the scan deviation sequence are calculated, compared, and the largest absolute value is selected, which is defined as the i-th shock extreme value. Here, the shock extreme value refers to the maximum deviation magnitude that measures a sudden change in the signal within a specified time period.

[0046] In this embodiment, it should be noted that in S21, the sampling interval is set to 10 milliseconds. Output values ​​are synchronously acquired at specific sampling times, and the spatial residual is calculated by combining this with the background calibration results. This process achieves baseline alignment and dynamic feature stripping of the original sensor data. The spatial residual is obtained by subtracting the output value synchronously acquired by the first cooperating node from the output value acquired by the node under test, and then subtracting the 0.5 kPa static deviation obtained in S13. This continuous subtraction calculation logic aims to eliminate fixed measurement differences caused by different spatial locations, while filtering out the pressure background fluctuations common to the entire industrial site. By calculating the spatial residual, the defect of easy misjudgment when directly using the original absolute values ​​for horizontal comparison is solved. The extracted spatial residual becomes a purely characterizing indicator of local dynamic physical fluctuations. It enables sensor data at different physical locations to be compared fairly and accurately on a unified reference plane, providing a high-quality standardized data source for subsequent extraction of high-frequency transient disturbance energy and impact extreme values.

[0047] In S22, the spatial residuals continuously calculated for the first collaborative node within an analysis period are stored and arranged in chronological order, forming a deviation sequence with temporal characteristics. An analysis period of 50 milliseconds is defined as K equals five consecutive 10-millisecond sampling intervals. Within this period, five spatial residuals are obtained sequentially: 0.2, 1.2, -0.8, 1.5, and 0.4, in kilopascals. The logic of organizing these discrete values ​​into a deviation sequence is that data points at a single moment are often susceptible to random interference and cannot reflect the true dynamic trend of signal evolution. By constructing a continuous sequence within a time window, the technical problem of jitter and false alarms easily generated by relying solely on single-point value exceeding the limit alarm is solved. This ordered data structure preserves the signal's gradient and frequency distribution information over a short period, upgrading the analysis perspective from static instantaneous value evaluation to dynamic temporal pattern recognition, thus providing sufficient data samples for subsequent extraction of energy level indicators reflecting the intensity of signal fluctuations.

[0048] In step S23, the first disturbance energy and the first impulse extremum are extracted based on the numerical calculation of the deviation sequence, thus achieving effective quantification of the abnormal signal characteristics. Specifically, in step S23, the characteristic expression is used. To calculate the first The perturbation energy of each cooperating node is calculated, and the computational logic aims to extract the pure dynamic fluctuation intensity from the original sequence containing background drift. In the expression, It represents the effective energy value of high-frequency transient disturbances, indicating the degree of drastic change in the signal within a specific time window; To determine the total number of spatial residuals within the analysis period, in a fluid pipeline application scenario, the following setting is used: A series of consecutive sampling points; and These represent the positions in the deviation sequence at the th position. The and the first Spatial residuals at each location.

[0049] The expression is executed first. The first-order difference operation, in its physical sense, calculates the rate of change of the pressure signal at adjacent sampling times. This subtraction process effectively filters out low-frequency baseline drift caused by slow adjustments in normal processes, retaining only rapidly changing transient components. Subsequently, a squaring operation is performed on the difference result. This step not only eliminates the directional sign difference between pressure increases and decreases but also amplifies anomalous jump signals with large deviations in mathematical weight. Based on specific data from the application scenario, the five spatial residuals are as follows: kPa kPa kPa kPa and kPa, and their adjacent first-order differences are respectively , , and kilopascal, after squaring, is converted to , , and Immediately afterwards, Symbols indicate this The total energy is obtained by cyclically summing the squared values. In the formula That is, divide by the degrees of freedom Used to calculate the arithmetic mean This ensures that the final calculation results are comparable across different time window lengths. Finally, the outer square root operation restores the dimensions to the original pressure units, yielding the perturbation energy. for kPa. This nonlinear feature extraction logic, which first differs and then squares to calculate the mean, solves the technical problem that existing monitoring methods, which rely on a single static threshold, struggle to distinguish between normal slow fluid changes and high-frequency circuit noise within the sensor. It provides an objective characteristic indicator for subsequent anomaly detection that is not affected by steady-state conditions.

[0050] like Figure 4As shown, in one embodiment, S3 includes: S31, initial state determination. For the i-th cooperating node, determine whether the i-th disturbance energy is greater than a first threshold, or whether the i-th impact extreme value is greater than a second threshold. If either condition is met, then the delay check is activated; if neither is greater than the threshold, then the current state is maintained and no marking operation is performed. The delay check refers to the process of identifying the authenticity of a fault by analyzing whether the time difference between the arrival of the abnormal signal at different spatial nodes conforms to the propagation speed of the physical medium. The method for determining the first threshold is as follows: retrieve the disturbance energy calculation records of the same type of cooperative nodes before and after a limited number of historical real intrinsic electrical fault events (such as 50 historical short circuit or aging jump events), extract the peak disturbance energy at the moment these faults occur, and remove the highest and lowest 10% extreme outliers. Calculate the arithmetic mean of the remaining effective historical peak values ​​and multiply it by a lower limit safety factor (such as 0.8) to ensure sensitivity. For example, by analyzing 50 historical pressure jump fault data, after processing and removing extreme values, the average disturbance energy is 1.25 kPa. After multiplying by a safety factor of 0.8 for data scaling, the first threshold value is finally determined to be 1.0 kPa. The method for determining the second threshold is as follows: collect the spatial residual absolute value sequence generated by nodes within the sensing area during a limited number of historical real-world operating condition fluctuations (e.g., 20 historical valve emergency opening and closing operations), extract the maximum absolute value during each operation as a historical impact sample, calculate the median of these 20 historical samples, and add an anti-disturbance margin to this median; for example, by extracting 20 historical valve operation records, the median of the maximum value of the spatial residual absolute value sequence is calculated to be 1.6 kPa, and an additional 0.4 kPa engineering anti-disturbance margin is added for data translation processing, and finally the value of the second threshold is set to 2.0 kPa.

[0051] S32. Calculate the straight-line distance. Based on the three-dimensional coordinates of the node to be tested and the three-dimensional coordinates of the i-th cooperating node obtained in Example 2, calculate the Euclidean distance between the two coordinate points in the three-dimensional coordinate system, and define the calculation result as the straight-line distance.

[0052] S33. Calculate the delay boundary. Calculate the delay boundary based on the following second formula:

[0053] In the second formula above, The delay boundary is calculated; where the delay boundary refers to the theoretical shortest time limit required for an anomalous disturbance to propagate through the physical medium to reach the observation point. The straight-line distance between the node to be tested and the i-th cooperative node is obtained in S32; The theoretical wave velocity of the physical medium is a preset value, and is a known physical constant. The preset compensation coefficient has a fixed real number between zero and one. The basic delay is calculated using the straight-line distance as the numerator and the theoretical wave velocity as the denominator. The compensation coefficient is determined as follows: Actual propagation time data of disturbance signals between the measured node and the cooperating node are extracted from a finite number of historical normal fluid fluctuation events (e.g., 100 normal process adjustment pressure wave transmissions). The relative deviation rate between the actual propagation time and the theoretical propagation time based on the theoretical wave velocity is calculated. Gaussian fitting is applied to these 100 relative deviation rates, and the upper limit corresponding to the 95% confidence level on one side is taken as the compensation standard. For example, statistical analysis of 100 historical fluctuation data reveals that, due to pipeline deformation and temperature, the actual propagation time is at most about 9.8% shorter than the theoretically calculated time. After Gaussian fitting and rounding, the final compensation coefficient is set to 0.1.

[0054] S34. Calculate the response time difference. Extract the target time when the output value of the node under test reaches the maximum jump rate, and extract the reference time when the value collected by the i-th cooperating node reaches the maximum jump rate. Calculate the absolute value of the difference between the target time and the reference time, and define it as the response time difference.

[0055] S35. Perform intrinsic fault determination. Compare the response time difference with the delay boundary. If the response time difference is less than the delay boundary, the signal abrupt change is determined to be inconsistent with the propagation speed law of the physical medium, belonging to an isolated intrinsic fault of the node itself, and the i-th cooperating node is recorded as a first-level anomaly. Here, an intrinsic fault refers to an isolated damage occurring in the node's hardware circuit itself, not a numerical anomaly caused by external actual process changes. If the response time difference is greater than or equal to the delay boundary, it is determined to conform to the physical propagation law, and is not recorded as a first-level anomaly.

[0056] In this embodiment, it should be noted that in S31, an initial state determination is performed on the first collaborative node. The extracted feature indicators are compared with a preset threshold to determine whether to activate the delay check. The initial first threshold is set to 1.0 kPa, and the second threshold is set to 2.0 kPa. Since the calculated first disturbance energy of 1.696 kPa is greater than the first threshold of 1.0 kPa, the logic triggered by either condition activates the subsequent delay check process. Setting such a threshold gating logic before entering complex spatiotemporal physical calculations is to optimize the allocation of computing resources for edge computing devices. This step solves the problem of computing power exhaustion caused by performing high-load spatiotemporal propagation physical calculations on every frame of data under the normal condition of massive concurrent operation of sensors in industrial sites. This coarse screening mechanism ensures that computing resources are only allocated to nodes that have shown clear fluctuations or abnormal signs, maintaining lightweight operation while not overlooking any potentially risky signal mutation events, thus improving overall operating efficiency.

[0057] In step S32, based on the entered 3D coordinates of the node under test and the first collaborating node, the Euclidean distance between the two coordinate points in the 3D coordinate system is calculated, and this result is defined as the straight-line distance. Specifically, using the extended formula of the Pythagorean theorem in 3D space, the distance between the node under test at the origin (0,0,0) and the first collaborating node at coordinates (10,0,0) is calculated, yielding a straight-line distance of 10 meters. The purpose of this calculation logic is to transform the abstract IT network node topology into spatial geometric data with real physical scale significance. This step solves the problem of the lack of physical spatial measurement indicators in existing data network monitoring methods, enabling the monitoring algorithm to perceive the physical distance between devices. Obtaining the straight-line distance is a key bridge connecting the data processing world and the real physical world. It directly determines the time basis required for subsequent process fluid disturbances to propagate between two monitoring points, providing an indispensable spatial scale parameter for assessing the rationality of abnormal signal propagation.

[0058] In S33, the delay boundary calculated based on the straight-line distance and theoretical wave velocity obtained in S32 sets a rigorous physical time lower limit for fault identification. Specifically, following the feature extraction process described above, when the disturbance energy exceeds the limit, S33 introduces an expression... By calculating the propagation delay time boundary of theoretical calculations, this logic establishes an objective benchmark for the time scale to distinguish the physical authenticity of abnormal signals.

[0059] In the formula, It represents the theoretical propagation delay time boundary, which is the theoretical shortest time required for a physical disturbance to propagate from one place to another; For the node to be tested and the first The straight-line distance between collaborative nodes, based on coordinates and This value is calculated using Euclidean geometry principles. rice; The theoretical perturbation propagation wave velocity, representing the physical medium in an industrial setting, is set as follows in a pipeline fluid scenario: meters per second; The preset compensation coefficient is set to [value]. The core part of the expression Following basic kinematic principles, the fundamental time for the fluid pressure wave to propagate between two nodes is obtained by dividing the spatial distance by the physical wave speed. Second.

[0060] However, in real industrial environments, temperature fluctuations in the medium, elastic deformation of the pipe material, and measurement tolerances at node positions can all cause slight deviations in the actual propagating wave velocity. Therefore, the formula introduces... This adjustment factor multiplies the base time by the scaling factor. The final delay boundary is thus obtained as The time difference of a second, mathematically speaking, is equivalent to reserving a safety margin for the theoretical limit, making the time boundary more rigorous. This operational logic solves the technical problem of the lack of physical space constraints in purely data-driven monitoring models, enabling the identification of the essential source of signal mutations. Since the actual response time difference extracted in the application scenario is only a fraction of the time difference... seconds, much smaller than the calculated value. The time limit for propagation is set at the second threshold, which determines that the jump exceeds the physical conduction limit of the fluid medium, thus confirming it as an intrinsic fault within the node. This time-bound calculation based on physical laws eliminates false isolation caused by changes in operating conditions, improving the accuracy of fault diagnosis.

[0061] In step S34, the transition moments between the tested node and the first cooperating node are extracted to calculate the response time difference, achieving precise capture of the time difference of anomaly events occurring in space. Using a high-frequency sampling sequence, the target moment when the output value of the tested node reaches its maximum transition rate is extracted as 0.050 seconds, and the reference moment when the value collected by the first cooperating node reaches its maximum transition rate is extracted as 0.052 seconds. Subtracting these two moments and calculating their absolute values ​​yields the actual response time difference of 0.002 seconds. This logic of extracting the moment of maximum transition rate accurately pinpoints the most intense characteristic instant of the abnormal signal at each sensor. This step solves the problem of accurately locating the trigger point of anomalies in continuous dynamic data streams. The calculated 0.002-second response time difference objectively quantifies the time interval manifested between the tested node and the cooperating node in this pressure surge event, providing a true and accurate measurement basis for ultimately determining whether the event conforms to the laws of fluid dynamics propagation.

[0062] In step S35, intrinsic fault determination is performed, and the final fault conclusion is derived by comparing the response time difference with the delay boundary. The actual response time difference of 0.002 seconds obtained in S34 is compared with the delay boundary of 0.009 seconds established in S33. Because 0.002 seconds is less than the physical lower limit of 0.009 seconds, this determination logic indicates that the data mutation rate exceeds the limit of physical propagation in the fluid medium. Therefore, it cannot be a real change in process conditions, but rather an intrinsic electrical fault occurring instantaneously in the internal circuitry of the node under test. Based on this, the first cooperating node is recorded as a level one anomaly. This step effectively solves the problem that existing monitoring methods easily misjudge normally operating sensors as faulty when faced with drastic changes in operating conditions. By introducing a reliable physical propagation delay law as the judgment criterion, interference caused by changes in process fluid is effectively isolated, significantly improving the accuracy and reliability of intrinsic fault identification of sensor hardware, and ensuring the authenticity of industrial monitoring results.

[0063] like Figure 5As shown, in one implementation, S4 includes: S41, calculating the degradation consensus. When the i-th cooperating node is not recorded as a Level 1 anomaly, the total number of current Level 1 anomalies is obtained. The total number of Level 1 anomalies is divided by the total number of cooperating nodes, which is defined as the degradation consensus. The degradation consensus reflects the percentage of nodes identified as having anomalies within the current sensing area.

[0064] S42. Generate a correction threshold based on the exponential damping of the noise floor extreme value. Calculate the first correction threshold based on the following third formula:

[0065] In the third formula above, This is the first corrected threshold generated; This represents the extracted noise floor extreme value of the node to be tested; The initial first threshold; It is a natural constant; The preset convergence coefficient is a constant with a value greater than zero. To obtain the degradation consensus, the same attenuation factor and noise floor extreme value calculation logic are used simultaneously to convert the initial second threshold and generate a second corrected threshold. The convergence coefficient is determined as follows: A limited number of historical regional fault propagation events are analyzed retrospectively (e.g., records of the past 10 large-scale sensor cluster aging and degradation events). The optimal threshold attenuation level required to successfully provide early warning of secondary degradation nodes when the degradation consensus reaches a specific proportion (e.g., 50%) is recorded for each event. The coefficient value is obtained by establishing an equation using the exponential function in the third formula. For example, historical data analysis reveals that when the degradation consensus is 0.5, the first threshold needs to be precisely attenuated to 50% of its original dynamic margin to optimally capture secondary anomalies (i.e., the target attenuation factor is 0.5). After substituting the data into the equation and performing logarithmic operations, the convergence coefficient is finally determined to be 1.386.

[0066] S43. Perform a secondary judgment. Compare the current node using the corrected threshold. If the i-th disturbance energy is greater than the first corrected threshold, or the i-th impact extreme value is greater than the second corrected threshold, record the i-th cooperative node as a level-two anomaly.

[0067] In this embodiment, it should be noted that in S41, for those cooperative nodes not recorded as Level 1 anomalies, the overall degradation level of the current sensing area is assessed to quantify the spread of spatial anomalies. First, the total number of nodes in the current network that have been clearly recorded as Level 1 anomalies is counted. In the specific application scenario, this number is two nodes, including the first cooperative node. Then, the number of these two Level 1 anomaly nodes is divided by the total number of cooperative nodes deployed in the network, 4, to calculate a real number ratio of 0.5, which is between zero and the constant one. This ratio is defined as the degradation consensus.

[0068] This division operation logic transforms discrete single-point fault events into continuous global spatial state evaluation indicators, solving the technical challenge of treating individual sensors as isolated information sources in existing monitoring systems, making it difficult to coordinate and perceive regional fault development trends. The resulting 0.5 degradation consensus objectively reflects the proportion of nodes identified as having intrinsic faults within the current monitoring range. This parameter directly serves as the core input variable for the subsequent dynamic threshold nonlinear evolution model, ensuring that the sensitivity for judging remaining nodes can be proportionally adjusted according to the actual fault impact range. This lays a data foundation for subsequently capturing secondary abnormal nodes on the verge of degradation.

[0069] In S42, using the acquired degraded consensus, an exponentially damped model based on the noise floor extreme value is employed to generate a corrected threshold, thereby adjusting the detection lower limit for the remaining nodes. The third formula invoked is that the first corrected threshold equals the noise floor extreme value plus the difference between the initial threshold and the noise floor extreme value multiplied by an exponential decay factor.

[0070] Specifically, for the remaining nodes that are not identified as intrinsic faults, S42 uses a characteristic expression. The detection threshold is dynamically adjusted, and this exponentially damped model ensures that the decision boundary is strictly constrained by the physical noise floor while improving sensitivity. In the expression, This is the first corrected threshold generated; The initial first threshold is set to... kPa; To calibrate the extracted noise floor extremes kPa; To degrade consensus, this represents the percentage of nodes currently exhibiting Level 1 anomalies in the scenario. There are nodes An anomaly, therefore for ; To control the convergence coefficient of the decay rate, it is set to... ; It is a natural constant.

[0071] In the formula Calculate The kilopascal threshold adjustment margin indicates that any threshold-lowering operation will only occur within the effective dynamic range above the physical noise floor. The core of the calculation lies in the natural exponential decay term. The negative sign indicates the proportion of faults. An increasing trend followed by a decreasing trend. Substituting the data into the calculation, the exponential part... equal natural constant of The power is approximately equal to the decay factor. The attenuation factor is compared with... Multiplying the remaining amount of kilopascals by the amount of kilopascals yields The compression amount of kilopascals is added back at the end. The extreme value of the noise floor in kPa is used to calculate the new first correction threshold. kPa.

[0072] This calculation logic, employing the natural exponential function, overcomes the technical flaw of existing algorithms that, when reducing the threshold linearly, cause the threshold to drop to zero when the proportion of anomalies is too high. This is because the natural exponential function possesses asymptotic properties, regardless of... How can we increase the threshold? It can only approach, but never fall below, the corrected threshold. This setup ensures that monitoring proactively lowers the judgment threshold to detect... While maintaining secondary degradation signals such as kPa, it also maintained the... The objective physical immunity of kilopascal thermal noise achieves a balance between detection sensitivity and anti-interference capability under complex fault backgrounds.

[0073] In S43, a secondary judgment is performed on the current node in the unmarked state using the generated first correction threshold to identify nodes that have not yet exceeded the initial threshold but have already shown performance degradation. In a specific monitoring scenario, for the third cooperating node, the disturbance energy of the node was initially calculated to be 0.8 kPa. In the initial state, 0.8 kPa is lower than the original first threshold of 1.0 kPa, and the node did not trigger the alarm check. At this time, it is compared with the newly calculated first correction threshold of 0.575 kPa. Since 0.8 kPa is greater than the correction judgment boundary of 0.575 kPa, the secondary comparison condition is met, and the third cooperating node is then recorded as a level two anomaly. This execution logic utilizes dynamically corrected parameters constrained by physical noise floor, solving the technical problem that in complex industrial sites, some sensors are in the early stages of slow degradation, and their weak abnormal characteristics are easily missed by conventional fixed thresholds. By implementing secondary judgment, in the context of the existence of abnormal nodes globally, the tolerance for the remaining nodes is proactively reduced, secondary abnormal signals that are affected by the fault or whose own performance begins to decline are captured, and a multi-dimensional state-aware protection is constructed.

[0074] like Figure 6 As shown, in one embodiment, S5 includes: S51, obtaining the anomaly percentage. The number of first-level anomalies and the number of second-level anomalies are counted, and the two values ​​are added together to obtain the total number of anomaly nodes. The total number of anomaly nodes is divided by the total number of cooperating nodes to obtain the anomaly percentage.

[0075] S52. Calculate the confirmation time. Obtain the preset initial time. Multiply the obtained anomaly percentage by the initial time to obtain the confirmation time. The method for determining the initial time is as follows: statistically analyze a limited number of historical equipment fault isolation and recovery files (e.g., review the actual time taken from the discovery of a single node's intrinsic fault to the complete elimination of physical interference in the past 80 cases), calculate the weighted average of the recovery times of these 80 cases, and add twice the standard deviation to cover the vast majority of long-tail recovery cases; for example, analyzing 80 historical fault files reveals that the average self-recovery or maintenance interference elimination time of a node is 85 seconds, and the statistical standard deviation is 17.5 seconds. Through a series of data processing steps involving the weighted average and the standard deviation, i.e., adding twice 17.5 seconds to 85 seconds, the final calculated initial time is set to 120 seconds.

[0076] S53. Perform replacement. When the anomaly percentage is greater than zero, send a command to the control unit to disconnect the data channel connected to the node under test and replace the data source with the backup source.

[0077] S54. Conditional Replacement. Start a timer with a set running time equal to the confirmation duration. When the countdown ends, acquire the disturbance energy and impact extreme value calculated by the node under test for all cooperating nodes in the latest analysis period. If the disturbance energy is less than the first threshold and the impact extreme value is less than the second threshold, send a command to replace the data source back to the node under test; if the conditions are not met, maintain the connection with the backup source and reset the timer to start counting again.

[0078] In this implementation, it should be noted that in S51, the global anomaly determination results are integrated to obtain the overall anomaly percentage, providing a quantitative basis for subsequent calculations of isolation and recovery times. The number of anomaly nodes at each level is aggregated, and the two collaborative nodes recorded as Level 1 anomalies and the one collaborative node recorded as Level 2 anomalies from the previous steps are added together to obtain a total of 3 anomaly nodes in the current monitoring network. Subsequently, this total number of 3 anomaly nodes is divided by the total number of deployed collaborative nodes (4), yielding a real-valued ratio of 0.75, which is defined as the anomaly percentage. This calculation logic, which merges anomaly nodes of different severity, solves the problem that considering only severe faults or minor degradation cannot accurately reflect the overall scale of industrial damage. The resulting anomaly percentage of 0.75 objectively quantifies the path share of quality problems within the current monitoring node group around the fluid pipeline. This indicator is no longer limited to the health status of a single sensor but extends to the availability assessment of the entire sensing area, enabling the understanding of the prevalence and spread of faults, providing reliable macro-level data support for subsequent risk mitigation strategies and control loop switching decisions.

[0079] In S52, based on the obtained global anomaly ratio, the required isolation and observation confirmation time is dynamically calculated to match the fault handling time with the actual risk scale. First, a preset initial time is obtained from the configuration parameters, which is set to 120 seconds based on the experience of historical pipeline fluid fault recovery. Then, the anomaly ratio of 0.75 is multiplied by this 120-second initial time to calculate the exact confirmation time of 90 seconds. This multiplicative calculation logic, which directly scales the base time using the anomaly ratio as a scaling factor, solves the technical problem of existing redundancy mechanisms using a fixed single dead time, which cannot adapt to fault scenarios of different scales. When the anomaly ratio is high, it indicates a wide fault impact range, and the calculated isolation time is correspondingly extended, thus allowing time buffer for the main node's troubleshooting and repair; conversely, the time is shortened to restore the main node's operation as quickly as possible. The resulting 90-second confirmation time allows for the automatic planning of an isolation period commensurate with the actual risk based on the current damage status of 0.75, ensuring business operation while avoiding excessive occupation of backup data source resources.

[0080] In S53, specific control loop actions are triggered based on the assessment results of the anomaly ratio to prevent unreliable data from negatively impacting industrial production control. Once the calculated anomaly ratio of 0.75 is determined to be greater than zero, a switching command is sent to the underlying industrial control unit. Upon receiving the command, the control unit disconnects the data input channel connected to the faulty monitored node and replaces the production control data source with a pre-prepared backup digital twin model. This hard-switching execution logic based on anomaly ratio solves the problem that distorted data from faulty sensors can lead to malfunctions in pipelines and valves or process control failures when the main sensor experiences intrinsic electrical faults or regional degradation. By immediately replacing the data source with a stable digital twin model, it ensures that even with widespread degradation of the actual monitoring node group, the core control algorithm in the industrial field can still obtain stable and expected input parameters, maintaining the operation of the production process and demonstrating the fault-tolerant protection capability of this monitoring architecture in complex environments.

[0081] In S54, a conditional data source replacement operation with real-time indicator constraints is performed to achieve closed-loop safe recovery of the control loop. Simultaneously with the switchover, an internal timer is started, with its runtime set to the 90-second confirmation duration obtained in S52. At the trigger moment when the 90-second countdown ends, recovery is not performed directly. Instead, the real-time disturbance energy and impact extreme values ​​calculated by the node under test for all cooperating nodes in the latest analysis period are reacquired. It is determined whether all the latest calculated disturbance energies have fallen below the initial first threshold of 1.0 kPa, and whether all impact extreme values ​​are below the second threshold of 2.0 kPa. Only when these conditions are simultaneously met is it confirmed that the physical interference or node anomaly has been truly eliminated, and a command is sent to replace the data source back to the node under test. If any condition is not met, the connection with the digital twin backup source is maintained, and the 90-second timer is reset. This switchover logic with multiple indicator verifications solves the problem of repeated control oscillations caused by forced switchovers before the fault is fully identified in existing simple timed recovery mechanisms, ensuring the smoothness of sensor recovery.

[0082] like Figure 7 As shown, a sensor status monitoring system based on the Industrial Internet of Things (IIoT) is also provided. The system includes a management platform, a sensor network platform, and an object platform that are connected in sequence. The management platform includes: The acquisition module is used to acquire the node under test, the cooperating node, and the backup source; perform background calibration and extract the background noise extreme values; The processing module is used to collect output values, calculate spatial residuals by combining them with the background calibration results, and form a deviation sequence; it also extracts the i-th disturbance energy and the i-th impact extremum based on the deviation sequence. The first judgment module is used to activate the delay test when the threshold is exceeded. By comparing the response time difference of the abnormal signal with the delay boundary, when the sudden change is determined to be an intrinsic fault, it is recorded as a level one anomaly. The second judgment module is used to calculate the degradation consensus when it is not recorded as a first-level anomaly; generate a first correction threshold and a second correction threshold using an exponential damping correction threshold based on the extreme value of the noise floor; and record it as a second-level anomaly based on the corrected threshold. The configuration module is used to obtain the confirmation time based on the number of first-level and second-level anomalies, replace the data source with the backup source, and conditionally replace it back with the node under test after the confirmation time has elapsed.

[0083] An electronic device is also provided, comprising a memory and a processor. The memory stores a computer program containing instructions corresponding to the various processing logics and model calculations described in the above embodiments. The processor is connected to the memory and is used to read and execute the computer program stored in the memory to implement the aforementioned sensor status monitoring method based on the Industrial Internet of Things.

[0084] To enable those skilled in the art to fully understand and implement the technical solutions described in this specification, the following section, in conjunction with a specific application scenario, provides a detailed deduction and data analysis of the entire implementation principle of a sensor status monitoring system based on the Industrial Internet of Things.

[0085] In a specific application scenario of fluid pipeline pressure monitoring in the Industrial Internet of Things (IIoT), the system executes S1 for initialization and baseline calibration. The site deployment includes one target node and N=4 collaborating nodes, with a digital twin model running on an edge computing device configured as a backup source. The system first establishes a spatial coordinate system, setting the 3D coordinates of the target node as (0,0,0) and the first collaborating node as (10,0,0), in meters. Through synchronization commands, the timestamp error of each node is controlled within 1 millisecond. Within a stable time window with a pipeline pressure of 100.0 kPa and no fluid fluctuations, the system performs background calibration. Data collection shows that the arithmetic mean of the sampled data from the target node is 100.0 kPa, and the arithmetic mean of the first collaborating node is 99.5 kPa. Subtracting the two yields a static deviation of 0.5 kPa for the collaborating node. Meanwhile, the standard deviation of the sampling sequence of the node under test within this time window is calculated to be 0.05 kPa. Multiplying it by the preset constant 3, the extreme value of the noise floor is extracted to be 0.15 kPa. This value represents the physical lower limit of the inherent thermal noise of the device at the current ambient temperature.

[0086] Entering S2, the system sets the sampling interval to 10 milliseconds, with K=5 consecutive sampling intervals constituting one analysis cycle. Within a certain analysis cycle, the system synchronously acquires 5 output values ​​from the measured node and the first cooperating node. By subtracting the value of the first cooperating node from the value of the measured node, and then subtracting a static deviation of 0.5 kPa, the system obtains the spatial residual sequence of the first cooperating node. The five spatial residuals are 0.2, 1.2, -0.8, 1.5, and 0.4, respectively, all in kPa. The system scans this deviation sequence, calculating the absolute values ​​of 0.2, 1.2, 0.8, 1.5, and 0.4, with the maximum value being 1.5 kPa. The system extracts 1.5 kPa and defines it as the first impact extreme value, used to characterize the maximum abrupt change in the pressure signal within this time period.

[0087] In the feature extraction stage of S2, the system extracts the first perturbation energy based on the deviation sequence using the first formula: For the first cooperating node, the differences in its adjacent spatial residuals are as follows: 1.2 minus 0.2 equals 1.0, -0.8 minus 1.2 equals -2.0, 1.5 minus -0.8 equals 2.3, and 0.4 minus 1.5 equals -1.1. Squaring these four differences yields 1.0, 4.0, 5.29, and 1.21 respectively. Summing these four squared values ​​gives 11.5. Since K equals 5 and the denominator K-1 is 4, dividing the sum 11.5 by 4 yields an arithmetic mean of 2.875. Rooting 2.875 gives the first disturbance energy as 1.696 kPa, accurately quantifying the high-frequency transient disturbance intensity of the pressure signal within the current observation period.

[0088] Upon execution up to S3, the system sets the initial first threshold to 1.0 kPa and the second threshold to 2.0 kPa. Since the calculated first disturbance energy of 1.696 kPa exceeds the first threshold of 1.0 kPa, the system immediately activates the delay check. Based on the spatial coordinates, the system calculates the straight-line distance between the node under test and the first cooperating node to be 10 meters. The theoretical wave velocity of the pipeline fluid medium is preset to 1000 m / s, and the compensation coefficient is set to 0.1. The system calls the second formula to calculate the delay boundary: The calculation process is as follows: 10 divided by 1000 yields a base delay of 0.01 seconds; a constant of 1 minus 0.1 yields a compensation ratio of 0.9; multiplying these two gives a delay boundary of 0.009 seconds. The system extracts the target time of the node's transition as 0.050 seconds, and the reference time of the first coordinating node's transition as 0.052 seconds; subtracting these and taking the absolute value yields a response time difference of 0.002 seconds. Because 0.002 seconds is less than the physical lower limit of 0.009 seconds, the system determines that this sudden change does not conform to the fluid pressure propagation law and belongs to an intrinsic fault occurring instantaneously within the node circuit; therefore, the first coordinating node is recorded as a Level 1 anomaly. Assuming similar verification, the second coordinating node is also recorded as a Level 1 anomaly.

[0089] In S4, for the 3rd and 4th coordinating nodes that were not recorded as Level 1 anomalies, the system counts the total number of Level 1 anomalies as 2. The system divides this total of 2 by the total number of coordinating nodes, 4, to calculate a degraded consensus of 0.5. The system then uses the third formula to generate the first correction threshold: Substituting the known data, the first threshold is 1.0 kPa, the extreme noise floor is 0.15 kPa, the preset convergence coefficient is 1.386, and the degradation consensus is 0.5. The exponential part, -1.386 multiplied by 0.5, equals -0.693, and the natural constant e raised to the power of -0.693 is approximately 0.5. Subtracting 0.15 from 1.0 gives 0.85, multiplying by 0.5 gives 0.425, and finally adding the extreme noise floor of 0.15, a new first corrected threshold of 0.575 kPa is obtained. This exponential damping mechanism ensures that the threshold decreases reasonably as the anomaly spreads, but remains above the physical noise floor of 0.15 kPa. Since the disturbance energy of the third cooperating node is calculated to be 0.8 kPa, although this value is lower than the original 1.0 kPa, it is greater than the current first corrected threshold of 0.575 kPa. The system accurately captures this secondary attenuation signal and records the third cooperating node as a second-level anomaly.

[0090] Finally, in the S5 decision-making and execution phase, the system statistics show that there are currently 2 Level 1 anomalies and 1 Level 2 anomaly, for a total of 3 anomalies. Dividing 3 by the total number of coordinating nodes (4) yields an anomaly percentage of 0.75. The initial system time is configured to be 120 seconds; multiplying this by the anomaly percentage of 0.75 yields a confirmation time of 90 seconds. The system immediately outputs control commands, replacing the data source of the control loop from the tested node experiencing the intrinsic fault with a digital twin backup source to ensure the continuity and safety of pipeline pressure regulation. The system synchronously starts a 90-second timer. After the 90-second countdown, the system extracts the real-time disturbance energy and impact extreme values ​​of the tested node in the latest cycle. Only when all indicators have stably fallen back to below the initial 1.0 kPa and 2.0 kPa values ​​does the system confirm that the fault has been eliminated and conditionally replace the data source back with the tested node, completing the fully closed-loop intelligent scheduling and control.

[0091] It should also be noted that the entire sensor condition monitoring system based on the Industrial Internet of Things can be applied to the optimized Industrial Internet of Things. Figure 7 This is a schematic diagram of the sensor status monitoring system based on the Industrial Internet of Things of the present invention. Figure 8 This is a schematic diagram illustrating the optimized industrial Internet of Things (IIoT) involved in this invention. (See diagram below.) Figure 7 and Figure 8 As shown, the optimized Industrial Internet of Things (IIoT) includes a user platform, a service platform, a management platform, a sensor network platform, and an object platform that establish communication in sequence. The user platform is configured to provide front-end services to users; users obtain the necessary perception service information through the user platform, process the perception service information, and transform it into user perception information; users analyze the user perception information and make corresponding decisions based on their own wishes, and transform the user perception information into user control information through the corresponding information system and send it to the service platform, thereby demonstrating the user's corresponding service needs and wishes.

[0092] The physical entities of the user platform include various user terminals, such as mobile phones, computers, and dedicated terminals, which provide user services through integration with user information system software.

[0093] The service platform is configured as an API server or other server used to establish communication between the management platform and the user platform to achieve corresponding functions; the physical entity of the service platform includes various servers.

[0094] The management platform is configured to perform at least one of the following: device operation status monitoring and management, data monitoring and management, device parameter management, and lifecycle management; the management platform is the overall operation platform for the Internet of Things, which may include various management sub-platforms, with different management sub-platforms performing different management tasks; the physical entities of the management platform include various servers.

[0095] The sensor network platform is configured to perform at least one of the following functions: network management, command management, device status management, data protocol management, data parsing, data classification, data transmission monitoring, and data transmission security management. The sensor network platform provides functions such as data communication, transmission, parsing, identification, and classification, avoiding the direct aggregation of data from various object platforms onto the management platform, which would otherwise result in data redundancy and low data processing efficiency. The physical entities of the object platforms include various gateways, edge computing devices, etc.

[0096] The object platform is configured to perform specific production control, detection, measurement and other production tasks; the physical entities of the production objects include various production equipment, sensors and so on.

[0097] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0098] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0099] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A sensor status monitoring method based on the Industrial Internet of Things, characterized in that, The methods include: Obtain the node to be tested, the cooperating node, and the backup source; Perform background calibration and extract the extreme values ​​of the noise floor of the node under test; The output values ​​of the node under test and the cooperative node are collected, and the spatial residuals are calculated in combination with the baseline calibration results. The spatial residuals within the analysis period are then arranged in time sequence to form a deviation sequence. Extract the i-th perturbation energy and the i-th impact extreme value based on the deviation sequence of the i-th cooperative node; If the i-th disturbance energy is greater than the first threshold, or the i-th impact extreme value is greater than the second threshold, then the delay test is activated; By comparing the response time difference and delay boundary of the abnormal signal, when the sudden change is determined to be an intrinsic fault, the i-th cooperative node is recorded as a first-level anomaly. If it is not recorded as a Level 1 anomaly, obtain the total number of current Level 1 anomalies and calculate the degradation consensus; use exponential damping based on the aforementioned noise floor extreme value to correct the first threshold and the second threshold, and generate a first corrected threshold and a second corrected threshold. If the i-th disturbance energy is greater than the first correction threshold or the i-th impact extreme value is greater than the second correction threshold, it is recorded as a level two anomaly; Based on the number of first-level and second-level anomalies, the confirmation time is obtained, the data source is replaced with a backup source, and after the confirmation time has elapsed, it is conditionally replaced back with the node under test.

2. The sensor status monitoring method based on the Industrial Internet of Things according to claim 1, characterized in that, The process of extracting the extreme values ​​of the background noise includes: Simultaneous sampling within a stable time window; The difference between the arithmetic mean of the node to be tested and the i-th cooperative node is calculated and defined as the static deviation; Calculate the standard deviation of the sampling sequence of the node to be tested, and define the product of the standard deviation and a preset constant as the noise floor extreme value.

3. The sensor status monitoring method based on the Industrial Internet of Things according to claim 2, characterized in that, The calculation process of the spatial residual includes: At the sampling time, the output value of the node to be tested is subtracted from the output value of the i-th cooperative node, and then the corresponding static deviation is subtracted. The calculation result is defined as the spatial residual.

4. The sensor status monitoring method based on the Industrial Internet of Things according to claim 1, characterized in that, The process of extracting the disturbance energy and impact extreme values ​​includes: The perturbation energy is obtained by calculating the square of the difference between two adjacent spatial residuals in the deviation sequence, taking the arithmetic mean of all squares, and then performing a square root operation on the arithmetic mean. The absolute value of all spatial residuals in the scan deviation sequence is calculated, and the largest absolute value is selected as the shock extreme value.

5. The sensor status monitoring method based on the Industrial Internet of Things according to claim 1, characterized in that, The process of obtaining the delay boundary includes: Obtain the spatial coordinates of the node to be tested and the i-th cooperative node, calculate the Euclidean distance between them, and define it as the straight-line distance; The basic delay is obtained by dividing the straight-line distance by the theoretical wave speed; the compensation ratio is obtained by subtracting the compensation coefficient from the constant; and the delay boundary is obtained by multiplying the basic delay by the compensation ratio.

6. The sensor status monitoring method based on the Industrial Internet of Things according to claim 5, characterized in that, The process of obtaining the response time difference and determining intrinsic faults includes: Extract the target time when the output value of the node under test reaches the maximum jump rate, and extract the reference time when the value collected by the i-th cooperative node reaches the maximum jump rate; The absolute value of the difference between the target time and the reference time is defined as the response time difference; When the response time difference is less than the delay boundary, the sudden change is determined to be the intrinsic fault.

7. The sensor status monitoring method based on the Industrial Internet of Things according to claim 1, characterized in that, The process of generating the first correction threshold and the second correction threshold includes: dividing the total number of the first-level anomalies by the total number of cooperating nodes, which is defined as the degraded consensus; The threshold margin is obtained by subtracting the noise floor extreme value from the initial first threshold. The threshold margin is then multiplied by the attenuation factor and added to the noise floor extreme value to generate the first corrected threshold. The attenuation factor is an exponent with the natural constant as the base and the negative of the product of the convergence coefficient and the deteriorated consensus as the exponent. Using the same attenuation factor, the initial second threshold is converted with the extreme value of the noise floor to generate the second corrected threshold.

8. The sensor status monitoring method based on the Industrial Internet of Things according to claim 1, characterized in that, The process of conditionally replacing the node to be tested includes: The sum of the number of first-level anomalies and second-level anomalies is divided by the total number of cooperating nodes to obtain the anomaly ratio, and the anomaly ratio is multiplied by the initial time to obtain the confirmation duration. A timer with a running time equal to the confirmation duration is started. When the countdown ends, the disturbance energy and impact extreme value of the node under test in the latest analysis period are obtained. When the disturbance energy is less than the first threshold and the impact extreme value is less than the second threshold, the data source is replaced back to the node under test.

9. A sensor condition monitoring system based on the Industrial Internet of Things, characterized in that, The system includes a management platform, a sensor network platform, and an object platform that are sequentially connected in communication. The management platform includes: The acquisition module is used to acquire the node under test, the cooperating node, and the backup source; perform background calibration and extract the background noise extreme values; The processing module is used to collect output values, calculate spatial residuals by combining them with the background calibration results, and form a deviation sequence; it also extracts the i-th disturbance energy and the i-th impact extremum based on the deviation sequence. The first judgment module is used to activate the delay test when the threshold is exceeded. By comparing the response time difference of the abnormal signal with the delay boundary, when the sudden change is determined to be an intrinsic fault, it is recorded as a level one anomaly. The second judgment module is used to calculate the degradation consensus when it is not recorded as a first-level anomaly; generate a first correction threshold and a second correction threshold using an exponential damping correction threshold based on the extreme value of the noise floor; and record it as a second-level anomaly based on the corrected threshold. The configuration module is used to obtain the confirmation time based on the number of first-level and second-level anomalies, replace the data source with the backup source, and conditionally replace it back with the node under test after the confirmation time has elapsed.

10. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the sensor status monitoring method based on the Industrial Internet of Things as described in any one of claims 1 to 8.

Images