Industrial internet data three-level connection method and system

Abstract

The invention discloses an industrial internet data three-level connection method and system, and the method comprises the steps: collecting equipment data at an equipment layer, generating an equipment abnormality index based on multi-source data in combination with a process stage label, and triggering edge preliminary screening; receiving each equipment anomaly index at the production line layer, and calculating the comprehensive disturbance intensity of the production line according to the equipment anomaly indexes and the process dependence; the method comprises the following steps: collecting a disturbance intensity sequence of each production line at a factory layer, and calculating a coupling risk index through dynamic time warping, mode entropy and energy consumption dispersion; carrying out marginal sensitivity analysis on the coupling risk indexes to obtain a key production line set, and obtaining a potential root cause conduction path according to the equipment anomaly indexes and the comprehensive disturbance intensity of the production lines in the set; and verifying the vibration spectrum collected by the core equipment in the path and the historical baseline, matching the verification result with the historical maintenance record, and realizing the self-adaptive updating of the rule according to the matching result and the subsequent tracking. According to the invention, the fault positioning precision of the industrial equipment is improved.

Description

Technical Field

[0001] This invention relates to the industrial field, and in particular to a method and system for three-level data connectivity in the industrial internet. Background Technology

[0002] In modern industrial systems, the sheer number of devices, complex coupling relationships, and dynamic changes in operating status make it difficult to determine whether an overall factory performance degradation or disturbance propagation is due to an individual device malfunction or a multi-device imbalance. This leads to ambiguous positioning and delayed response. Sensor data, production line operation data, and factory scheduling data are scattered across different systems, lacking a unified analytical framework, resulting in a situation where data is visible but faults are difficult to locate. Most systems focus only on the alarm thresholds of individual devices, making it difficult to detect cascading problems caused by other devices that affect the current device, leading to misjudgments of the root cause. Fault handling relies on experienced operators, requiring extensive troubleshooting from scratch after a fault occurs, consuming significant time. Against this backdrop, there is an urgent need for a fault tracing method that can take a global perspective, delve step-by-step, accurately locate, and continuously learn from the past. Based on this, this invention proposes a three-level data integration method and system for the Industrial Internet. Summary of the Invention

[0003] This invention provides a method for three-level data connectivity in the industrial internet, characterized by comprising: S10. Collect real-time multi-source data of the equipment at the equipment layer, and generate an equipment anomaly index based on the multi-source data and the current equipment process stage label, and trigger edge screening. S20. Receive the abnormal index uploaded by each device at the production line level, and calculate the overall disturbance intensity of the production line based on the abnormal index of each device and the process dependency relationship through dynamic weight aggregation and topology influence diffusion algorithm. S30. Collect disturbance intensity sequences of each production line at the factory level, and calculate the coupled risk index through dynamic time warping, mode entropy and energy consumption dispersion to conduct system-level risk early warning. S40, the factory coupling risk index obtains a set of key production lines through marginal sensitivity analysis. Based on the equipment anomaly index and the comprehensive disturbance intensity of the production lines within the set, potential root cause transmission paths are obtained through time-series cross-correlation and topological inversion. S50. Collect vibration spectrum data from core equipment along the path and verify it against historical baselines. Match the verification results with historical maintenance records. Based on the matching results and subsequent tracking, achieve adaptive updates of the rule system.

[0004] The above-described three-level data integration method for the Industrial Internet involves collecting multi-source real-time data from equipment at the equipment level. Based on this multi-source data and the current equipment's process stage tag, an equipment anomaly index is generated, triggering edge screening. Specifically, it comprises the following sub-steps: For each device, four types of signals are collected: vibration, temperature, current, and acoustic emission. The variance normalization amplitude within the sliding window is calculated for each signal, and the original abnormal potential energy value is generated by weighted sum of squares. By introducing process stage labels, abnormal potential energy is aligned with stage benchmarks, eliminating the influence of operating condition drift, and outputting an abnormal index that is adaptive to operating conditions. The piecewise linear slope change rate and peak-to-valley ratio are calculated based on the abnormal index sequence. If both exceed the dynamic threshold at the same time, it is determined to be a local mutation event, triggering an initial alarm at the edge layer.

[0005] The above-described three-level data connectivity method for the Industrial Internet involves receiving anomaly indices uploaded by each device at the production line level. Based on the anomaly indices of each device and their process dependencies, a dynamic weight aggregation and topology influence diffusion algorithm is used to calculate the overall disturbance intensity of the production line. Specifically, this method comprises the following sub-steps: Construct a production line equipment topology map, set initial connection weights based on the process flow direction and equipment output dependencies, and when a certain equipment alarms, spread the influence factors along the topology path; Update the operational contribution of each device, dynamically adjust its weight in the aggregate calculation, and form a real-time importance matrix; The abnormal index of all equipment is multiplied by its dynamic weight and then summed to generate the comprehensive disturbance intensity of the production line. At the same time, a Fourier transform is performed on the sequence to extract the dominant frequency and identify whether the production line is in a periodic oscillation mode.

[0006] The above-described three-level data integration method for the Industrial Internet involves collecting disturbance intensity sequences from each production line at the factory level. A coupled risk index is calculated using dynamic time warping, pattern entropy, and energy consumption dispersion to provide system-level risk early warning. Specifically, this method comprises the following sub-steps: Collect the disturbance intensity sequences of each production line, calculate the dynamic time warping distance between each pair, construct the production line behavior similarity matrix, and cluster them into operation mode groups; The plant-wide operating mode entropy is calculated based on the similarity matrix to reflect the consistency of production line collaboration, and the dispersion coefficient of energy consumption per unit output is also calculated. A coupling risk index is constructed. When the adaptive threshold is reached, it is determined to be a system-level imbalance, and a three-level through-response is initiated. The above-described three-level data connectivity method for the Industrial Internet involves using a factory coupling risk index to obtain a set of key production lines through marginal sensitivity analysis. Based on the equipment anomaly index and overall disturbance intensity of these production lines, potential root cause transmission paths are derived through temporal cross-correlation and topological inversion. Specifically, this method comprises the following sub-steps: Define risk sensitivity, estimate the marginal contribution of each production line to the total risk through numerical differentiation, and screen the contribution to obtain the set of key production lines; Within key production lines, a causal response delay map is constructed based on the time-series cross-correlation peak value between equipment anomaly index and overall production line disturbance intensity to identify the core equipment that responds fastest to production line disturbances. By combining equipment topology weights and response latency, the longest impact path is traced in reverse to deduce the potential transmission chain from the source equipment to the factory-level imbalance.

[0007] The above-described three-level data connectivity method for the Industrial Internet combines device topology weights and response latency to trace the longest impact path in reverse, deducing the potential transmission chain of imbalances from source devices to the factory level. Specifically, it consists of the following sub-steps: Based on the reverse of the production line equipment topology diagram, an impact tracing diagram is generated. Based on the core response equipment in the diagram, all possible upstream devices are searched along the reverse edge. Update edge weights to give higher weights to the paths of devices that experienced anomalies earlier in time. Using Dijkstra's algorithm, we search for paths in the influence source graph and select the path with the highest cumulative influence score as the potential root cause transmission chain.

[0008] The above-described three-level data connectivity method for the Industrial Internet involves verifying the vibration spectrum of core equipment along the path against historical baselines, matching the verification results with historical maintenance records, and adaptively updating the rule system based on the matching results and subsequent tracking. Specifically, it comprises the following sub-steps: For the core equipment along the simulation path, extract its high-frequency sampled vibration signals to confirm whether there are any quantifiable physical anomalies. Match the abnormal equipment data with historical maintenance records. If the match is successful, solidify the abnormal pattern, fault type and handling strategy as a new rule and write it into the knowledge base. The rules will be updated adaptively based on subsequent actual verification. This invention also provides an industrial internet data three-level interconnection system, comprising: Data Acquisition and Anomaly Module: Collects real-time data from multiple sources at the equipment level, and generates an equipment anomaly index based on the multi-source data and the current equipment's process stage label, triggering edge screening. The receiving and calculation module receives the abnormality index uploaded by each device at the production line level, and calculates the overall disturbance intensity of the production line based on the abnormality index of each device and the process dependency relationship through dynamic weight aggregation and topology influence diffusion algorithm. Risk warning module: Collects disturbance intensity sequences of each production line at the factory level, and calculates a coupled risk index through dynamic time warping, mode entropy and energy consumption dispersion to conduct system-level risk warning; The reverse path module: The factory coupling risk index obtains a set of key production lines through marginal sensitivity analysis. Based on the equipment anomaly index and the comprehensive disturbance intensity of the production lines within the set, the potential root cause transmission path is obtained through time-series cross-correlation and topological reverse inference. Verification and update module: Verify the vibration spectrum collected from the core equipment in the path with the historical baseline, match the verification results with historical maintenance records, and realize the adaptive update of the rule system based on the matching results and subsequent tracking.

[0009] The beneficial effects achieved by this invention are as follows: This invention improves the accuracy of locating faults in industrial equipment. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0011] Figure 1 This is a flowchart of a three-level data connectivity method for the Industrial Internet provided in Embodiment 1 of this application.

[0012] Figure 2 This is a schematic diagram of a three-level data interconnection system for the industrial internet provided in Embodiment 2 of this application. Detailed Implementation

[0013] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides a method for three-level data connectivity in the industrial internet, including: S10. Collect multi-source real-time data of the equipment at the equipment layer. Based on the multi-source data and the current process stage label of the equipment, generate the equipment anomaly index and trigger edge screening.

[0015] S11. Collect four types of signals for each device: vibration, temperature, current, and acoustic emission. Calculate the normalized variance amplitude of each signal within the sliding window and generate the original abnormal potential energy value through weighted sum of squares.

[0016] High-frequency data acquisition is performed on four types of sensors deployed on each industrial device: vibration, temperature, current, and acoustic emission, to obtain raw operating signals. Within a fixed-length sliding time window, the variance of each type of signal is calculated to reflect its fluctuation intensity. To eliminate the influence of different physical dimensions and individual device differences, the variance of each signal is normalized by dividing it by the reference variance of the same type of signal under historical steady-state conditions, resulting in a variance ratio. Finally, based on preset weights according to equipment type and process characteristics, a weighted sum of squares is performed on the normalized variance ratios to generate the original abnormal potential energy value. The higher this value, the greater the overall deviation of the current multi-source signals of the device.

[0017] S12. Introduce process stage labels, perform stage benchmark alignment for abnormal potential energy, eliminate the influence of operating condition drift, and output an abnormal index that adapts to the operating condition.

[0018] The system introduces the current process stage label for the equipment, including startup, loading, steady-state operation, deceleration, and shutdown. It then retrieves historical baseline data for the corresponding stage from a pre-stored process knowledge base, including the average and standard deviation of the original abnormal potential energy values ​​at that stage. These baseline parameters are used to standardize the current original abnormal potential energy values, resulting in a condition-adaptive abnormality index after operating condition alignment. If this value is less than zero, it is set to zero to ensure that the index under normal operating conditions approaches the baseline level. This alignment mechanism effectively eliminates parameter drift interference caused by normal equipment startup and shutdown, load changes, and other operations.

[0019] S13. Calculate the piecewise linear slope change rate and peak-to-valley ratio based on the abnormal index sequence. If both exceed the dynamic threshold at the same time, it is determined to be a local mutation event, triggering an initial alarm at the edge layer.

[0020] Based on the continuously output adaptive anomaly index sequence, a piecewise linear regression method is used to divide it into two segments, fitting two linear slopes for each segment. The ratio of the absolute difference between these two slopes to the time interval is calculated to obtain the slope change rate, which measures the acceleration of the abnormal trend. Simultaneously, the peak-to-valley ratio (the ratio of the maximum to the minimum value in the sequence) is calculated to capture the amplitude abrupt changes in abnormal fluctuations. Two thresholds are dynamically set based on historical data: a multiple of the median slope change rate and an empirical multiple of the peak-to-valley ratio. When both the current slope change rate and the peak-to-valley ratio exceed their respective dynamic thresholds, a significant local abrupt event is determined to have occurred in the equipment, indicating a sharp deterioration in its operating status. An initial alarm is then triggered at the edge, and information such as the equipment ID, abrupt change intensity, and timestamp is packaged into an event packet and uploaded to the production line layer for further aggregation and analysis.

[0021] S20. Receive the abnormal indexes uploaded by each device at the production line level, and calculate the overall disturbance intensity of the production line based on the abnormal indexes of each device and the process dependency relationship through dynamic weight aggregation and topology influence diffusion algorithm.

[0022] S21. Construct a production line equipment topology diagram, set initial connection weights based on the process flow direction and equipment output dependencies, and spread the influence factors along the topology path when a certain equipment alarms.

[0023] Based on the production line's process flow diagram and the material, energy, and control logic relationships between equipment, a directed weighted graph structure is constructed, serving as the production line's equipment topology. Nodes in the graph represent individual devices, edges represent output dependencies between devices, and the direction of the edges indicates the direction of influence propagation. Initial connection weights are set based on the device's position and criticality in the process chain. When a device triggers a local mutation event alarm at the device level, an influence propagation mechanism is activated. Starting from that device, the disturbance influence factor propagates forward along the directed edges of the topology graph. First-level downstream devices receive 100% of the influence value, second-level devices receive 70%, and third-level and further downstream devices receive 50%, forming an influence distribution map reflecting the range of disturbance propagation. This map is used for adjusting the pre-weights in subsequent weighted aggregation.

[0024] S22. Update the operational contribution of each device, dynamically adjust its weight in the aggregate calculation, and form a real-time importance matrix.

[0025] To avoid rigid analysis caused by fixed weights, the operational contribution of each device in the current time period is dynamically updated every 5 minutes, which serves as its real-time weight in perturbation aggregation.

[0026] The formula for calculating operational contribution is as follows: ,in, Assign coefficients to weight efficiency and health. The segment efficiency of equipment k refers to the product of the availability segment and the performance segment of equipment k within a specific short time period. The availability segment is the proportion of time that equipment k is actually in operation during that time period, and the performance segment is the ratio of the actual production speed of equipment k during operation to the standard cycle speed. This represents the average efficiency of the production line. It is an abnormal inhibitory factor. denoted as the anomaly index of device k.

[0027] Calculation results Real-time importance weights are generated after normalization. The weights of all devices constitute a real-time importance matrix that evolves over time, reflecting the dominant role of highly efficient and low-abnormal devices in production line status assessment.

[0028] S23. Multiply the abnormal index of all equipment by its dynamic weight and sum them to generate the comprehensive disturbance intensity of the production line. At the same time, perform Fourier transform on the sequence to extract the dominant frequency and identify whether the production line is in a periodic oscillation mode.

[0029] Obtain the adaptive anomaly index of each device's operating condition. and its corresponding dynamic weights Combined with the influence factors generated by topological diffusion in step S21 The overall disturbance intensity of the production line is generated using the following formula: Where n is the number of devices in the production line. This is the alarm enhancement factor, used to amplify the impact of known disturbance sources. Indicates whether device k is affected by alarms. This is a distance attenuation factor, which reduces the effect as the propagation distance increases. Let k be the topology hop count from the initial alarm source to device k. This allows the impact to decrease along the path, preventing remote devices from being over-amplified.

[0030] A higher overall disturbance intensity value indicates greater instability in the production line's operation. Simultaneously, a Fast Fourier Transform is performed on the continuously generated overall disturbance intensity sequence to analyze its spectral characteristics and extract the dominant frequency with the largest amplitude. If the maximum dominant frequency falls within the range of 0.001–0.01 Hz and its amplitude exceeds three times the average background noise, the production line is determined to be in a periodic oscillation mode, caused by control loop oscillation, mechanical resonance, or cycle mismatch, requiring close monitoring.

[0031] S30. Collect disturbance intensity sequences of each production line at the factory level, and calculate the coupled risk index through dynamic time warping, mode entropy and energy consumption dispersion to conduct system-level risk early warning.

[0032] S31. Collect the disturbance intensity sequences of each production line, calculate the dynamic time regularization distance between each pair, construct the production line behavior similarity matrix, and cluster them into operation mode groups.

[0033] The plant synchronously collects time series data on the comprehensive disturbance intensity of each production line, with each series containing 12 data points from the most recent hour. To overcome the misjudgment of Euclidean distance caused by inconsistent disturbance rhythms across different production lines, a dynamic time warping algorithm is used to calculate the minimum alignment distance between any two comprehensive disturbance intensity series of production lines. This distance reflects the similarity of the two production lines on the disturbance evolution path; the smaller the distance, the closer the operating fluctuation patterns of the two production lines are. Subsequently, all minimum alignment distances are combined into an N×N distance matrix, where N is the total number of production lines, and transformed into a behavioral similarity matrix. Based on this matrix, a hierarchical clustering algorithm is used to divide all production lines in the plant into four operating mode groups: steady-state low disturbance group, periodic oscillation group, high disturbance sudden change group, and random fluctuation group.

[0034] S32. Calculate the plant-wide operating mode entropy based on the similarity matrix to reflect the consistency of production line collaboration, and at the same time calculate the dispersion of energy consumption per unit output.

[0035] The percentage of production lines included in each operating mode group is statistically analyzed. Based on this, the entropy of the entire plant's operating mode is calculated, and the specific formula is as follows: K represents the number of operating mode groups. The larger the value, the more dispersed the production line operation mode and the worse the coordination. If all production lines belong to the same group, then... This indicates a high degree of consistency.

[0036] Collect energy consumption data per unit output from each production line during the same time period, and calculate its dispersion, which is the standard deviation of energy consumption per unit output for each production line divided by the average energy consumption per unit output for each production line. The larger the value, the higher the energy consumption of some production lines, indicating a risk of resource waste or inefficient equipment operation. S33. Construct a coupling risk index. When the index value exceeds the adaptive threshold, it is determined to be a system-level imbalance, and a three-level through-response is initiated.

[0037] To achieve a unified quantitative assessment of system-level risks, a coupled risk index R is constructed, which integrates synergy, energy efficiency consistency, and maximum disturbance level. The specific formula is as follows: ,in, For the entropy of the operating mode, The energy consumption dispersion per unit output, The adjustment factor for the maximum disturbance term. This represents the highest overall disturbance intensity across all production lines. For reference disturbance intensity, The gain coefficient of the average disturbance response term. This is the slope control factor for the average disturbance response term. The sum of the overall disturbance intensity of all production lines is given by N, where N is the total number of production lines.

[0038] Set adaptive threshold This is a series of historical coupling risk indices calculated hourly over the past 7 days. The median of the historical coupling risk index series. The median absolute deviation of the historical coupling risk index series is used to determine if the current coupling risk index R is greater than the adaptive threshold. If the system is deemed to be in a state of imbalance, a three-level interconnected response mechanism will be triggered, a reverse tracing process will be initiated, and an early warning message will be sent to the factory management.

[0039] S40, the factory coupling risk index, obtains a set of key production lines through marginal sensitivity analysis. Based on the equipment anomaly index and the comprehensive disturbance intensity of the production lines within the set, potential root cause transmission paths are obtained through time-series cross-correlation and topological inference.

[0040] S41. Define risk sensitivity, estimate the marginal contribution of each production line to the total risk through numerical differentiation, and screen out the set of key production lines.

[0041] After determining system-level imbalance at the plant level, to identify the main sources of risk, it is necessary to quantify the marginal contribution of each production line to the overall coupled risk. Therefore, risk sensitivity is defined as the partial derivative of the plant-level coupled risk index R with respect to the comprehensive disturbance intensity of the k-th production line. Since R is a combination of multiple nonlinear functions, its analytical derivative cannot be directly calculated; therefore, a numerical differentiation method is used for estimation. The estimation process is as follows: Obtain the comprehensive disturbance intensity value of each production line at the current moment. Apply a small disturbance to each production line k, first slightly increasing its comprehensive disturbance intensity by a fixed amount, while keeping the data of all other production lines unchanged. Substitute this adjusted data back into the coupled risk calculation formula to obtain the comprehensive disturbance intensity +. Then, correspondingly decrease the disturbance intensity of the production line by the same small amount and recalculate the risk value to obtain the comprehensive disturbance intensity -. By comparing the difference in risk values ​​before and after these two disturbances, and dividing this difference by twice the amount of disturbance, the rate of influence of that production line on the overall risk is obtained, i.e., the risk sensitivity. The larger the value, the more sensitive the operating state of that production line is to the overall plant risk.

[0042] After calculating the risk sensitivity of all production lines, they are sorted in descending order, and the top 20% of production lines with the highest sensitivity are selected to form the critical production line set.

[0043] S42. Within key production lines, based on the peak value of the time-series cross-correlation between the equipment anomaly index and the overall disturbance intensity of the production line, construct a causal response delay map to identify the core equipment that responds fastest to the disturbance in the production line.

[0044] After identifying the set of critical production lines, the core response equipment for disturbances within each production line is identified by analyzing the time series of the adaptive anomaly index of each piece of equipment in each critical production line and the time-series dynamic relationship between the overall disturbance intensity of that production line.

[0045] Cross-correlation analysis was used to calculate the time delays between the two. The correlation is shown in the following formula: , Let the combined disturbance intensity of the i-th device and the production line be within a time delay of . The correlation coefficient at that time. For the correlation coefficient function, Let be the adaptive anomaly index of the i-th device at time t. For the production line at all times The overall disturbance intensity. A positive value indicates that the equipment anomaly leads the overall disturbance intensity of the production line, while a negative value indicates that it lags behind.

[0046] For each device i, extract the delay corresponding to its maximum cross-correlation coefficient. And record the peak correlation value. All devices and Construct a causal response delay graph, with the horizontal axis representing response delay. The smaller the value, the earlier the response; the vertical axis represents the correlation strength. The higher the cross-correlation value, the stronger the impact. Prioritize equipment with strong correlation and early response. Select the equipment with the largest cross-correlation peak and the earliest response in the production line and identify them as core response equipment.

[0047] S43. By combining equipment topology weights and response delays, the longest impact path is traced in reverse to deduce the potential transmission chain from the source equipment to the factory-level imbalance.

[0048] After identifying the core response device, the possible propagation path of the disturbance is traced backward by combining the device topology and response timing information. Based on the production line device topology map established in step S21, the edge directions are reversed to form an impact tracing map. Starting from the current core response device, all possible upstream devices are searched along the reverse edges. The weight of each edge is... , These are the original topological weights. It is the difference between the abnormal peak time of equipment j and the peak time of the overall disturbance intensity of the production line. It is the difference between the abnormal peak time of equipment i and the peak time of the overall disturbance intensity of the production line. As a decay coefficient, this design assigns higher weights to paths of devices that exhibit anomalies earlier in time. Using the Dijkstra algorithm, paths from the core response device to all possible sources are searched in the impact sourcing graph. By combining the topological and time-weighted impact values ​​of each edge on the path and subtracting the length penalty, the cumulative impact score of each sourcing path is calculated. The path with the highest score is selected as the most likely disturbance propagation chain. This path represents the potential root cause propagation chain from local devices to system-level imbalances, serving as the focus for maintenance, optimization, or control interventions.

[0049] S50. Collect vibration spectrum data from core equipment along the path and verify it against historical baselines. Match the verification results with historical maintenance records. Based on the matching results and subsequent tracking, achieve adaptive updates of the rule system.

[0050] S51. For the core equipment along the simulation path, extract its high-frequency sampled vibration signal to confirm whether there is any quantifiable physical anomaly.

[0051] For critical equipment identified as core responses along the path, high-frequency sampling sensor data is automatically retrieved for refined physical characteristic analysis to confirm the presence of quantifiable physical anomalies. Combining the equipment's operating speed, structural parameters, and operating condition information, characteristic frequencies related to typical faults are extracted, and the energy level of the characteristic frequency bands and their time-varying trends are calculated in the frequency domain. By comparing this data with historical baseline data and statistical thresholds for normal operation of similar equipment, it is determined whether there is significant energy accumulation, increased harmonic components, or modulation phenomena. If the detected abnormal characteristics exceed a preset threshold, the equipment is determined to have physical degradation or operational anomalies, verifying it as the actual carrier of the current risk.

[0052] S52. Match the abnormal equipment data with historical maintenance records. If the match is successful, solidify the abnormal mode, fault type and handling strategy as a new rule and write it into the knowledge base.

[0053] The system intelligently compares the identified anomaly data with historical maintenance records. It extracts the vibration characteristic patterns, associated disturbance propagation paths, and operating conditions of the current abnormal equipment, and matches these against the existing anomaly pattern-fault type-response strategy triplet in the knowledge base. If highly similar historical cases are found, with consistent time, operating conditions, and response behavior, a successful match is determined. At this point, the current analysis path and related features are automatically solidified into a new diagnostic rule or an existing rule is updated, and along with recommended handling suggestions, are written into the knowledge base, achieving structured accumulation and reuse of experience.

[0054] S53. Adaptively update the rules based on subsequent tracking and actual verification.

[0055] For abnormal path rules already written into the knowledge base, their subsequent prediction performance is continuously tracked. If a rule triggers a risk warning for three consecutive warning cycles, but no corresponding maintenance or manually confirmed fault occurs in actual production, the prediction accuracy of the rule is considered to have decreased. An adaptive degradation mechanism is automatically activated to reduce the default weight of the transmission path in subsequent risk aggregation and tracing, and a penalty factor is introduced to gradually weaken its influence in future analysis. If a rule triggers a risk warning for three consecutive warning cycles and maintenance is frequent, the existing rule is maintained, equipment aging factors are assessed, and load reduction, rotation, or replacement is considered. If an anomaly suddenly occurs without any warning, it is necessary to check whether the rule is missing and supplement new knowledge in a timely manner.

[0056] Example 2 like Figure 2 As shown, Embodiment 2 of this application provides an industrial internet data three-level interconnection system, including: Data Acquisition and Anomaly Module: This module collects real-time data from multiple sources at the equipment level. Based on this multi-source data and the current equipment's process stage label, it generates an equipment anomaly index and triggers initial edge screening. Specifically, it consists of the following sub-modules: Acquisition Submodule: Acquires four types of signals for each device: vibration, temperature, current, and acoustic emission. Calculates the variance-normalized amplitude of each signal within a sliding window and generates the original abnormal potential energy value through weighted sum of squares.

[0057] Alignment Submodule: Introduces process stage labels, performs stage benchmark alignment for abnormal potential energy, eliminates the influence of operating condition drift, and outputs an abnormality index that adapts to operating conditions.

[0058] Anomaly and Alarm Submodule: Calculates the piecewise linear slope change rate and peak-to-valley ratio based on the anomaly index sequence. If both exceed the dynamic threshold simultaneously, it is determined to be a local mutation event, triggering an initial alarm at the edge layer.

[0059] The receiving and calculation module receives anomaly indices uploaded by each device at the production line level. Based on the anomaly indices of each device and their process dependencies, it calculates the overall disturbance intensity of the production line using dynamic weight aggregation and topology influence diffusion algorithms. Specifically, it consists of the following sub-modules: Constructing the topology submodule: Constructing the production line equipment topology diagram, setting initial connection weights based on the process flow direction and equipment output dependencies, and spreading the impact factors along the topology path when a certain device alarms.

[0060] Update submodule: Updates the operational contribution of each device, dynamically adjusts its weight in the aggregate calculation, and forms a real-time importance matrix.

[0061] The generation and identification submodule multiplies the abnormal index of all equipment with its dynamic weight and sums them to generate the comprehensive disturbance intensity of the production line. At the same time, it performs a Fourier transform on the sequence to extract the dominant frequency and identify whether the production line is in a periodic oscillation mode.

[0062] Risk warning module: Collects disturbance intensity sequences for each production line at the factory level, and calculates a coupled risk index using dynamic time warping, mode entropy, and energy consumption dispersion to provide system-level risk warnings. Specifically, it consists of the following sub-modules: Collection and Segmentation Submodule: Collect the disturbance intensity sequences of each production line, calculate the dynamic time warping distance between each pair, construct the production line behavior similarity matrix, and cluster and segment the operation mode groups.

[0063] Pattern Entropy and Discrete Submodule: Calculates the plant-wide operating pattern entropy based on the similarity matrix, reflecting the consistency of production line collaboration, and simultaneously calculates the discrete coefficient of energy consumption per unit output.

[0064] Coupling Risk Submodule: Constructs a coupling risk index. When the adaptive threshold is reached, it is determined to be a system-level imbalance, and a three-level through-response is initiated.

[0065] The reverse path deduction module: The factory coupling risk index, through marginal sensitivity analysis, obtains a set of critical production lines. Based on the equipment anomaly index and overall disturbance intensity of these production lines, potential root cause transmission paths are obtained through time-series cross-correlation and topological reverse deduction. Specifically, it is divided into the following sub-modules: Screening submodule: Defines risk sensitivity, estimates the marginal contribution of each production line to the total risk through numerical differentiation, and screens the contributions to obtain a set of key production lines.

[0066] Identification Submodule: Within key production lines, based on the peak value of the time-series cross-correlation between equipment anomaly index and the overall disturbance intensity of the production line, a causal response delay map is constructed to identify the core equipment that responds fastest to production line disturbances.

[0067] The tracking submodule combines device topology weights and response latency to trace the longest impact path in reverse, deducing the potential transmission chain from the source device to the factory-level imbalance.

[0068] Verification and Update Module: This module verifies the vibration spectrum collected from core equipment along the path against historical baselines. The verification results are then matched with historical maintenance records. Based on the matching results and subsequent tracking, the rule system is adaptively updated. Specifically, it consists of the following sub-modules: Extraction and Confirmation Submodule: For core equipment along the simulation path, extract its high-frequency sampled vibration signal to confirm whether there are any quantifiable physical anomalies.

[0069] Matching and updating submodule: Matches abnormal equipment data with historical maintenance records. If the match is successful, the abnormal pattern, fault type, and handling strategy are fixed as new rules and written to the knowledge base.

[0070] Verification Update Submodule: Adaptively updates rules based on subsequent actual verification.

[0071] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method and system for three-level data connectivity in the industrial internet, characterized in that, include: S10. Collect real-time multi-source data of the equipment at the equipment layer, and generate an equipment anomaly index based on the multi-source data and the current equipment process stage label, and trigger edge screening. S20. Receive the abnormal index uploaded by each device at the production line level, and calculate the overall disturbance intensity of the production line based on the abnormal index of each device and the process dependency relationship through dynamic weight aggregation and topology influence diffusion algorithm. S30. Collect disturbance intensity sequences of each production line at the factory level, and calculate the coupled risk index through dynamic time warping, mode entropy and energy consumption dispersion to conduct system-level risk early warning. S40. By coupling the risk index through marginal sensitivity analysis, a set of key production lines is obtained. Based on the equipment anomaly index and the comprehensive disturbance intensity of the production lines within the set, potential root cause transmission paths are obtained through time-series cross-correlation and topological inversion. S50. Collect vibration spectrum data from core equipment along the path and verify it against historical baselines. Match the verification results with historical maintenance records. Based on the matching results and subsequent tracking, achieve adaptive updates of the rule system.

2. The method for three-level data connectivity in the industrial internet as described in claim 1, characterized in that, The system collects real-time data from multiple sources at the equipment level. Based on this multi-source data and the current equipment's process stage label, it generates an equipment anomaly index and triggers edge screening. This process is divided into the following sub-steps: For each device, four types of signals are collected: vibration, temperature, current, and acoustic emission. The variance normalization amplitude within the sliding window is calculated for each signal, and the original abnormal potential energy value is generated by weighted sum of squares. By introducing process stage labels, abnormal potential energy is aligned with stage benchmarks, eliminating the influence of operating condition drift, and outputting an abnormal index that is adaptive to operating conditions. The piecewise linear slope change rate and peak-to-valley ratio are calculated based on the abnormal index sequence. If both exceed the dynamic threshold at the same time, it is determined to be a local mutation event, triggering an initial alarm at the edge layer.

3. The method for three-level data connectivity in the industrial internet as described in claim 1, characterized in that, The production line layer receives the anomaly indices uploaded by each device. Based on the anomaly indices of each device and their process dependencies, it calculates the overall disturbance intensity of the production line using dynamic weight aggregation and topology influence diffusion algorithms. This process is divided into the following sub-steps: Construct a production line equipment topology map, set initial connection weights based on the process flow direction and equipment output dependencies, and when a certain equipment alarms, spread the influence factors along the topology path; Update the operational contribution of each device, dynamically adjust its weight in the aggregate calculation, and form a real-time importance matrix; The abnormal index of all equipment is multiplied by its dynamic weight and then summed to generate the comprehensive disturbance intensity of the production line. At the same time, a Fourier transform is performed on the sequence to extract the dominant frequency and identify whether the production line is in a periodic oscillation mode.

4. The method for three-level data connectivity in the industrial internet as described in claim 1, characterized in that, At the factory level, disturbance intensity sequences for each production line are collected. A coupled risk index is calculated using dynamic time warping, mode entropy, and energy consumption dispersion to conduct system-level risk early warning. This process involves the following sub-steps: Collect the disturbance intensity sequences of each production line, calculate the dynamic time warping distance between each pair, construct the production line behavior similarity matrix, and cluster them into operation mode groups; The plant-wide operating mode entropy is calculated based on the similarity matrix to reflect the consistency of production line collaboration, and the dispersion coefficient of energy consumption per unit output is also calculated. A coupling risk index is constructed. When the adaptive threshold is reached, it is determined to be a system-level imbalance, and a three-level through-response is initiated.

5. The method for three-level data connectivity in the industrial internet as described in claim 1, characterized in that, The factory coupling risk index, through marginal sensitivity analysis, yields a set of critical production lines. Based on the equipment anomaly index and overall disturbance intensity of these production lines, potential root cause transmission paths are obtained through time-series cross-correlation and topological inversion. This process involves the following sub-steps: Define risk sensitivity, estimate the marginal contribution of each production line to the total risk through numerical differentiation, and screen the contribution to obtain the set of key production lines; Within key production lines, a causal response delay map is constructed based on the time-series cross-correlation peak value between equipment anomaly index and overall production line disturbance intensity to identify the core equipment that responds fastest to production line disturbances. By combining equipment topology weights and response latency, the longest impact path is traced in reverse to deduce the potential transmission chain from the source equipment to the factory-level imbalance.

6. The method for three-level data connectivity in the industrial internet as described in claim 5, characterized in that, By combining device topology weights and response latency, the longest impact path is traced in reverse to deduce the potential transmission chain from the source device to the factory-level imbalance, which is divided into the following sub-steps: Based on the reverse of the production line equipment topology diagram, an impact tracing diagram is generated. Based on the core response equipment in the diagram, all possible upstream devices are searched along the reverse edge. Update edge weights to give higher weights to the paths of devices that experienced anomalies earlier in time. Using Dijkstra's algorithm, we search for paths in the influence source graph and select the path with the highest cumulative influence score as the potential root cause transmission chain.

7. The method for three-level data connectivity in the industrial internet as described in claim 1, characterized in that, The vibration spectrum of the core equipment in the path is collected and verified against the historical baseline. The verification results are matched with historical maintenance records. Based on the matching results and subsequent tracking, the rule system is adaptively updated. The process is divided into the following sub-steps: For the core equipment along the simulation path, extract its high-frequency sampled vibration signals to confirm whether there are any quantifiable physical anomalies. Match the abnormal equipment data with historical maintenance records. If the match is successful, solidify the abnormal pattern, fault type and handling strategy as a new rule and write it into the knowledge base. The rules will be updated adaptively based on subsequent actual verification.

8. A three-level data interconnection system for the industrial internet, characterized in that, include: Data Acquisition and Anomaly Module: Collects real-time data from multiple sources at the equipment level, and generates an equipment anomaly index based on the multi-source data and the current equipment's process stage label, triggering edge screening. The receiving and calculation module receives the abnormality index uploaded by each device at the production line level, and calculates the overall disturbance intensity of the production line based on the abnormality index of each device and the process dependency relationship through dynamic weight aggregation and topology influence diffusion algorithm. Risk warning module: Collects disturbance intensity sequences of each production line at the factory level, and calculates a coupled risk index through dynamic time warping, mode entropy and energy consumption dispersion to conduct system-level risk warning; The reverse path module: The factory coupling risk index obtains a set of key production lines through marginal sensitivity analysis. Based on the equipment anomaly index and the comprehensive disturbance intensity of the production lines within the set, the potential root cause transmission path is obtained through time-series cross-correlation and topological reverse inference. Verification and update module: Verify the vibration spectrum collected from the core equipment in the path with the historical baseline, match the verification results with historical maintenance records, and realize the adaptive update of the rule system based on the matching results and subsequent tracking.

Images