Copper foil production monitoring method and system based on data storage optimization

Abstract

The invention relates to the technical field of industrial data processing, and discloses a copper foil production monitoring method and system based on data storage optimization, and the method comprises the following steps: carrying out the periodic aggregation of high-frequency data when a production process is in a steady state, and carrying out the calculation of a newly recognized production process event when the production process is monitored to deviate from the steady state. Whether causal association exists or not is judged by calculating the correlation between the new event and the current activity event in time and physical dimensions, and if the association exists, a chained event ID inheriting a precursor event identifier is generated for the new event; if the event is an independent event, a brand-new root event ID is generated, context data before the event occurs is extracted in a backtracking mode, and the generated event ID is marked on a data set in a unified mode in combination with real-time data during the event occurrence period for permanent storage. According to the method, the storage configuration is obviously optimized, and the efficiency of fault causal chain tracing is improved through the structured chain ID.

Description

Technical Field

[0001] This invention relates to the field of industrial data processing technology, specifically to a copper foil production monitoring method and system based on data storage optimization. Background Technology

[0002] In the production processes of precision manufacturing industries such as electrolytic copper foil, deploying a large number of sensors to monitor key process parameters (such as electrolyte temperature, main roller current, electrolytic cell voltage, and foil tension) at high frequency in real time has become an indispensable technical means to ensure the quality stability of the final product and the safety of the process. These sensors continuously generate massive amounts of time-series data at millisecond or second-level frequencies, providing a foundation for transparent management and refined control of the production process.

[0003] However, how to economically and efficiently process and store these high-frequency data streams is a core technological challenge currently facing the industry. One direct and seemingly comprehensive approach is to indiscriminately store all high-frequency data generated by all sensors in a full and permanent manner. But this strategy quickly leads to an explosive growth in data volume, placing a heavy burden on enterprise storage hardware costs, data center energy consumption, and network transmission bandwidth. More importantly, massive amounts of redundant data can severely slow down database query and analysis performance, making it difficult for engineers to respond quickly and inefficient in data retrieval when troubleshooting or optimizing processes.

[0004] To avoid the high costs of full data storage, the industry has generally adopted a compromise: downsampling and storing high-frequency data at fixed time intervals. For example, the system might only record the average or instantaneous value of all sensor readings once per minute. While this method significantly reduces data storage, it comes at the cost of data fidelity and integrity. Many transient anomalies in the production process, such as a brief current surge or a rapid pressure fluctuation, may last much shorter than the downsampling interval. These fleeting details are crucial precursors to major equipment failures or batch quality issues in products. If this information is missed due to the sampling interval, technicians will struggle to accurately trace the root cause when a serious failure occurs later due to a lack of complete, high-resolution data on the event's evolution.

[0005] Furthermore, existing data storage systems typically lack an understanding of the inherent logic of the production process when organizing data. Even if a system captures multiple consecutive abnormal data points, they are merely isolated records arranged in time-stamp order in the database. The inherent causal chain of a complex fault process—from a small disturbance gradually fermenting and spreading to ultimately triggering a chain reaction—is not identified at the structural level of data storage. This leads to fault diagnosis heavily relying on the personal experience of senior engineers, who must spend a significant amount of time manually comparing and correlating vast amounts of data from different sensors and time periods in an attempt to reconstruct the logic of the event. This process is not only inefficient but also highly error-prone. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a copper foil production monitoring method and system based on data storage optimization. This solves the contradiction between the high cost of full storage of high-frequency data and the loss of key information during downsampling storage in industrial monitoring, as well as the problem that traditional data storage methods are difficult to effectively trace the cause-and-effect chain of faults.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a copper foil production monitoring method based on data storage optimization, comprising the following steps: Acquire high-frequency sensor data during the copper foil production process; Real-time monitoring of the high-frequency sensor data to determine whether new production process events have occurred; If a new production process event is detected, an event-driven high-resolution storage process is executed, which includes: a) Perform a correlation assessment between the new production process event and the currently active production process event to determine whether there is a causal relationship between the new production process event and any of the active production process events; b) Based on the results of the correlation assessment, when a causal relationship exists, a chain event ID is generated for the new production process event; when no causal relationship exists, an independent root event ID is generated for the new production process event. c) Temporarily store the high-frequency sensor data within the time period corresponding to the new production process event, and store the high-frequency sensor data after associating it with the chain event ID or the root event ID; if no new production process event is detected, under steady-state conditions, aggregate the high-frequency sensor data to generate and store low-resolution baseline data.

[0008] Through the above approach, this invention establishes a dual-modal data storage strategy. Under stable production conditions, only low-resolution aggregated data is stored to conserve storage resources; once a potential process event is detected, it immediately switches to an event-driven, high-resolution storage mode. The core innovation of this mode lies in the fact that it does not record each event in isolation, but rather constructs a causal chain between events through correlation assessment and indexes the data using chained IDs. This structured storage method can completely preserve high-resolution data along the entire evolutionary path from the initiation of a fault to its eventual occurrence without storing all the original data, thus providing logically clear and informationally complete data support for subsequent fault tracing.

[0009] In one specific embodiment, the correlation assessment step in step a) includes: calculating temporal similarity based on the temporal proximity between the new production process event and the production process event in the active state; and calculating variable similarity based on the set of sensor variables associated with the new production process event and the production process event in the active state. The final correlation score is obtained by weighting the time similarity and the variable similarity, and the existence of the causal relationship is determined based on the correlation score.

[0010] ; in, This indicates the final comprehensive metric or overall score. For time similarity, For variable similarity, and These are the preset weighting coefficients.

[0011] ; in, This is the formula for calculating similarity, which uses an exponential function. It is a parameter or weight that reflects the degree to which the time interval affects the similarity. Indicates the start time of the new time period. Indicates the end time of the activity period. This represents the difference between the start time of the new time period and the end time of the activity period.

[0012] The similarity of the variables is calculated using the Jaccard index, which represents the ratio of the intersection to the union of the sensor variable sets associated with two events. The formula is as follows: ; in, It is a set of sensor variables associated with events in the new production process. It is a set of sensor variables associated with events in the production process and their active states. Represents a set and The size of the intersection, Represents a set and The size of the union of the sets.

[0013] Preferably, step c) of temporarily storing high-frequency sensor data specifically includes: retrospectively extracting high-frequency sensor data within a preset retrospective time length prior to the start time of the new production process event; synchronously tracking real-time high-frequency sensor data during the continuous active state of the new production process event; and uniformly storing the retrospectively extracted high-frequency sensor data and the synchronously tracked high-frequency sensor data after associating them with the chain event ID or the root event ID.

[0014] In one implementation, a variable storage priority is defined for each high-frequency sensor data associated with the chained event ID stored in step c); and when the severity level of any production process event in the event chain corresponding to the chained event ID is determined to be a preset abnormal level, the storage priority of the data associated with all preceding events in the event chain is retrospectively increased.

[0015] Preferably, in step a), any of the production process events is defined as an event object containing the following attributes: event start time, event end time, associated set of sensor variables, and event severity level.

[0016] In one specific embodiment, the method further includes a query step: in response to receiving a query request, the query request containing an event ID of the final production process event; based on the chain structure of the event ID, retrieving all data records with the event ID prefix in the event ID field, and combining the retrieved data records into a complete high-frequency sensor dataset and returning it, the high-frequency sensor dataset containing all data associated with the final production process event and all its preceding events.

[0017] Preferably, the aggregation operation of the high-frequency sensor data includes at least one of the mean, variance, maximum value, or minimum value.

[0018] In one implementation, the step of determining whether a new production process event has occurred is achieved by monitoring at least one of the following data patterns: univariate anomalies based on statistical process control rules, changes in cross-correlation among multiple variables, or changes in the frequency domain characteristics of signals.

[0019] A second aspect of the present invention provides a copper foil production monitoring system based on data storage optimization, the system being used to perform the methods described in any of the foregoing embodiments.

[0020] The copper foil production monitoring system includes: The data monitoring module is used to acquire and monitor high-frequency sensor data in real time during the copper foil production process to determine whether a production process event has occurred. An event handling module is configured to be activated in response to the data monitoring module determining that a new production process event has occurred. The event handling unit includes: The causal analysis module is used to assess the correlation between the new production process event and the currently active production process event to determine whether there is a causal relationship between them. The event ID generation module is used to generate chained event IDs or independent root event IDs based on the determination result of the causal analysis module. A high-resolution storage module is used to temporarily store high-frequency sensor data within the time period corresponding to the new production process event, and to store the high-frequency sensor data in association with the event ID generated by the event ID generation module. The baseline storage module is configured to aggregate the high-frequency sensor data and generate and store low-resolution baseline data under steady-state conditions where the data monitoring module has not determined that a production process event has occurred.

[0021] This invention provides a method and system for monitoring copper foil production based on data storage optimization. It has the following beneficial effects: 1. This invention establishes a dynamic, on-demand storage resource allocation mechanism by setting low-resolution baseline data storage under steady-state conditions and an event-driven high-resolution storage process that is triggered only when a production process event is detected. This method of dynamically switching storage strategies according to process status avoids the continuous storage of all high-frequency data, thereby achieving effective utilization of storage resources without losing abnormal transient details.

[0022] 2. This invention assesses the correlation between newly occurring production process events and events in the activity state to determine whether there is a causal relationship. Based on this, it generates chained event IDs to index high-resolution data. This overcomes the shortcomings of traditional monitoring methods, where event records are isolated and cannot reflect the logic of fault development. This ensures that the high-resolution data of the entire process, from the precursor event in the fault initiation stage to the final abnormal outbreak, is structured, correlated and saved, providing complete contextual information for root cause analysis.

[0023] 3. Based on the constructed chain-like event ID structure, this invention allows analysts to retrieve all data records with the same ID prefix by providing only the event ID of the final abnormal event with a single query request. This method can directly return a high-resolution dataset that is logically complete, temporally continuous, and covers the entire fault evolution chain, avoiding inefficient searches in massive amounts of irrelevant data, thereby greatly shortening the fault diagnosis cycle and improving the accuracy of cause localization. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the system functional modules of the present invention; Figure 2 This is a schematic diagram of the baseline data generation and storage process under steady-state conditions according to the present invention; Figure 3 This is a schematic diagram of the production process event identification and instantiation process of the present invention; Figure 4 This is a schematic diagram of the event causal chain construction and ID generation process of the present invention; Figure 5 This is a schematic diagram of the high-resolution data extraction and associated storage process of the present invention; Figure 6 This is a schematic diagram of the storage priority traceability enhancement process of the present invention. Detailed Implementation

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0026] See attached document Figure 1 The copper foil production monitoring system based on data storage optimization provided by the present invention may include: a data monitoring module, an event processing module, and a baseline storage module.

[0027] The data monitoring module is used to acquire high-frequency sensor data from various sensors at the copper foil production site and monitor the high-frequency sensor data in real time to determine whether new production process events have occurred.

[0028] The baseline storage module is connected to the data monitoring module. Under steady-state conditions where the data monitoring module has not detected any new production process events, the baseline storage module is configured to perform aggregation operations on high-frequency sensor data, generate low-resolution baseline data, and then store it.

[0029] An event handling module is configured to be activated when the data monitoring module determines that a new production process event has occurred. In one specific implementation, the event handling module may include: a cause-effect analysis module, an event ID generation module, and a high-resolution storage module.

[0030] The overall data processing flow of this invention is as follows: The raw high-frequency sensor data stream acquired from the sensor is input to the data monitoring module. The data monitoring module continuously analyzes the data stream to determine the current production process status.

[0031] When the production process is in a steady state, meaning the data monitoring module has not detected any production process events that meet preset conditions, the data stream is directed to the baseline storage module. The baseline storage module aggregates the data at fixed time intervals and writes the aggregated low-resolution baseline data to permanent storage. This process corresponds to a conventional, low-overhead data recording mode.

[0032] When the data monitoring module detects a new production process event, the system executes an event-driven, high-resolution storage process. At this point, the event processing module is activated. First, the causal analysis module begins, assessing the correlation between the new production process event and all currently active production process events in the system. This assessment determines whether the new production process event occurred independently or is a continuation or evolution of an existing event.

[0033] Correlation assessment can be based on a quantified correlation score. To complete this, the calculation method is as follows: ; in, This indicates the final comprehensive metric or overall score. For time similarity, For variable similarity, and These are the preset weighting coefficients.

[0034] Next, the event ID generation module generates an event ID for the new production process event based on the judgment result output by the causal analysis module. If the judgment result indicates that a causal relationship exists, a chained event ID is generated, which is an extension of its predecessor event ID; if the judgment result indicates that no causal relationship exists, a completely new and independent root event ID is generated.

[0035] Finally, the high-resolution storage module is triggered. This module retrieves all high-frequency sensor data that is temporally relevant to the new production process event from the memory buffer, associates this data with the chained event IDs or root event IDs generated by the event ID generation module, and then writes the associated high-resolution data records to the permanent storage area. This process ensures that all detailed information related to the event is completely preserved.

[0036] See attached document Figure 2 When the data monitoring module analyzes the real-time high-frequency sensor data stream and does not identify any production process events that meet preset conditions, the system confirms that the production process is in a steady state. Under this condition, the data stream is continuously transmitted to the baseline storage module to execute a low-resolution baseline data storage process.

[0037] The core function of the baseline storage module is to convert continuous, high-frequency time-series data into discrete, low-resolution data points. To achieve this, the module first divides the time axis into continuous and non-overlapping aggregated time windows, with each window having a preset duration. In a specific implementation, The value can be set to 1 minute, 5 minutes, or other values ​​determined according to the specific stage characteristics of the copper foil production process. For any aggregation time window, the baseline storage module collects all high-frequency sensor data points arriving within that time period.

[0038] At the end of each aggregation time window, the baseline storage module performs aggregation operations on the set of data points collected within the window. This is done using a specific sensor variable (e.g., electrolyzer temperature). For example, the set of data points within this window is: Aggregation operations can include calculating the mean of a set of data points. ,variance Maximum value and minimum value One or more of the following. This aggregation operation is performed in parallel on all monitored sensor variables, thereby compressing a large number of high-frequency data points within a time window into a set of statistical characteristic values ​​that can characterize the process state during that period.

[0039] After the aggregation operation is completed, the system generates a low-resolution baseline data record and stores it in the database. Each stored baseline data record has a well-defined structure, which may include: a timestamp field, the value of which can be the start or end time of the aggregation time window; and a set of aggregation value fields, storing the calculation results of each sensor variable within that time window (e.g., ...). , The data record includes a resolution identifier field to explicitly mark it as low-resolution baseline data, and an event ID field. In the steady-state storage process of this invention, since the data record is not associated with any production process event, its event ID field is set to a preset null value to distinguish it from the high-resolution data stored in the event-driven process.

[0040] See attached document Figure 3 When the data monitoring module determines that the production process deviates from a steady state, the system switches to an event-driven, high-resolution storage core process. The first step in this process is to identify and instantiate production process events. The data monitoring module continuously analyzes the acquired high-frequency sensor data streams to monitor data anomalies that conform to preset patterns.

[0041] In one specific implementation, the event monitoring mechanism is based on Statistical Process Control (SPC) rules. The data monitoring module monitors each key sensor variable (e.g., the main roller current of the foil production machine). Maintain a dynamically updated statistical baseline that includes the mean over the most recent time window. and standard deviation For each newly collected data point The data monitoring module compares it with dynamic control limits calculated based on a statistical baseline. Preset event triggering conditions can be defined as a set of rules. For example, when a data point... The value satisfies or When a single point of failure occurs, it is determined that an anomaly has occurred; or, when all 8 consecutive data points fall within the mean. When the data sequence of any sensor variable satisfies any rule in the rule set, a chain (run) anomaly is determined to have occurred. Once the data sequence of any sensor variable satisfies any rule in the rule set, the data monitoring module determines that a new production process event has occurred.

[0042] Upon determining that a new production process event has occurred, the system immediately performs an event object instantiation operation, creating a structured data object to represent this event. This event object contains a set of predefined attributes used to fully describe the event. The event object instantiation process is as follows: First, the system precisely assigns the data point's acquisition timestamp that triggered the event to the event object's event start time attribute. Simultaneously, the sensor variable that triggered the preset rule is added to the event object's associated sensor variable collection attribute.

[0043] Secondly, the system assigns an initial severity level to the event object. This level can be set according to the specific SPC rule that triggered it. For example, an event triggered by a single point of failure rule can have its initial severity level set to warning; while an event triggered by a chain anomaly rule can have its initial severity level set to concern.

[0044] Finally, the event end time attribute of the event object is set to null or a special activity status identifier at this point. This indicates that the event has started but has not yet ended. The system will continue to monitor the sensor variables associated with this event until these variables return to a statistically stable state and remain so for a preset period of time, at which point this attribute will be updated to mark the formal end of the event. The instantiated event object is then passed to the event processing module for subsequent causal correlation analysis.

[0045] See attached document Figure 4 After the data monitoring module completes the identification and instantiation of production process events, the generated event objects are transmitted to the event processing module. The causal analysis module in the event processing module is then triggered, and its core task is to determine whether there is a causal relationship between the newly occurred production process event and any currently active production process event.

[0046] Upon receiving a new event object, the causal analysis module retrieves a set containing all currently active production process event objects. Then, the module pairs the new event object with each active event object in this set, calculating a quantified correlation score for each pair. This score characterizes the strength of the association between the two events in both the temporal and physical dimensions.

[0047] In one specific implementation, the relevance score is calculated by combining two dimensions: temporal similarity and variable similarity. Temporal similarity... The calculation uses an exponential decay function, whose mathematical expression is: ; in, This is the formula for calculating similarity, which uses an exponential function. It is a parameter or weight that reflects the degree to which the time interval affects the similarity. Indicates the start time of the new time period. Indicates the end time of the activity period. This represents the difference between the start time of the new time period and the end time of the activity period.

[0048] Variable similarity The calculation uses the Jaccard Index, whose mathematical expression is: ; in, It is a set of sensor variables associated with events in the new production process. It is a set of sensor variables associated with events in the production process and their active states. Represents a set and The size of the intersection, Represents a set and The size of the union of the sets.

[0049] This index quantifies the physical correlation between two events by calculating the ratio of the number of elements in the intersection to the number of elements in the union of the sets of sensor variables associated with them. If the two events are triggered or affected by the same sensor variable anomaly, the value approaches 1.

[0050] The causal analysis module will calculate and By preset weighting coefficients and The final relevance score is obtained by performing a weighted summation. .

[0051] In obtaining relevance scores Then, the event ID generation module executes the causal determination and ID generation logic. This module will calculate the... The value is correlated with a preset correlation threshold. Compare them.

[0052] like The system determines that a new production process event is causally related to the current activity event. At this point, the event ID generation module defines the activity event as a predecessor event of the new event and generates a chained event ID for the new event. In a specific implementation, if the predecessor event's ID is E881, the chained ID of the new event can be constructed as E801.S81, where S is a separator and 81 is a sequence number, indicating that this is the first successor event in the E001 event chain.

[0053] If the calculation is performed for all activity events The values ​​are all no greater than The system then determines the new production process event as an independent event. At this point, the event ID generation module will generate a completely new, independent root event ID for it. In a specific implementation, this root event ID can be generated using a global counter, such as E002.

[0054] After ID generation is complete, the chained event ID or root event ID is assigned to the ID attribute of the new event object. At this point, the construction of the event causal chain and the generation of the dynamic ID are finished. The event object carrying this unique ID will then be passed to the high-resolution storage module for further processing.

[0055] See attached document Figure 5 After the event ID generation module generates a unique chained event ID or root event ID for a new production process event, the event object carrying this ID is transferred to the high-resolution storage module. Once triggered, this module executes a dual-channel data acquisition process that includes retrospective extraction and synchronous tracking to ensure that all high-frequency sensor data related to the event are completely recorded.

[0056] First, the high-resolution storage module performs retrospective data extraction. In one specific implementation, the system maintains a fixed-size circular buffer in memory to temporarily store all raw high-frequency sensor data from the most recent period. The module reads the event start time from the received event object. The attribute, and based on a preset backtracking time length. (For example, 30 seconds), calculate the start time for data extraction. Subsequently, the module retrieves and extracts timestamps located in the specified interval from the memory circular buffer. All high-frequency sensor data points within the system. This step ensures that contextual data reflecting the incubation process of an event is fully captured before it is formally identified.

[0057] Simultaneously, the high-resolution storage module initiates synchronous data tracking. As long as the event object's state is active (i.e., its event end time attribute has not yet been assigned a value), this module directly receives real-time high-frequency data streams from the sensor data source. This synchronous tracking process continues until the data monitoring module determines that the event has ended (e.g., the relevant sensor variable values ​​have returned to their steady-state range) and updates the event object's event end time attribute. This step ensures that all dynamic changes during the event are recorded at high resolution.

[0058] Finally, this module performs data association and unified storage. Whether historical data points are obtained through retrospective extraction or real-time data points are obtained through synchronous tracking, they all undergo a formatting process before being written to the permanent storage area. The system creates a high-resolution data record for each data point. This record contains the data point's timestamp, sensor variable identifier, specific value, and a crucial event ID field. The high-resolution storage module then fills the event ID field of these data records with the chained event ID or root event ID (e.g., E001.Se1 or E002) generated in the previous step, carried by the current event object.

[0059] After the association operation is completed, all data records tagged with the same event ID are treated as a whole and written to a dedicated high-resolution database table. In this way, two data streams originating before and during the event are logically merged to form a complete high-resolution dataset that is temporally continuous, has a clear causal relationship, and can be directly retrieved and analyzed later.

[0060] See attached document Figure 6 Embodiments of the present invention also provide a traceable enhancement mechanism for storage priority. In one specific implementation, each stored high-resolution data record, in addition to containing an event ID field, also contains a variable storage priority field. The initial value of this field is set to a preset default level, such as the standard.

[0061] When the severity level attribute of any production process event in an event chain is updated to a preset abnormal level (this update can be manually confirmed by the operator or automatically determined by another diagnostic analysis module), the system triggers a priority escalation process. The system first obtains the chain event ID corresponding to the event that has been determined to be abnormal, such as E001.S01.S02.

[0062] The system then parses the chained event ID, identifying all its predecessor event IDs within the event chain. For IDE001.501.502, its predecessor event IDs are the root event E001 and the intermediate event E001.S01. Based on this, the system generates a database update instruction. This instruction aims to find all high-resolution data records where the value of the event ID field exactly matches E001. Or the record for EOO1.SO2.

[0063] For all retrieved data records, the database update instruction updates the value of its storage priority field from standard to high. Through this mechanism, all high-resolution data associated with the eventual serious anomaly and all its preceding triggering events are marked as high priority. This high-priority identifier can be used for subsequent data lifecycle management, such as preventing this critical data from being prematurely archived or deleted by automated data cleanup strategies.

[0064] Embodiments of the present invention also provide a fault tracing query implementation method based on chained IDs. When it is necessary to perform root cause analysis on a certain production anomaly, the system can receive a query request containing the event ID of the final production process event.

[0065] Upon receiving the query request, the system first extracts the root event ID portion, E001, from the event ID contained in the request (e.g., Eeer.ser.se2). Then, the system constructs a database retrieval query that uses string prefix matching to retrieve all data records in the high-resolution database table whose event ID field begins with the root event ID IDEeoo1. In SQL (Structured Query Language), this operation can be achieved using the LIKE 'E001%' clause.

[0066] The query will return all high-resolution data records associated with the root event E001, subsequent events E801.S01, and final events E001.S01.S02. The system processes these retrieved data records as a whole, sorts them in ascending order according to their timestamp attributes, and thus reconstructs the discrete data records into a temporally continuous and complete high-resolution sensor dataset.

[0067] Finally, this complete dataset, encompassing the entire evolutionary path from the inception of a fault to its eventual occurrence, is returned as a query result to the requester (e.g., an engineer's analytics terminal or a higher-level application). This query method leverages the inherent logical structure of event IDs to achieve accurate and complete extraction of data related to the causal chain of a specific fault.

[0068] The monitoring system provided by this invention can be deployed in a hardware environment consisting of edge computing nodes and a central server. The edge computing node can be an industrial personal computer (IPC) or server deployed on the copper foil production line, equipped with a multi-core processor for real-time data processing, random access memory (RAM) for temporarily storing high-frequency data, and a solid-state drive (SSD) for deploying the operating system and applications. The central server can be one or more high-performance servers deployed in a data center, equipped with a large-capacity disk array for permanent data storage. The edge computing nodes and the central server are connected via an industrial Ethernet network, forming a distributed system for data acquisition and storage.

[0069] In this embodiment, the various functional modules of the system can be mapped to specific software components. The functions of the data monitoring module, event processing module, and baseline storage module can be integrated and implemented as a data acquisition and processing service that runs continuously on an edge computing node.

[0070] The data acquisition and processing service establishes communication with the programmable logic controller (PLC) or supervisory control and data acquisition system (SCADA) on the copper foil production line via industrial fieldbus protocols or OPCUA (Open Platform Communication Unified Architecture) protocols to acquire raw high-frequency sensor data streams. After receiving the data, the service first writes it into a memory circular buffer allocated in the RAM of the edge computing node with a preset time length (e.g., the most recent 5 minutes).

[0071] The data monitoring module, a core component of the service, performs real-time analysis of data flowing into the circular buffer. When it fails to detect an event, it periodically (e.g., every minute) triggers the baseline storage module's logic. This module extracts data from the circular buffer corresponding to the time window, performs aggregation operations, and sends the generated low-resolution baseline data records to the central server for storage via the network.

[0072] When the data monitoring module detects a new production process event, it triggers the event handling module. This module performs a series of operations, including event instantiation, correlation assessment with active events, and generation of chained or root event IDs. Upon completion, it immediately triggers the high-resolution storage module. This module retrieves backtracking data from the memory circular buffer based on the event's start timestamp and ID, continuously receives real-time data, tags it with event IDs, and sends it to the central server for storage via the network.

[0073] The system's permanent storage portion can consist of a time-series database or a relational database deployed on a central server. Within this database, two core data tables can be created: a baseline data table and a high-resolution event data table. Fields in the baseline data table may include: timestamps, aggregated statistical values ​​of each sensor variable (such as mean and variance). Fields in the high-resolution event data table may include: precise timestamps, sensor variable identifiers, raw sensor values, event IDs, and storage priority. Through this combined hardware and software implementation, the present invention achieves efficient and intelligent monitoring and storage of data from the copper foil production process.

Claims

1. A copper foil production monitoring method based on data storage optimization, characterized in that, Includes the following steps: Acquire high-frequency sensor data during the copper foil production process; Real-time monitoring of the high-frequency sensor data to determine whether new production process events have occurred; If a new production process event is detected, an event-driven high-resolution storage process is executed, which includes: a) Perform a correlation assessment between the new production process event and the currently active production process event to determine whether there is a causal relationship between the new production process event and any of the active production process events; b) Based on the results of the correlation assessment, when a causal relationship exists, a chain event ID is generated for the new production process event; when no causal relationship exists, an independent root event ID is generated for the new production process event. c) Temporarily store high-frequency sensor data within the time period corresponding to the new production process event, and then store the high-frequency sensor data after associating it with the chain event ID or the root event ID; If no new production process event is detected, under steady-state conditions, the high-frequency sensor data is aggregated to generate and store low-resolution baseline data.

2. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, The steps for correlation assessment in step a) include: Based on the temporal proximity between the new production process event and the production process event of the active state, time similarity is calculated; Calculate variable similarity based on the set of sensor variables associated with the new production process event and the production process event of the activity state; The final correlation score is obtained by weighting the time similarity and the variable similarity, and the existence of the causal relationship is determined based on the correlation score.

3. The copper foil production monitoring method based on data storage optimization according to claim 2, characterized in that, The time similarity is calculated using an exponential decay function, which indicates that the longer the time interval between two events, the lower the time similarity. The similarity of the variables is calculated using the Jaccard index, which represents the ratio of the intersection to the union of the sensor variable sets associated with two events.

4. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, The step of temporarily storing and storing high-frequency sensor data in step c) specifically includes: High-frequency sensor data is retrospectively extracted from a preset retrospective time period prior to the start time of the new production process event; Synchronously track real-time high-frequency sensor data during the ongoing active state of the new production process event; The high-frequency sensor data extracted retrospectively and the high-frequency sensor data tracked synchronously are then associated with the chain event ID or the root event ID and stored uniformly.

5. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, Define a variable storage priority for each high-frequency sensor data associated with the chained event ID stored in step c); Furthermore, when the severity level of any production process event in the event chain corresponding to the chained event ID is determined to be a preset abnormal level, the storage priority of the data associated with all preceding events in the event chain is retrospectively increased.

6. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, In step a), each of the aforementioned production process events is defined as an event object containing the following attributes: The event start time, event end time, associated set of sensor variables, and event severity level.

7. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, The copper foil production monitoring method also includes a query step: In response to receiving a query request, the query request containing the event ID of the final production process event; Based on the chain structure of the event ID, all data records with the event ID prefix in the event ID field are retrieved, and the retrieved data records are combined into a complete high-frequency sensor dataset and returned. The high-frequency sensor dataset contains all data associated with the final production process event and all its precursor events.

8. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, The aggregation operation of the high-frequency sensor data includes at least one of the following: mean, variance, maximum value, or minimum value.

9. The copper foil production monitoring method based on data storage optimization according to claim 1, characterized in that, The step of determining whether a new production process event has occurred is achieved by monitoring at least one of the following data patterns: This involves univariate anomalies, changes in cross-correlation among multiple variables, or changes in the frequency domain characteristics of signals, all based on statistical process control rules.

10. A copper foil production monitoring system based on data storage optimization, characterized in that, include: The data monitoring module is used to acquire and monitor high-frequency sensor data in real time during the copper foil production process to determine whether a production process event has occurred. An event handling module is configured to be activated in response to the data monitoring module determining that a new production process event has occurred. The event handling unit includes: The causal analysis module is used to assess the correlation between the new production process event and the currently active production process event to determine whether there is a causal relationship between them. The event ID generation module is used to generate chained event IDs or independent root event IDs based on the determination result of the causal analysis module. A high-resolution storage module is used to temporarily store high-frequency sensor data within the time period corresponding to the new production process event, and to store the high-frequency sensor data in association with the event ID generated by the event ID generation module. The baseline storage module is configured to aggregate the high-frequency sensor data and generate and store low-resolution baseline data under steady-state conditions where the data monitoring module has not determined that a production process event has occurred.

Images