An industrial data governance method and device

Abstract

The application relates to an industrial data governance method and device, and belongs to the technical field of industrial data processing. The method comprises the following steps: receiving multi-source heterogeneous data, and uniformly converting the multi-source heterogeneous data into standardized event data containing physical timestamps and logical serial numbers; utilizing a state machine to perform anti-noise identification on the standardized event data, dynamically adjusting the aggregation window width in combination with a bus load rate, and outputting an effective event stream; querying a pre-established business model dependency graph, determining a change source node according to the effective event stream, traversing the business model dependency graph in reverse from the change source node, and generating an event-orchestration routing table; and triggering business orchestration execution according to the event-orchestration routing table. The application solves the problems of multi-source data synchronization difficulty and business response fault, significantly reduces system maintenance cost, and improves response sensitivity and operation stability in a high-load scene.

Description

Technical Field

[0001] This application relates to the field of industrial data processing technology, and in particular to an industrial data governance method and apparatus. Background Technology

[0002] With the rapid development of smart manufacturing and industrial IoT technologies, the sources of data within factories are becoming increasingly diversified, covering various equipment and platforms such as CNC machine tools, sensors, upper-level management platforms, and production execution management systems.

[0003] Existing industrial data governance solutions typically employ general event-driven architectures or traditional data integration tools for data integration. In practice, data from different devices is generally collected via protocol adapters, converted into a unified format, and then published to a message queue. Downstream business systems then subscribe to relevant events and respond according to pre-configured rules. This model, to a certain extent, achieves data collection and distribution, meeting basic information management needs.

[0004] However, existing technologies have significant limitations when facing specific industrial scenarios such as mold making and injection molding. First, industrial environments present challenges due to the vast differences in timescales of multi-source, heterogeneous data. For example, millisecond-level high-frequency vibration signals coexist with hourly heat treatment furnace temperature curves. General frameworks struggle to automatically adapt to such cross-scale timescale conflicts, easily leading to data synchronization difficulties or delays in detecting critical changes. Second, high-frequency data can easily trigger "event storms," ​​causing excessive bus load or even system congestion, while low-frequency critical events may be buried, resulting in missed alarms. Finally, the dependencies between events and business orchestration typically rely on manual maintenance. When business models change, routing configurations need to be manually modified, resulting in high development and maintenance costs and a high risk of configuration omissions leading to gaps in business response. Summary of the Invention

[0005] This application provides an industrial data governance method and apparatus to solve the technical problems in related technologies, such as the difficulty in automatically adapting cross-scale timescale conflicts of multi-source heterogeneous data, the difficulty in balancing event storm suppression and critical event response under high load scenarios, and the reliance on manual configuration for business linkage, which can easily lead to response gaps.

[0006] Firstly, an industrial data governance method is provided, comprising: receiving multi-source heterogeneous data and uniformly converting it into standardized event data containing physical timestamps and logical sequence numbers; using a state machine to perform noise resistance identification on the standardized event data, dynamically adjusting the aggregation window width in conjunction with bus load rate, and outputting a valid event stream; querying a pre-established business model dependency graph, determining the change source node based on the valid event stream, traversing the business model dependency graph in reverse from the change source node, and generating an event-orchestration routing table; triggering business orchestration execution based on the event-orchestration routing table; calculating a governance health index, and correcting the aggregation window width based on the governance health index.

[0007] In some embodiments, the unified conversion into standardized event data containing physical timestamps and logical sequence numbers includes: performing feature extraction processing on high-frequency sampled data to obtain feature event data; performing linear interpolation padding processing on low-frequency sampled data to obtain padding event data; and encapsulating the feature event data and the padding event data into the standardized event data.

[0008] In some embodiments, the unified conversion into standardized event data containing physical timestamps and logical sequence numbers further includes: obtaining the logical sequence number of the multi-source heterogeneous data; storing the multi-source heterogeneous data in a rearrangement buffer; and sorting the data in the rearrangement buffer from smallest to largest according to the logical sequence number to obtain ordered data; wherein the feature extraction processing and the linear interpolation completion processing are performed based on the ordered data.

[0009] In some embodiments, the use of a state machine to perform noise-resistant identification on the standardized event data includes: setting a normal state, a watch state, an early warning state, and a recovery confirmation state; when the number of times the monitored value of the standardized event data exceeds the soft threshold consecutively reaches a preset consecutive number, the system jumps from the normal state to the watch state; when the duration reaches the confirmation window width, the system jumps from the watch state to the early warning state.

[0010] In some embodiments, the dynamic adjustment of the aggregated window width based on the bus load rate includes: obtaining the current bus load rate, the baseline window width, and the load sensitivity coefficient; calculating the sum of the baseline window width and the load adjustment amount to obtain an intermediate window value, wherein the load adjustment amount is the product of the baseline window width, the load sensitivity coefficient, and the current bus load rate; and limiting the intermediate window value between the minimum window width and the maximum window width to obtain the aggregated window width.

[0011] In some embodiments, the output of the valid event stream includes: establishing a high-frequency queue, a medium-frequency queue, and a low-frequency queue; allocating the noise-resistant identified events to the corresponding queues according to the event type; prioritizing the consumption of alarm events in the low-frequency queue to generate the valid event stream.

[0012] In some embodiments, querying the pre-established business model dependency graph, determining the change source node based on the valid event flow, and traversing the business model dependency graph backwards from the change source node to generate an event-orchestration routing table includes: constructing the business model dependency graph with business entities as nodes and dependencies as edges; traversing the business model dependency graph backwards from the change source node to determine the set of affected nodes; extracting the event types of interest to the set of affected nodes, and generating the event-orchestration routing table.

[0013] In some embodiments, the calculation of the governance health index includes: recording event reception timestamps and execution completion timestamps; calculating the time difference between the event reception timestamps and the execution completion timestamps; calculating a percentile latency value based on the distribution of the time difference; counting the total number of received events and the number of unacknowledged events, and calculating the event loss rate based on the ratio of the two; and using at least one of the percentile latency value and the event loss rate as the governance health index.

[0014] In some embodiments, the feedback correction of the aggregation window width includes: comparing the governance health index with a preset threshold; when the governance health index is greater than the preset threshold, increasing the adjustment step size of the baseline window width; and when the governance health index is less than the preset threshold, decreasing the adjustment step size of the baseline window width.

[0015] Secondly, an industrial data governance device is provided, comprising: an event access module for receiving multi-source heterogeneous data and uniformly converting it into standardized event data containing physical timestamps and logical sequence numbers; a change perception module for using a state machine to perform noise-resistant identification on the standardized event data, dynamically adjusting the aggregation window width in conjunction with bus load rate, and outputting a valid event stream; a model orchestration module for querying a pre-established business model dependency graph, determining the change source node based on the valid event stream, traversing the business model dependency graph backwards from the change source node, and generating an event-orchestration routing table; an execution feedback module for triggering business orchestration execution based on the event-orchestration routing table; and a governance feedback module for calculating a governance health index and correcting the aggregation window width based on the governance health index.

[0016] The beneficial effects of the technical solution provided in this application include: This application provides an industrial data governance method and apparatus. First, it achieves cross-scale timescale alignment of multi-source heterogeneous data by uniformly converting it into standardized event data, thus resolving time conflict issues. Second, it utilizes state machine noise immunity recognition combined with dynamic adjustment of the aggregation window width based on bus load rate, and introduces a governance health index feedback correction mechanism to form adaptive closed-loop control. This suppresses high-frequency event storms while ensuring that low-frequency critical events are not missed and that the system load is dynamically optimized. Finally, it automatically generates an event-orchestration routing table by querying the business model dependency graph and traversing backwards from the change source node, transforming the dependency relationship between events and business from manual maintenance to automatic derivation. Ultimately, it achieves a low-latency, low-load, high-reliability, and self-optimizing automatic closed-loop governance effect from industrial data changes to business execution, reducing maintenance costs and avoiding response gaps caused by configuration omissions. It realizes real-time and accurate perception of industrial data changes and automatic closed-loop governance of business logic, significantly improving system stability, response efficiency, and operational flexibility. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall method steps provided in the embodiments of this application; Figure 2 A detailed flowchart illustrating step S1 is provided for an embodiment of this application; Figure 3 A flowchart illustrating the state machine configuration in S2 is provided for an embodiment of this application. Figure 4 A schematic diagram provided for illustrating the adjustment of the reference window in S2, as an embodiment of this application; Figure 5 This is a schematic diagram provided in an embodiment of the present application to illustrate the queue rearrangement process in S2; Figure 6 A detailed flowchart illustrating step S3 is provided for an embodiment of this application. Figure 7 This is a detailed flowchart illustrating S5, provided for an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] This application provides an industrial data governance method and apparatus that can solve the technical problems in related technologies, such as the difficulty in automatically adapting cross-scale timescale conflicts of multi-source heterogeneous data, the difficulty in balancing event storm suppression and critical event response under high load scenarios, and the reliance on manual configuration for business linkage which can easily lead to response gaps.

[0021] Reference Figure 1-7 An industrial data governance method, comprising: S1: Receives multi-source heterogeneous data and converts it into standardized event data containing physical timestamps and logical sequence numbers. Multi-source heterogeneous data refers to raw data from different levels, protocols, and time scales in the industrial field, including but not limited to: equipment-level data, operating status data from CNC machine tools, injection molding machines, and PLC controllers, and physical quantity monitoring data from sensors. This type of data typically features high frequency and millisecond-level sampling. System-level data: business data from manufacturing execution systems, enterprise resource planning, or quality management systems, such as work order status, material batches, and heat treatment furnace temperature curves. This type of data typically features low frequency, hourly-level, or event-triggered characteristics. Heterogeneity is manifested in the heterogeneity of protocol formats, time scales, and semantic structures among the data. The receiving process includes, but is not limited to: configuring multiple protocol adapters within the data access module, each listening to different network ports or message middleware topics.

[0022] The process of uniformly converting the data into standardized event data containing physical timestamps and logical sequence numbers includes: S101: performing feature extraction processing on high-frequency sampled data to obtain feature event data; S102: performing linear interpolation and padding processing on low-frequency sampled data to obtain padding event data; S103: encapsulating the feature event data and padding event data into standardized event data.

[0023] The data access module first identifies the sampling frequency attribute of the data to be processed. For high-frequency sampling data at the millisecond level, such as the spindle vibration signal of a CNC machine tool, feature extraction processing is used to extract key feature values ​​that can effectively characterize the operating status of the equipment, and compress and reduce them into feature event data to avoid high-frequency redundant data directly impacting subsequent processing links. For low-frequency sampling data at the hour level, such as the temperature curve of a heat treatment furnace, linear interpolation completion processing is used to calculate the state at intermediate logical moments based on the data values ​​of adjacent sampling points, generating continuous completion event data, thereby eliminating data gaps and timescale conflicts caused by excessively large physical sampling intervals. Finally, the feature event data and completion event data that have undergone the above differentiated processing are encapsulated into a unified format data packet and bound with the corresponding precise physical timestamp to form standardized event data. This enables heterogeneous industrial data at different time scales to undergo subsequent change perception and business orchestration under the same logical time scale, realizing the normalization of multi-source data on the logical time axis.

[0024] As a further optimization of the above conversion process, this embodiment also includes a data rearrangement step to ensure that feature extraction processing and linear interpolation completion processing are performed based on ordered data. Specifically, the data access module temporarily stores the received multi-source heterogeneous data into a rearrangement buffer. This buffer serves as an intermediate cache area to accommodate data packets arriving within a preset time window. It then retains the logical sequence number assigned by the data source, which represents the global order of data generation. The processing unit sorts the data in the rearrangement buffer from smallest to largest according to the logical sequence number, rearranging data whose physical arrival order may be disordered into an ordered sequence that conforms to the logical generation order. After sorting, this ordered data is used as the input data for steps S101 to S103, thereby ensuring that the downstream state machine identification and business orchestration logic can be executed based on the correct timing, avoiding misjudgments of state or business logic conflicts caused by data inversion.

[0025] S2: Utilize a state machine to perform noise-resistant identification on standardized event data, dynamically adjust the aggregation window width based on bus load rate, and output a valid event stream. The noise-resistant identification of standardized event data using a state machine includes: S201: Configure the state machine with normal, watch, warning, and recovery confirmation states; S202: When the monitored value of standardized event data exceeds the soft threshold for a preset number of consecutive times, the state machine transitions from the normal state to the watch state; S203: When the duration reaches the confirmation window width, the state machine transitions from the watch state to the warning state.

[0026] In this embodiment, the specific implementation process of using a state machine to perform noise-resistant identification of standardized event data and dynamically adjusting the aggregation window width in conjunction with bus load rate is as follows: First, a multi-level state machine is configured, including normal state, attention state, warning state, and recovery confirmation state. The system reads the monitoring values ​​of standardized event data in real time. When the number of times the monitoring value exceeds the preset soft threshold reaches the preset consecutive number, it is determined that there is an abnormal trend in the data. The state machine is controlled to jump from the normal state to the attention state. At this time, no business response is triggered temporarily to observe subsequent changes. If the duration of the attention state reaches the confirmation window width, the abnormality is confirmed to be valid. The state machine is controlled to jump from the attention state to the warning state and marked as a valid event, thereby filtering out false alarms caused by transient interference.

[0027] In addition, the aggregation window width is dynamically adjusted in conjunction with the bus load rate, including: S211: obtaining the current bus load rate, the reference window width, and the load sensitivity coefficient; S212: calculating the sum of the reference window width and the load adjustment amount to obtain the intermediate window value, where the load adjustment amount is the product of the reference window width, the load sensitivity coefficient, and the current bus load rate; S213: limiting the intermediate window value between the minimum window width and the maximum window width to obtain the aggregation window width.

[0028] First, the current bus load rate, the preset baseline window width, and the load sensitivity coefficient, which characterizes the impact of load changes on the window, are acquired in real time. Then, the sum of the baseline window width and the load adjustment amount is calculated to obtain the intermediate window value, where the load adjustment amount is the product of the baseline window width, the load sensitivity coefficient, and the current bus load rate. This achieves dynamic mapping of the window width with the bus load state. Next, the intermediate window value is limited between the preset minimum and maximum window widths. If the intermediate window value is less than the minimum window width, it is forcibly adjusted to the minimum window width; if it is greater than the maximum window width, it is forcibly adjusted to the maximum window width. Finally, the aggregated window width is obtained. This limiting process avoids the aggregated window from becoming too small and losing its buffering effect due to extreme fluctuations in the load rate, or from becoming too large and introducing excessive latency. It ensures that the dynamically adjusted window width is always within a safe and effective range, achieving a balance between system throughput and real-time performance.

[0029] In addition, this step outputs a valid event stream, including: S221: establishing a high-frequency queue, a medium-frequency queue, and a low-frequency queue; S222: assigning the noise-recognized events to the corresponding queues according to the event type; S223: prioritizing the consumption of alarm events in the low-frequency queue to generate a valid event stream.

[0030] First, a three-level independent caching channel is established, consisting of a high-frequency queue, a medium-frequency queue, and a low-frequency queue. Events, after noise reduction and identification, are assigned to the corresponding queues based on their event type, ensuring that data with different characteristics are physically isolated and stored. Then, during the event consumption phase, a priority scheduling strategy is adopted to prioritize the consumption of alarm events in the low-frequency queue, generating an effective event stream. Even under high-concurrency and high-frequency data influx, it can ensure that low-frequency critical alarm events are not blocked and can be processed with priority, thereby avoiding delays in critical business response and achieving the diversion and precise delivery of events with different priorities.

[0031] S3: Query the pre-established business model dependency graph, determine the change source node based on the valid event flow, and traverse the business model dependency graph in reverse from the change source node to generate an event-orchestration routing table. This step specifically includes: S31: Construct a business model dependency graph with business entities as nodes and dependencies as edges; S32: Starting from the source node of the change, traverse the business model dependency graph in reverse to determine the set of affected nodes; S33: Extract the event types of interest to the affected node set and generate an event-orchestration routing table.

[0032] First, the model orchestration module constructs a business model dependency graph with business entities as nodes and dependencies as edges. This graph structurally stores the logical connections and dependency directions between various business units in the industrial production process. When a data change event is included in the valid event stream, the system identifies the change source node corresponding to the event. Starting from the change source node, it traverses the business model dependency graph backward, searching for all downstream business entities that directly or indirectly depend on the source node along the reverse path of the dependency relationship, thereby determining the set of affected nodes. Subsequently, it extracts the event types that each node in the affected node set is interested in, maps and associates the event types with the corresponding business orchestration logic, and generates an event-orchestration routing table. This routing table is used to guide subsequent event distribution and business triggering. In this way, the system can automatically deduce event routing paths based on business dependencies without the need for manual configuration of subscription rules one by one. When the business model changes, only the dependency graph needs to be updated to automatically refresh the routing table, effectively reducing maintenance costs and avoiding business response gaps caused by manual configuration omissions.

[0033] S4: Trigger business orchestration execution based on the event-orchestration routing table. In this application, the orchestration execution engine reads event objects from the valid event stream in real time, extracts the event type identifier and change source node information, and queries the event-orchestration routing table. It then retrieves the routing entry matching the event type identifier and obtains the associated business orchestration logic script or workflow definition. Subsequently, the engine instantiates the business orchestration logic and calls the preset execution interface to trigger relevant business actions. These actions include, but are not limited to, adjusting downstream device operating parameters, generating maintenance work orders, sending early warning notifications, or updating the status records of the management system. Through this automatic triggering mechanism based on the routing table, the system can automatically execute the response actions of affected business nodes according to the relationships determined by the business model dependency graph without requiring manual hard-coded event consumption logic. This achieves data-driven automated business governance, ensuring the accuracy and timeliness of business responses and avoiding business response gaps caused by manual configuration omissions.

[0034] S5: Calculate the governance health index and adjust the aggregation window width based on the feedback from the governance health index; wherein, the calculation of the governance health index includes: S50: Record the event reception timestamp and execution completion timestamp; S51: Calculate the time difference between the event reception timestamp and the execution completion timestamp; S52: Calculate percentile delay values ​​based on the distribution of time differences; S53: Calculate the event loss rate based on the ratio of the total number of received events to the number of unacknowledged events; S54: Use at least one of percentile latency and event loss rate as a governance health indicator.

[0035] During the execution of business orchestration, the event reception timestamp and execution completion timestamp for each event are recorded. The time difference between the event reception timestamp and the execution completion timestamp is calculated. Subsequently, based on the distribution of the time difference, the percentile latency value is calculated, the total number of received events and the number of unacknowledged events are counted, and the event loss rate is calculated based on the ratio of the two. Finally, at least one of percentile latency and event loss rate is used as a governance health indicator, where percentile latency is used to characterize the system's response performance in most cases, and event loss rate is used to characterize the system's reliability.

[0036] Percentile latency refers to the latency value at a specific percentile position, such as the 95th percentile, corresponding to the processing time of all successfully processed events sorted from smallest to largest within a statistical period. For example, a P95 latency value of 100ms means that 95% of events take less than or equal to 100ms to process, with only 5% of events taking longer than this value. This metric is chosen instead of the "average latency value" because the average is easily affected by a few extremely fast or slow outliers and cannot reflect the true level; this metric is chosen instead of the "maximum latency value" because the maximum value is easily distorted by occasional network jitter or hardware anomalies. Therefore, percentile latency value can eliminate the interference of extreme outliers, objectively characterize the system's response performance under most normal conditions, ensure that the feedback mechanism is adjusted based on the system's normal performance, and avoid frequent oscillations of the aggregation window due to occasional fluctuations.

[0037] Event loss rate refers to the proportion of events that the system failed to process or accidentally discarded within a statistical period, out of the total number of events received. Event loss is typically caused by buffer overflows, processing timeouts, or system overload. This metric directly reflects the completeness and reliability of the data governance process. A high loss rate means that critical industrial data, such as alarm signals, may be missed, leading to gaps in business response. Therefore, the event loss rate is used to characterize system reliability. During feedback and correction processes, if the loss rate exceeds a threshold, the system must prioritize expanding the aggregation window or reducing the load to prevent data loss and ensure the accuracy and security of industrial data governance.

[0038] In addition, the feedback correction of the aggregation window width includes: comparing the governance health index with a preset threshold; When the governance health indicator is greater than the preset threshold, the baseline window width is increased by the adjustment step; when the governance health indicator is less than the preset threshold, the baseline window width is decreased by the adjustment step. Specifically: when the governance health indicator is a percentile latency value, the preset threshold is the preset latency threshold; when the governance health indicator is an event loss rate, the preset threshold is the preset loss rate threshold; when the governance health indicator includes both percentile latency value and event loss rate, the system normalizes the percentile latency value and event loss rate to eliminate dimensional differences, and calculates a comprehensive score value through weighted summation, in which case the preset threshold is the preset score threshold.

[0039] The calculated governance health indicators are compared with preset thresholds to quantitatively assess whether the current system latency exceeds the acceptable range for business operations. When using a single indicator, for example, if the percentile latency value is greater than the preset latency threshold, it indicates that the system's current processing latency is too high, with a risk of congestion or untimely response. In this case, the system increases the baseline window width by a preset adjustment step, expanding the aggregation window to merge more events and reduce the frequency of business triggers per unit time, thereby reducing system load and alleviating latency pressure. When the percentile latency value is less than the preset latency threshold, it indicates that the system currently has surplus resources and strong processing capacity. In this case, the system decreases the baseline window width by a preset adjustment step, shrinking the aggregation window to reduce event waiting time and improve the real-time performance of data processing. When using a combined indicator, the system compares the calculated comprehensive score value with a preset score threshold. If the comprehensive score value is greater than the preset score threshold, it indicates that the overall system health is poor, and the baseline window width is also increased to protect system stability. This dynamic adjustment mechanism based on actual execution delay feedback enables the aggregation window width to adapt to the system's operating state, ensuring that the system can maximize its response speed while meeting latency requirements under different load conditions, thus achieving stable closed-loop optimization of system performance.

[0040] The implementation principle of this application embodiment is as follows: First, multi-source heterogeneous data is unified into ordered standardized event data through time normalization and rearrangement buffering mechanisms, solving the problems of cross-scale timescale conflicts and data out-of-order. Second, a multi-level state machine is used for noise resistance identification, and the aggregation window width is dynamically adjusted in combination with bus load rate to effectively filter noise and balance system load and real-time performance, preventing event storms. Third, an event-orchestration routing table is automatically generated based on the reverse topology traversal of the business model dependency graph, realizing the automatic derivation and decoupling of business linkage relationships and eliminating manual configuration dependencies. Finally, the aggregation window is dynamically corrected through feedback closed loop of governance health indicators to ensure that the system operates in the optimal state. Its beneficial effects are reflected in significantly reducing data synchronization latency and critical event false negative rate, greatly reducing business system CPU usage and event storm risk, while reducing the development and maintenance cost of business linkage from man-days to hours, realizing real-time and accurate perception of industrial data changes and automatic closed-loop governance of business logic, and improving system stability, response efficiency and operational flexibility.

[0041] Reference Figure 1-7An industrial data governance device includes: an event access module, a change sensing module, a model orchestration module, an execution feedback module, and a governance feedback module. The event access module receives multi-source heterogeneous data and converts it into standardized event data containing physical timestamps and logical sequence numbers. The change sensing module uses a state machine to perform noise-resistant identification on the standardized event data, dynamically adjusts the aggregation window width based on bus load rate, and outputs a valid event stream. The model orchestration module queries a pre-established business model dependency graph, determines the change source node based on the valid event stream, and traverses the business model dependency graph backwards from the change source node to generate an event-orchestration routing table. The execution feedback module triggers business orchestration execution based on the event-orchestration routing table. The governance feedback module calculates governance health indicators and adjusts the aggregation window width based on these indicators. The industrial data governance device and industrial data governance methods address the same technical problems and employ the same technical means; therefore, further details are omitted here.

[0042] In the description of this application, it should be noted that the terms "upper," "lower," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication between two elements. For those skilled in the art, the specific meaning of the above terms in this application can be understood according to the specific circumstances.

[0043] It should be noted that in this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0044] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An industrial data governance method, characterized in that, include: It receives heterogeneous data from multiple sources and converts it into standardized event data containing physical timestamps and logical sequence numbers. The standardized event data is subjected to noise-resistant identification using a state machine, and the aggregation window width is dynamically adjusted in combination with the bus load rate to output a valid event stream. Query the pre-established business model dependency graph, determine the change source node based on the effective event flow, traverse the business model dependency graph in reverse from the change source node, and generate an event-orchestration routing table. Service orchestration execution is triggered based on the aforementioned event-orchestration routing table; Calculate the governance health index, and adjust the width of the aggregation window based on the feedback from the governance health index.

2. The industrial data governance method according to claim 1, characterized in that, The unified conversion into standardized event data containing physical timestamps and logical sequence numbers includes: Feature extraction processing is performed on high-frequency sampled data to obtain feature event data; Linear interpolation is performed on the low-frequency sampled data to obtain the completed event data; The feature event data and the completion event data are encapsulated into the standardized event data.

3. The industrial data governance method according to claim 2, characterized in that, The unified conversion into standardized event data containing physical timestamps and logical sequence numbers also includes: Obtain the logical sequence number of the multi-source heterogeneous data; The multi-source heterogeneous data is stored in a rearrangement buffer, and the data in the rearrangement buffer is sorted from smallest to largest according to the logical sequence number to obtain ordered data. The feature extraction process and the linear interpolation completion process are performed based on the ordered data.

4. The industrial data governance method according to claim 1, characterized in that, The step of using a state machine to perform noise-resistant identification on the standardized event data includes: Set the status to normal, watch, alert, and recovery confirmation; When the number of times the monitored value of the standardized event data exceeds the soft threshold consecutively reaches a preset number of consecutive times, the system switches from the normal state to the attention state. When the duration reaches the confirmation window width, the system transitions from the attention state to the warning state.

5. The industrial data governance method according to claim 1, characterized in that, The dynamic adjustment of the aggregation window width based on bus load rate includes: Get the current bus load rate, baseline window width, and load sensitivity coefficient; The sum of the baseline window width and the load adjustment amount is calculated to obtain the intermediate window value, wherein the load adjustment amount is the product of the baseline window width, the load sensitivity coefficient, and the current bus load rate; The aggregated window width is obtained by limiting the intermediate window value between the minimum window width and the maximum window width.

6. The industrial data governance method according to claim 1, characterized in that, The output valid event stream includes: Establish high-frequency queues, medium-frequency queues, and low-frequency queues; The noise-reduced events are assigned to the corresponding queues based on their event type. Prioritize consuming alarm events in the low-frequency queue to generate the valid event stream.

7. The industrial data governance method according to claim 1, characterized in that, The query pre-established business model dependency graph is used to determine the change source node based on the valid event flow. Starting from the change source node, the business model dependency graph is traversed in reverse to generate an event-orchestration routing table, including: Construct the business model dependency graph with business entities as nodes and dependencies as edges; Starting from the source node of the change, traverse the business model dependency graph in reverse to determine the set of affected nodes; Extract the event types that the affected node set is interested in, and generate the event-orchestration routing table.

8. The industrial data governance method according to claim 5, characterized in that, The calculation of the governance health index includes: Record the event reception timestamp and execution completion timestamp; Calculate the time difference between the event reception timestamp and the execution completion timestamp; Calculate the percentile delay value based on the distribution of the time difference; The total number of received events and the number of unacknowledged events are counted, and the event loss rate is calculated based on the ratio of the two. At least one of percentile latency and event loss rate is used as the governance health indicator.

9. The industrial data governance method according to claim 8, characterized in that, The feedback corrects the width of the aggregation window, including: The governance health index is compared with a preset threshold. When the governance health index is greater than the preset threshold, the width of the baseline window will be increased by the adjustment step. When the governance health index is less than the preset threshold, the width of the baseline window is reduced by the adjustment step.

10. An industrial data governance device, characterized in that, include: The event access module is used to receive heterogeneous data from multiple sources and convert it into standardized event data containing physical timestamps and logical sequence numbers. The change perception module is used to perform noise-resistant identification on the standardized event data using a state machine, dynamically adjust the width of the aggregation window in conjunction with the bus load rate, and output a valid event stream. The model orchestration module is used to query the pre-established business model dependency graph, determine the change source node based on the effective event flow, traverse the business model dependency graph in reverse from the change source node, and generate an event-orchestration routing table. The execution feedback module is used to trigger service orchestration execution based on the event-orchestration routing table; The governance feedback module is used to calculate the governance health index and adjust the width of the aggregation window based on the governance health index.

Images