New energy automobile battery assembly line fault early warning method and system

Abstract

The application provides a new energy automobile battery assembly line fault early warning method and system, collects parameter characteristic values of each key process in the new energy automobile battery assembly line, and then pushes all parameter characteristic values after time stamp alignment to the corresponding digital twin virtual model of the assembly line to determine the exponentially weighted moving average and standard deviation of each parameter characteristic value, and then determine the upper and lower limits of each parameter characteristic value, real-time monitor the parameter characteristic values of each key process in the assembly line and the corresponding upper and lower limits of the early warning, determine the integrated deviation of the current assembly line running state relative to the historical health benchmark, and when the integrated deviation exceeds the preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding early warning upper and lower limits within a single batch, a fault early warning signal of the new energy automobile battery assembly line is generated. By adopting the scheme of the application, early fault early warning of the battery assembly line can be performed based on the weak abnormal trend of multiple parameters.

Description

Technical Field

[0001] This application relates to the field of fault early warning technology, and in particular to a fault early warning method and system for a new energy vehicle battery assembly production line. Background Technology

[0002] The rapid development of new energy vehicles has placed extremely high demands on the safety and reliability of their core component, the power battery. As a key link in the manufacturing process, the stability of the battery assembly production line directly determines the quality of the final product. This production line covers many key stations such as cell welding, module tightening, and sealing testing. The process parameters of each station must be strictly monitored to prevent batch quality defects caused by equipment performance degradation, material batch differences, or occasional failures.

[0003] Currently, fault warnings for such production lines mainly rely on two types of technologies. The first type is real-time alarms based on fixed thresholds. This method is simple and weak, but too rigid and cannot adapt to parameter baseline drift caused by the natural aging of equipment. It is also difficult to distinguish parameter changes caused by normal batch fluctuations from early fault budding. The second type is based on statistical process control methods, such as using control charts to monitor parameter mean and range. Although this method can reflect process fluctuations to some extent, its control limits are usually based on long historical data and are updated slowly. It is not sensitive enough to complex anomaly patterns with multiple parameters coupled, and the real-time and early warning capabilities are insufficient. Therefore, how to provide early fault warnings for battery assembly production lines based on weak anomaly trends of multiple parameters has become a challenge for the industry. Summary of the Invention

[0004] Based on this, this application provides a method and system for early fault warning of battery assembly line in new energy vehicle battery assembly line based on weak abnormal trends of multiple parameters.

[0005] In a first aspect, this application provides a method for early warning of faults in a new energy vehicle battery assembly line, comprising the following steps:

[0006] Collect parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line;

[0007] All parameter feature values ​​are timestamped and then pushed to the corresponding digital twin virtual model of the production line to realize the mapping and visualization of the production line operation status in the virtual space.

[0008] Based on the historical health operation data accumulated by the digital twin virtual model, a sliding time window mechanism is adopted to determine the exponentially weighted moving average and standard deviation of each parameter feature value, and then to determine the upper and lower warning limits of each parameter feature value that are adaptive with production batch and equipment operation time.

[0009] Real-time monitoring of the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits. When multiple parameter characteristic values ​​enter the warning zone between the corresponding warning upper and lower limits and show a continuous deviation trend, the degree of integrated deviation of the current production line operation status from the historical health benchmark is determined.

[0010] When the integration deviation exceeds the preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit in a single batch, a fault warning signal for the new energy vehicle battery assembly production line is generated.

[0011] In some embodiments, timestamp alignment of all parameter feature values ​​specifically includes:

[0012] Retrieve structured data packets with local timestamps corresponding to all parameter feature values;

[0013] The local timestamps in all structured data packets are aligned using a precision clock synchronization protocol to obtain all parameter feature values ​​after timestamp alignment.

[0014] In some embodiments, pushing all timestamp-aligned parameter feature values ​​to the digital twin virtual model corresponding to the production line to achieve the mapping and visualization of the production line's operating status in virtual space specifically includes:

[0015] All timestamp-aligned parameter feature values ​​are sent to the data access layer of the corresponding digital twin virtual model of the production line via industrial communication protocol;

[0016] Based on the workstation identifier and data content of all parameter feature values, update the state attributes of the corresponding virtual workstation entity in the digital twin virtual model;

[0017] The visualization engine that drives the digital twin virtual model renders and displays a three-dimensional virtual mapping of the production line's operating status based on the real-time changes in the stated state attributes.

[0018] In some embodiments, based on historical health operation data accumulated by the digital twin virtual model, the exponentially weighted moving mean and standard deviation of each parameter feature value are determined using a sliding time window mechanism, specifically including:

[0019] Historical health operation data within a specified time range is extracted from the time-series database associated with the digital twin virtual model;

[0020] For each parameter feature value, based on the set window length and sliding step size, the corresponding sliding time window data sequence is extracted from the historical health operation data;

[0021] Calculate the exponentially weighted moving mean and exponentially weighted moving standard deviation for each sliding time window data series.

[0022] In some embodiments, determining the upper and lower limits of the early warning system that adapt to production batch and equipment uptime for each parameter characteristic value specifically includes:

[0023] For each parameter feature value, the corresponding upper and lower warning limits are determined based on the preset confidence coefficient and the exponentially weighted moving mean and exponentially weighted moving standard deviation of the parameter feature value.

[0024] By associating the upper and lower warning limits with the current production batch identifier and the cumulative equipment running time, we obtain upper and lower warning limits that adapt the parameter feature values ​​to the production batch and equipment running time.

[0025] In some embodiments, the parameter characteristic values ​​of each key process in the production line and the corresponding upper and lower warning limits are monitored in real time:

[0026] Real-time acquisition of parameter characteristic values ​​of each key process generated during production line operation, aligned with timestamps;

[0027] Query and retrieve the latest upper and lower warning limits for each real-time parameter feature value associated with the current production batch and equipment uptime.

[0028] In some embodiments, when multiple parameter characteristic values ​​enter the warning zone between the corresponding upper and lower warning limits and show a continuous deviation trend, determining the integrated deviation of the current production line operating status relative to the historical health benchmark specifically includes:

[0029] For each parameter feature value, determine whether the real-time acquired parameter feature value has entered the warning zone between its corresponding upper and lower warning limits;

[0030] For the parameter characteristic values ​​that enter the warning zone, determine whether they show a continuous deviation trend;

[0031] If multiple parameter feature values ​​are determined to show a continuous deviation trend, the integrated deviation degree of the current production line operation state relative to the historical health benchmark is determined based on the current state of all parameter feature values ​​determined to show a continuous deviation trend and the statistical benchmark of the historical health operation data.

[0032] In some embodiments, when the integration deviation exceeds a preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit within a single batch, generating a fault warning signal for the new energy vehicle battery assembly production line specifically includes:

[0033] Continuously monitor whether the integration deviation exceeds the preset deviation threshold, and monitor whether any parameter feature value continuously exceeds its corresponding warning upper and lower limits within a single batch;

[0034] When any monitoring condition is met, a fault warning signal containing complete warning information is generated based on the specific triggering condition and relevant data.

[0035] The fault warning signal is sent to the designated warning receiver and the digital twin virtual model.

[0036] In some embodiments, the parameter characteristic values ​​include the effective value of welding current, the peak value of tightening torque, and the sealing pressure attenuation rate.

[0037] Secondly, this application provides a fault early warning system for a new energy vehicle battery assembly line, comprising:

[0038] The data acquisition module is used to collect parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line;

[0039] The processing module is used to align all parameter feature values ​​with timestamps, and then push all the time-stamped parameter feature values ​​to the digital twin virtual model corresponding to the production line, so as to realize the mapping and visualization of the production line operation status in the virtual space.

[0040] The processing module is also used to determine the exponentially weighted moving average and standard deviation of each parameter feature value based on the historical health operation data accumulated by the digital twin virtual model, using a sliding time window mechanism, and then determine the upper and lower limits of the early warning for each parameter feature value that are adaptive with production batch and equipment running time.

[0041] The processing module is also used to monitor the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits in real time. When multiple parameter characteristic values ​​enter the warning zone between the corresponding warning upper and lower limits and show a continuous deviation trend, the integrated deviation of the current production line operation status from the historical health benchmark is determined.

[0042] The execution module is used to generate a fault warning signal for the new energy vehicle battery assembly production line when the integration deviation exceeds a preset deviation threshold or when any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit in a single batch.

[0043] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0044] The fault early warning method and system for new energy vehicle battery assembly lines provided in this application first collects the parameter characteristic values ​​of each key process in the new energy vehicle battery assembly line, aligns all parameter characteristic values ​​with timestamps, and then pushes all timestamp-aligned parameter characteristic values ​​to the corresponding digital twin virtual model of the production line to realize the mapping and visualization of the production line's operating status in virtual space. This step enables standardized collection of multi-source heterogeneous process data, high-precision time synchronization, and real-time push to the digital twin model, thereby establishing an accurate and real-time digital mirror of the physical production line, providing a spatiotemporal unified basis for all subsequent analyses. First, a reliable data foundation for virtual-real correlation; second, based on the historical health operation data accumulated by the aforementioned digital twin virtual model, a sliding time window mechanism is used to determine the exponentially weighted moving average and standard deviation of each parameter's characteristic value, thereby determining the upper and lower limits of the early warning for each parameter's characteristic value that adapt to production batches and equipment operating time. This step can dynamically generate statistical early warning boundaries based on recent health data, thus establishing an adaptive and personalized normal range for each key parameter that changes with equipment aging and production batches. This overcomes the drawbacks of a fixed threshold "one-size-fits-all" approach and improves the early warning system's tolerance to normal process fluctuations. The system targets abnormal drift to reduce false alarms and missed alarms. Then, it monitors the parameter characteristics and corresponding warning limits of key processes in the production line in real time. When multiple parameter characteristics enter the warning zone between the corresponding warning limits and show a continuous deviation trend, it determines the integrated deviation of the current production line operating status from the historical health benchmark. This step can capture parameter trends that are still within acceptable range but continue to deteriorate, and perform comprehensive quantification when multiple parameters are abnormally linked. This enables keen perception and quantitative assessment of early, complex failure risks. This step significantly advances the warning timing and improves the ability to detect and quantify gradually changing, system-wide failures. The system's ability to detect systematic quality degradation enables truly predictive early warning. Finally, when the integration deviation exceeds a preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit within a single batch, a fault warning signal for the new energy vehicle battery assembly production line is generated. This step can automatically generate a complete structured warning based on dual intelligent criteria and simultaneously drive the virtual model for visual alarms, thereby achieving accurate, graded response and virtual-real linkage alarms for different fault modes. In summary, the solution of this application can provide early fault warnings for battery assembly production lines based on the weak abnormal trends of multiple parameters. Attached Figure Description

[0045] Figure 1 This is an exemplary flowchart of a fault early warning method for a new energy vehicle battery assembly line according to some embodiments of this application;

[0046] Figure 2 This is a schematic diagram illustrating an application scenario of an assembly line fault early warning data processing system according to some embodiments of this application;

[0047] Figure 3 This is a flowchart illustrating the process of determining integration deviation according to some embodiments of this application;

[0048] Figure 4 This is a structural schematic diagram of a fault early warning system for a new energy vehicle battery assembly line, as shown in some embodiments of this application.

[0049] Figure 5 This is a schematic diagram of the structure of a computer device for implementing a fault early warning method for a new energy vehicle battery assembly line, according to some embodiments of this application. Detailed Implementation

[0050] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0051] refer to Figure 1 The figure is an exemplary flowchart of a fault early warning method for a new energy vehicle battery assembly line according to some embodiments of this application. The fault early warning method for a new energy vehicle battery assembly line mainly includes the following steps:

[0052] In step 101, the parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line are collected.

[0053] In specific implementation, the parameter characteristic values ​​of each key process in the new energy vehicle battery assembly line can be collected in the following way: First, standard industrial data interfaces are configured on the field control units of the cell welding station, module tightening station, and sealing test station of the new energy vehicle battery assembly line; for the cell welding station, the root mean square value of the current in each welding point working cycle is calculated in real time by the signal processing unit integrated inside the welding power controller, and the calculation result is used as the effective value of the welding current and periodically output through PROFINET or EtherCAT industrial fieldbus; for the module tightening station, the high-precision torque sensor built into the servo tightening shaft controller records the maximum value on its torque curve after each tightening action, and the maximum value is used as the tightening torque peak value and output through the same industrial fieldbus protocol; for the sealing test station, the root mean square value of the current in each welding point working cycle is calculated in real time by the signal processing unit integrated inside the welding power controller, and the calculation result is used as the effective value of the welding current and periodically output through PROFINET or EtherCAT industrial fieldbus; for the module tightening ... At the sealing test station, a high-frequency sampling absolute pressure sensor continuously acquires raw pressure waveform data within the test chamber and sends this raw waveform data to an industrial edge computing gateway deployed at the station. A lightweight algorithm running within the edge computing gateway analyzes the pressure change curve over time in real time, calculating the pressure drop per unit time to obtain the sealing pressure decay rate. The effective value of the welding current, the peak value of the tightening torque, and the sealing pressure decay rate are the parameter characteristic values ​​of each key process in the new energy vehicle battery assembly line. Finally, the effective value of the welding current, the peak value of the tightening torque, and the sealing pressure decay rate are encapsulated into structured data packets with original station identifiers and local timestamps, awaiting subsequent transmission and processing, thus completing the acquisition of the characteristic values ​​of each key process parameter. Other methods can also be used in other embodiments, which are not limited here.

[0054] It should be noted that the parameter characteristic values ​​in this application refer to numerical indicators extracted from each key process of the new energy vehicle battery assembly production line, used to quantitatively characterize the core operating status and quality level of the process.

[0055] In some embodiments, reference Figure 2 As shown in the figure, this figure is a schematic diagram of the application scenario of the assembly line fault early warning data processing system shown in some embodiments of this application. The figure includes three main components: acquisition equipment, server and data storage equipment. The acquisition equipment is responsible for collecting the parameter feature values ​​of each key process in the new energy vehicle battery assembly line, and sending the collected parameter feature values ​​of each key process to the server through the communication network. The assembly line fault early warning data processing system runs on the server, and the server stores the processing results in the data storage device and visualizes them.

[0056] In step 102, all parameter feature values ​​are timestamped and then pushed to the corresponding digital twin virtual model of the production line to realize the mapping and visualization of the production line operation status in the virtual space.

[0057] In some embodiments, timestamp alignment of all parameter feature values ​​can be achieved using the following steps:

[0058] Retrieve structured data packets with local timestamps corresponding to all parameter feature values;

[0059] The local timestamps in all structured data packets are aligned using a precision clock synchronization protocol to obtain all parameter feature values ​​after timestamp alignment.

[0060] In specific implementation, obtaining the structured data packets with local timestamps corresponding to all parameter feature values ​​can be achieved in the following way: Deploy a data aggregation service, which continuously subscribes to and receives structured data packets periodically released by the field control units of the cell welding station, module tightening station, and sealing test station through an OPC UA client or a dedicated industrial protocol driver; wherein each structured data packet contains a station identifier of the data source, a local timestamp generated by the local clock of the data acquisition terminal, and specific parameter feature values; the data aggregation service temporarily caches and queues the structured data packets from different stations and different time points, waiting for unified time base processing. Other methods can also be used in other embodiments, which are not limited here.

[0061] In specific implementation, the local timestamps in all structured data packets are aligned based on a precise clock synchronization protocol. The resulting timestamp-aligned parameter characteristic values ​​can be achieved in the following way: the server hosting the data aggregation service and all devices providing the structured data packets in the production line network are connected to an IEEE-based network. The industrial Ethernet is synchronized with the 1588 precision clock protocol, thus providing a unified high-precision clock source. The data aggregation service converts the local timestamp attached to each of the cached structured data packets into an absolute timestamp based on the unified high-precision clock source. Then, the data aggregation service sets a fixed time alignment period, using this time alignment period as a time window, and aggregates all the structured data packets that have completed timestamp conversion and arrive at all workstations within this window. If there are multiple data packets for a certain workstation within the window, the data packet with the timestamp closest to the center point of the window is selected as the representative. For workstations with missing data within the window, valid data from the previous time window is used for interpolation to fill the gaps. Finally, a set containing the latest parameter feature values ​​of all key workstations at the same alignment time point is generated. This set is the set of all parameter feature values ​​after timestamp alignment. Other methods can also be used in other embodiments, which are not limited here.

[0062] In some embodiments, pushing all timestamp-aligned parameter feature values ​​to the digital twin virtual model corresponding to the production line to achieve the mapping and visualization of the production line's operating status in virtual space can be achieved through the following steps:

[0063] All timestamp-aligned parameter feature values ​​are sent to the data access layer of the corresponding digital twin virtual model of the production line via industrial communication protocol;

[0064] Based on the workstation identifier and data content of all parameter feature values, update the state attributes of the corresponding virtual workstation entity in the digital twin virtual model;

[0065] The visualization engine that drives the digital twin virtual model renders and displays a three-dimensional virtual mapping of the production line's operating status based on the real-time changes in the stated state attributes.

[0066] It should be noted that the digital twin virtual model in this application refers to a high-fidelity model constructed in virtual space that corresponds one-to-one with the elements of the physical production line and maintains a real-time data connection. Its core function is to serve as a synchronous digital mirror of the physical production line, realizing the virtual-real mapping and real-time visualization of the operating status.

[0067] In specific implementation, sending all timestamp-aligned parameter feature values ​​to the data access layer of the digital twin virtual model corresponding to the production line via an industrial communication protocol can be achieved in the following way: The data aggregation service encapsulates all timestamp-aligned parameter feature values ​​into a new data packet according to a predefined data format, as a synchronization data frame; in addition to the aligned timestamp and parameter feature values ​​of each workstation, the synchronization data frame also includes a frame sequence number to ensure transmission order; subsequently, the data aggregation service, as a publisher, publishes the synchronization data frame to a specific topic through OPC UA publish / subscribe mode or MQTT protocol; the data access layer of the digital twin virtual model corresponding to the production line, as a subscriber to the topic, listens to the network and receives the synchronization data frame in real time, completing the data transmission from the physical data layer to the virtual model layer. Other methods can also be used in other embodiments, which are not limited here.

[0068] In specific implementation, updating the state attributes of the corresponding virtual workstation entities in the digital twin virtual model based on the workstation identifier and data content of all parameter feature values ​​can be achieved in the following way: After receiving the synchronization data frame, the data access layer of the digital twin virtual model parses the synchronization data frame and extracts the alignment timestamp, as well as each workstation identifier and its corresponding parameter feature value; the simulation kernel of the digital twin virtual model locates the corresponding virtual workstation entity in the virtual scene based on the workstation identifier; the simulation kernel assigns the parsed parameter feature values ​​to the predefined state attribute variables in the virtual workstation entity that characterize its process operation state. For example, the effective value of the welding current is assigned to the "current current" attribute of the virtual welding gun entity, and the peak value of the tightening torque is assigned to the "current torque" attribute of the virtual tightening gun entity. This update process is event-driven, ensuring that the state attributes of all relevant virtual workstation entities can be refreshed in real time each time a new synchronization data frame is received. Other methods can also be used in other embodiments, which are not limited here.

[0069] In specific implementation, the visualization engine driving the digital twin virtual model can render and display a 3D virtual mapping of the production line's operating status based on the real-time changes in the state attributes. This can be achieved in the following way: the visualization engine of the digital twin virtual model continuously monitors the state attributes of each virtual workstation entity; when the state attributes are updated, the visualization engine drives the 3D rendering pipeline to update the virtual scene according to predefined visual mapping rules. Specific visualization updates include: dynamically refreshing and displaying the latest parameter characteristic values ​​of the data dashboard next to the 3D model of the corresponding virtual workstation entity; changing the color of the virtual workstation entity or its associated indicator light according to whether the parameter characteristic values ​​are within the normal range; and plotting the historical changes of key parameter characteristic values ​​on the monitoring panel of the virtual scene in the form of trend curves. Through the combination of the above multi-dimensional visual feedback, the overall operating status and detailed process data of the physical production line are mapped and displayed in real-time and intuitively in the 3D interface of the digital twin virtual model. Other methods can also be used in other embodiments, which are not limited here.

[0070] It should be noted that the above steps enable standardized acquisition of multi-source heterogeneous process data, high-precision time synchronization, and real-time push to the digital twin model, thereby establishing an accurate and real-time digital mirror of the physical production line, providing a reliable data foundation for all subsequent analyses that is spatiotemporally unified and correlates the virtual and real worlds.

[0071] In step 103, based on the historical health operation data accumulated by the digital twin virtual model, a sliding time window mechanism is used to determine the exponentially weighted moving average and standard deviation of each parameter feature value, and then determine the upper and lower warning limits of each parameter feature value that are adaptive with production batch and equipment operation time.

[0072] In some embodiments, the determination of the exponentially weighted moving mean and standard deviation of each parameter feature value based on historical health operation data accumulated by the digital twin virtual model, using a sliding time window mechanism, can be achieved through the following steps:

[0073] Historical health operation data within a specified time range is extracted from the time-series database associated with the digital twin virtual model;

[0074] For each parameter feature value, based on the set window length and sliding step size, the corresponding sliding time window data sequence is extracted from the historical health operation data;

[0075] Calculate the exponentially weighted moving mean and exponentially weighted moving standard deviation for each sliding time window data series.

[0076] It should be noted that the exponentially weighted moving average in this application is a dynamic statistic that reflects the estimate of the central trend of the recent operating level of the parameter characteristic value. The data closer to the current time contributes more to the mean, thus being able to sensitively track the slow drift of the parameter level. The exponentially weighted moving standard deviation quantifies the fluctuation range and dispersion of the parameter characteristic value around its recent central trend, and is also more sensitive to recent changes, used to measure the recent stability of the process.

[0077] In specific implementation, extracting historical health operation data within a specified time range from the time-series database associated with the digital twin virtual model can be achieved in the following way: First, access the industrial time-series database that is synchronized with the digital twin virtual model. This industrial time-series database continuously stores all parameter feature values ​​and their corresponding metadata that have been timestamped and passed in by the synchronization data frame. Then, according to a preset data filtering strategy, extract data records within a specified time range from the database. The specified time range can be set by the system administrator according to the equipment evaluation cycle or process stability requirements. The core of the data filtering strategy is to extract only the time series of parameter feature values ​​associated with the batch of "qualified product output" recorded in the manufacturing execution system, ensuring that the extracted data all represent the healthy operation of the production line under fault-free conditions, thereby forming a pure historical health operation dataset for statistical modeling. Other methods can also be used in other embodiments, which are not limited here.

[0078] In specific implementation, for each parameter feature value, the corresponding sliding time window data sequence can be extracted from the historical health operation data according to the set window length and sliding step size. This can be achieved in the following way: For each parameter feature value, firstly, an independent sliding time window mechanism is maintained for each parameter feature value to be monitored. The window length is set according to the stability cycle of the process or the number of production batches, for example, set to the data of the most recent 500 production batches. The sliding step size is defined as the amount of new data included each time the window slides forward, usually set to 1 production batch. Then, according to the current calculation time, for the historical health operation data sequence of each parameter feature value, starting from the latest data point, backtracking to select data points that meet the window length requirement, forming the data sequence of the current window. Finally, when a new batch of health data is generated, the window moves forward according to the sliding step size, removing the earliest batch data and adding the latest batch data, thereby forming the updated sliding time window data sequence corresponding to each parameter feature value. This mechanism is to ensure that the data used for calculation always reflects the recent production status. Other methods can also be used in other embodiments, which are not limited here.

[0079] In specific implementation, the exponentially weighted moving average and exponentially weighted moving standard deviation of each sliding time window data sequence can be calculated as follows: For any parameter feature value of the current sliding time window data sequence, a recursive exponentially weighted moving average statistical algorithm can be used. First, a smoothing coefficient is set, which determines the rate at which the weights of historical data decay. When calculating the exponentially weighted moving average, starting from the beginning of the sequence, the first data point is used as the initial mean. The inclusion of each new data point will update the current mean according to the smoothing coefficient, so that the newer the data, the greater its contribution weight to the mean. Based on the exponentially weighted moving average, the same exponentially weighted logic is used to calculate the squared deviation of each data point from the current estimated mean, and then the weighted average and square root are performed to obtain the exponentially weighted moving standard deviation. This exponentially weighted moving standard deviation also gives higher weight to recent fluctuations. Through the above recursive calculation, the exponentially weighted moving average and exponentially weighted moving standard deviation corresponding to the current sliding time window are finally output, which can sensitively reflect the recent data level and fluctuations. Other methods can also be used in other embodiments, which are not limited here.

[0080] In some embodiments, determining the upper and lower limits of the warning system that adapt to production batches and equipment uptime for each parameter characteristic value can be achieved by the following steps:

[0081] For each parameter feature value, the corresponding upper and lower warning limits are determined based on the preset confidence coefficient and the exponentially weighted moving mean and exponentially weighted moving standard deviation of the parameter feature value.

[0082] By associating the upper and lower warning limits with the current production batch identifier and the cumulative equipment running time, we obtain upper and lower warning limits that adapt the parameter feature values ​​to the production batch and equipment running time.

[0083] It should be noted that the warning upper and lower limits in this application refer to a pair of boundary values ​​dynamically calculated for each parameter feature value based on the exponentially weighted moving average and standard deviation. Their common function is to define a dynamic and statistically significant normal operating range for the corresponding parameter, serving as a direct benchmark for judging whether real-time data is abnormal. This range can be adaptively adjusted with the production process.

[0084] In specific implementation, determining the corresponding upper and lower warning limits based on the preset confidence coefficient and the exponentially weighted moving average and exponentially weighted moving standard deviation corresponding to the parameter characteristic value can be achieved in the following way: First, a preset confidence coefficient is set for each parameter characteristic value. This confidence coefficient is a constant determined based on process quality requirements and historical data statistical analysis, used to control the severity of the warning range. Other methods can also be used in other embodiments, and are not limited here. Then, the exponentially weighted moving average and exponentially weighted moving standard deviation calculated for the parameter characteristic value at the current calculation time are obtained. The specific value of the upper warning limit can be determined by adding the preset confidence coefficient and the exponentially weighted moving standard deviation corresponding to the parameter characteristic value to the exponentially weighted moving average. The upper warning limit is obtained by multiplying the standard deviations, which reflects the theoretical upper limit of the parameter's normal fluctuation under the current data volatility level. This calculation ensures that the upper warning limit not only follows the change in the mean level but also adaptively adjusts according to the recent volatility of the data. The specific value of the lower warning limit can be obtained by subtracting the product of the preset confidence coefficient and the exponentially weighted moving standard deviation from the exponentially weighted moving average. That is, the lower warning limit reflects the theoretical lower limit of the parameter's normal fluctuation under the current data volatility level. Through this symmetrical calculation logic, together with the upper warning limit, a dynamic warning interval is formed with the exponentially weighted moving average as the center and the exponentially weighted moving standard deviation as the width. Other methods can also be used in other embodiments, which are not limited here.

[0085] In specific implementation, the upper and lower warning limits are associated with the current production batch identifier and the cumulative equipment running time to obtain the upper and lower warning limits that adapt to the production batch and equipment running time of the parameter feature value. This can be achieved in the following way: After calculating the specific values ​​of the upper and lower warning limits of a certain parameter feature value at the current moment, the production batch identifier corresponding to the driving data that triggered this calculation and the cumulative running time of the workstation equipment to which the parameter feature value belongs at the calculation moment are immediately obtained. Subsequently, the five data items of the upper warning limit value, the lower warning limit value, the corresponding parameter feature value identifier, the production batch identifier, and the cumulative equipment running time are stored as a complete record in a configuration library or time-series database dedicated to managing dynamic thresholds, forming a dynamic threshold storage unit. Through this associated storage, the upper and lower warning limits used by any parameter feature value at any historical moment have clear equipment cumulative running time and production batch dimension markers, so that the warning boundary has traceability that dynamically evolves with the progress of production and equipment wear, thereby realizing the adaptive characteristics of the upper and lower warning limits. Other methods can also be used in other embodiments, which are not limited here.

[0086] It should be noted that the above steps can dynamically generate statistical early warning boundaries based on recent health data, thereby establishing an adaptive and personalized normal range for each key parameter that varies with equipment aging and production batches. This overcomes the drawbacks of a fixed threshold and improves the early warning system's tolerance for normal process fluctuations and its targeting of abnormal drifts, reducing false alarms and missed alarms.

[0087] In step 104, the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits are monitored in real time. When a parameter characteristic value enters the warning zone between the corresponding warning upper and lower limits and shows a continuous deviation trend, the integrated deviation of the current production line operation status from the historical health benchmark is determined.

[0088] In some embodiments, real-time monitoring of the parameter characteristic values ​​of each key process in the production line and the corresponding early warning upper and lower limits can be achieved by the following steps:

[0089] Real-time acquisition of parameter characteristic values ​​of each key process generated during production line operation, aligned with timestamps;

[0090] Query and retrieve the latest upper and lower warning limits for each real-time parameter feature value associated with the current production batch and equipment uptime.

[0091] In practice, the parameter characteristic values ​​of each key process generated by the production line operation and aligned with the timestamp can be obtained in real time in the following way: obtain the parameter characteristic values ​​of all key workstations at the latest alignment time point, including the effective value of welding current, the peak value of tightening torque and the attenuation rate of sealing pressure. These parameter characteristic values ​​have a unified time reference. Other methods can also be used in other embodiments, which are not limited here.

[0092] In specific implementation, querying and obtaining the latest warning upper and lower limits corresponding to each real-time parameter feature value associated with the current production batch and equipment running time can be achieved in the following way: while obtaining the parameter feature values ​​of each key process generated by the production line operation in real time and aligned with the timestamp, the current production batch identifier is obtained from the metadata of the synchronous data frame; at the same time, the current cumulative equipment running time of the corresponding workstation equipment is obtained through the query system; subsequently, using the parameter feature value identifier (such as "effective value of welding current"), the current production batch identifier, and the current cumulative equipment running time as joint query conditions, the configuration library storing dynamic thresholds is accessed in real time. According to the query conditions, the configuration library returns the latest set of warning upper limit values ​​and warning lower limit values ​​that were calculated and stored beforehand and associated with the parameter feature value and the closest production batch and equipment running time status, thereby completing the real-time matching and acquisition of the spatiotemporally adaptive dynamic warning boundary corresponding to each real-time parameter feature value. Other methods can also be used in other embodiments, which are not limited here.

[0093] In some embodiments, reference Figure 3 As shown in the figure, this is a flowchart illustrating the process of determining the integration deviation in some embodiments of this application. In this embodiment, when multiple parameter feature values ​​enter the warning zone between the corresponding upper and lower warning limits and show a continuous deviation trend, the integration deviation of the current production line operating status relative to the historical health benchmark can be determined by the following steps:

[0094] In step 1031, for each parameter feature value, it is determined whether the real-time acquired parameter feature value enters the warning zone between its corresponding upper and lower warning limits;

[0095] In step 1032, for the parameter characteristic values ​​that enter the warning zone, it is determined whether they show a continuous deviation trend;

[0096] In step 1033, if multiple parameter feature values ​​are determined to show a continuous deviation trend, the integrated deviation degree of the current production line operation state relative to the historical health benchmark is determined based on the current state of all parameter feature values ​​determined to show a continuous deviation trend and the statistical benchmark of the historical health operation data.

[0097] It should be noted that the integration deviation in this application is a comprehensive indicator that quantifies the overall deviation of the current production line's overall process status from the historical health benchmark. It can reveal the compound risks of coordinated anomalies among multiple parameters and is the core criterion for achieving early warning of compound faults.

[0098] In specific implementation, for each parameter feature value, determining whether the real-time acquired parameter feature value has entered the warning zone between its corresponding upper and lower warning limits can be achieved in the following way: After acquiring each real-time parameter feature value and its corresponding upper and lower warning limits, a fixed warning zone ratio is set for each parameter feature value; based on this warning zone ratio, such as 0.9, the specific warning zone boundary is calculated. For example, the value obtained by multiplying the upper warning limit by the warning zone ratio is used as the upper boundary of the warning zone, and the value obtained by dividing the lower warning limit by the warning zone ratio is used as the lower boundary of the warning zone; subsequently, the real-time parameter feature value is compared with the calculated upper and lower boundaries of the warning zone respectively; if the real-time parameter feature value is greater than or equal to the lower boundary of the warning zone and less than or equal to the lower warning limit, or greater than or equal to the upper warning limit and less than or equal to the upper boundary of the warning zone, then it is determined that the real-time parameter feature value has entered the warning zone between the upper and lower warning limits. Other methods can also be used in other embodiments, which are not limited here.

[0099] In specific implementation, determining whether a parameter characteristic value entering the warning zone exhibits a continuous deviation trend can be achieved in the following way: First, after determining that a parameter characteristic value has entered the warning zone, immediately extract the time-series data points within the most recent period, such as the most recent 30 sampling periods, from the real-time historical cache of that parameter characteristic value; then, use a linear regression analysis method, such as the least squares method, to fit the time-series data points, calculate a trend line, and obtain its slope. This slope quantifies the average direction and rate of change of the parameter value over time; if the absolute value of the calculated slope is greater than a preset trend sensitivity threshold, and its direction of change points away from the center of the upper and lower warning limits, for example, the slope is positive in the warning zone on the upper warning limit side, or negative in the warning zone on the lower warning limit side, and the trend is statistically significant, then it is determined that the parameter characteristic value exhibits a continuous deviation trend. Other methods can also be used in other embodiments, which are not limited here.

[0100] In specific implementation, if multiple parameter feature values ​​are determined to show a continuous deviation trend, the integrated deviation degree of the current production line operating state relative to the historical health benchmark, based on the current state of all parameter feature values ​​determined to show a continuous deviation trend and the statistical benchmark of the historical health operation data, can be determined in the following way: when two or more key parameter feature values ​​are detected to show a continuous deviation trend, the integrated deviation degree calculation is triggered. Specifically, this calculation involves first extracting the historical data sequences of these parameter feature values ​​showing a continuous deviation trend under healthy conditions from the historical health operation dataset, calculating their mean to form a historical mean vector, and calculating the covariance between them. A historical covariance matrix is ​​constructed. Then, the real-time observed values ​​of these parameter eigenvalues ​​that show a continuous deviation trend at the current moment are used to construct a current observation vector. Finally, the Mahalanobis distance of the current observation vector with respect to the historical mean vector and the historical covariance matrix is ​​calculated. The Mahalanobis distance is calculated by first calculating the difference vector between the current observation vector and the historical mean vector, and then transforming and weighting the difference vector using the inverse matrix of the historical covariance matrix to obtain a scalar distance value. This scalar distance value is the integrated deviation, which comprehensively reflects the degree to which the overall state of multiple related parameters deviates from its historical health benchmark. Other methods can also be used in other embodiments, which are not limited here.

[0101] It should be noted that the above steps can capture the trend of parameters that are still within the acceptable range but continue to deteriorate, and perform comprehensive quantification when multiple parameters are abnormal in synergy. This enables keen perception and quantitative assessment of early and complex failure risks. This step allows the early warning time to be significantly advanced, improves the ability to detect slow-changing and systemic quality degradation, and achieves true predictive early warning.

[0102] In step 105, when the integration deviation exceeds the preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper and lower limits in a single batch, a fault warning signal for the new energy vehicle battery assembly production line is generated.

[0103] In some embodiments, when the integration deviation exceeds a preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit within a single batch, generating a fault warning signal for the new energy vehicle battery assembly production line can be achieved through the following steps:

[0104] Continuously monitor whether the integration deviation exceeds the preset deviation threshold, and monitor whether any parameter feature value continuously exceeds its corresponding warning upper and lower limits within a single batch;

[0105] When any monitoring condition is met, a fault warning signal containing complete warning information is generated based on the specific triggering condition and relevant data.

[0106] The fault warning signal is sent to the designated warning receiver and the digital twin virtual model.

[0107] It should be noted that the deviation threshold in this application is used as a benchmark for triggering the "integration deviation exceeds the limit" type of early warning. It is used to distinguish between normal fluctuations in the overall process status and abnormal deviations that require intervention. It is a key parameter for controlling the sensitivity and specificity of the early warning system.

[0108] In specific implementation, continuously monitoring whether the integration deviation exceeds a preset deviation threshold and whether any parameter feature value continuously exceeds its corresponding warning upper or lower limit within a single batch can be achieved in the following way: In the real-time monitoring process, two independent monitoring threads run synchronously; the first monitoring thread continuously receives and checks the value of the integration deviation calculated each time, and compares it with a deviation threshold set in advance by analyzing historical fault case data. If the integration deviation exceeds the deviation threshold in several consecutive calculation cycles, it is determined that the condition of "integration deviation exceeding the preset deviation threshold" is met; the second monitoring thread maintains a continuous over-limit counter for each real-time parameter feature value within the life cycle of its single production batch. Whenever the real-time value of a parameter feature value is determined to exceed its corresponding warning upper or lower limit, the counter is incremented by one. If the continuous over-limit counter of any parameter feature value reaches a preset number of times, such as 3 times, before the end of the batch, it is immediately determined that the condition of "any parameter feature value continuously exceeding the corresponding warning upper or lower limit within a single batch" is met. Other methods can also be used in other embodiments, which are not limited here.

[0109] It should be noted that the deviation threshold can be preset in the following way: After system deployment or major process changes, an initial threshold calibration phase needs to be performed first. In this initial threshold calibration phase, historical operating data covering the known fault occurrence periods over a sufficiently long period of time is extracted from the historical database associated with the digital twin virtual model. Using the historical operating data, the integration deviation values ​​at a series of time points before the fault occurred are calculated retrospectively, thereby forming a fault precursor integration deviation sample set. Subsequently, statistical analysis methods are used, such as calculating a specific high quantile of the sample set or combining domain expert experience. An initial deviation threshold is determined from the set of integrated deviation samples of the fault precursors, such that the integrated deviation value of most fault precursor samples exceeds the initial deviation threshold, while the integrated deviation value calculated during most healthy operation periods is lower than the initial deviation threshold. This initial deviation threshold is recorded in the system configuration library as the initial value of the preset deviation threshold. During subsequent operation, the preset deviation threshold can be recalculated and updated periodically based on newly accumulated fault cases and healthy operation data to achieve continuous optimization of the threshold. Other methods can also be used in other embodiments, which are not limited here.

[0110] In specific implementation, when any monitoring condition is met, generating a fault warning signal containing complete warning information based on the specific triggering conditions and related data can be achieved in the following way: Once any of the above monitoring threads triggers the judgment condition, the warning generation service is immediately awakened. The warning generation service first determines the trigger type, such as whether it is "integration deviation exceeding the threshold" or "single parameter continuous exceeding the limit", and collects the key context data at the current moment. This key context data includes: trigger timestamp, trigger station identifier (for single parameter trigger) or the identifier of the involved multi-station (for integration deviation trigger), a list of specific abnormal parameter characteristic values ​​and their real-time values, the current integration deviation value, the current production batch identifier, and the relevant warning upper and lower limit values; then, the warning generation service fills and encapsulates the above information into a structured data object according to a predefined JSON or XML data architecture. This data object is defined as the fault warning signal containing complete warning information. The information dimension of the fault warning signal ensures that subsequent processing can accurately locate the source of fault risk and assess its severity. Other methods can also be used in other embodiments, which are not limited here.

[0111] In specific implementation, the fault warning signal can be published to the designated warning receiver and the digital twin virtual model in the following ways: After generating the fault warning signal, it can be published through two parallel channels. The first channel is to push it to the designated external warning receiver, that is, to send the structured fault warning signal to the manufacturing execution system, workshop Kanban system, or mobile terminal of maintenance personnel through the enterprise message bus, standard API interface, or SMS / email gateway. The second channel is to publish it to the internal digital twin virtual model, that is, to send the fault warning signal to the model through the internal event interface or data update interface provided by the digital twin virtual model. After receiving the signal, the digital twin virtual model will drive the corresponding virtual workstation entity in the three-dimensional scene to change its visual state, such as flashing or changing color, and display detailed warning information in the warning panel of the virtual interface, thereby realizing synchronous and intuitive warning of physical space fault risks in the virtual space. Other methods can also be used in other embodiments, which are not limited here.

[0112] It should be noted that the above steps can automatically generate a complete structured early warning based on dual intelligent criteria, and simultaneously drive the virtual model to perform visual alarms, thereby achieving accurate and hierarchical response and virtual-real linkage alarms for different fault modes.

[0113] In another aspect, in some embodiments, this application provides a fault early warning system for a new energy vehicle battery assembly line, referring to... Figure 4 The figure is a schematic diagram of a fault early warning system for a new energy vehicle battery assembly line according to some embodiments of this application. The fault early warning system for a new energy vehicle battery assembly line includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below:

[0114] The acquisition module 401 in this application is mainly used to acquire the parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line.

[0115] Processing module 402, in this application, is mainly used to align the timestamps of all parameter feature values, and then push all the timestamp-aligned parameter feature values ​​to the digital twin virtual model corresponding to the production line, so as to realize the mapping and visualization of the production line operation status in the virtual space.

[0116] The processing module 402 described in this application is also used to determine the exponentially weighted moving average and standard deviation of each parameter feature value based on the historical health operation data accumulated by the digital twin virtual model, using a sliding time window mechanism, and then determine the upper and lower limits of the early warning for each parameter feature value that are adaptive with production batch and equipment running time.

[0117] The processing module 402 described in this application is also used to monitor the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits in real time. When multiple parameter characteristic values ​​enter the warning zone between the corresponding warning upper and lower limits and show a continuous deviation trend, the integrated deviation of the current production line operation status from the historical health benchmark is determined.

[0118] The execution module 403 in this application is mainly used to generate a fault warning signal for the new energy vehicle battery assembly production line when the integration deviation exceeds a preset deviation threshold or when any parameter feature value continuously exceeds the corresponding warning upper and lower limits in a single batch.

[0119] The various modules in the aforementioned fault early warning system for new energy vehicle battery assembly lines can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0120] In another embodiment, this application provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 5 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores fault warning data for the new energy vehicle battery assembly line. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a fault warning method for the new energy vehicle battery assembly line.

[0121] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0122] In one embodiment, a computer device is also provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above embodiment of the fault early warning method for new energy vehicle battery assembly line.

[0123] In one embodiment, a computer-readable storage medium is provided, storing a computer program that, when executed by a processor, implements the steps in the above-described embodiment of the fault early warning method for new energy vehicle battery assembly lines.

[0124] In one embodiment, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps described in the embodiment of the fault early warning method for new energy vehicle battery assembly lines.

[0125] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0126] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0127] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A fault early warning method for a new energy vehicle battery assembly production line, characterized in that, Includes the following steps: Collect parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line; All parameter feature values ​​are timestamped and then pushed to the corresponding digital twin virtual model of the production line to realize the mapping and visualization of the production line operation status in the virtual space. Based on the historical health operation data accumulated by the digital twin virtual model, a sliding time window mechanism is adopted to determine the exponentially weighted moving average and standard deviation of each parameter feature value, and then to determine the upper and lower warning limits of each parameter feature value that are adaptive with production batch and equipment operation time. Real-time monitoring of the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits. When multiple parameter characteristic values ​​enter the warning zone between the corresponding warning upper and lower limits and show a continuous deviation trend, the degree of integrated deviation of the current production line operation status from the historical health benchmark is determined. When the integration deviation exceeds the preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit in a single batch, a fault warning signal for the new energy vehicle battery assembly production line is generated. The sliding time window mechanism has a sliding step size of one production batch. When a new batch of health data is generated, the sliding time window moves forward according to the sliding step size, removing the earliest batch data and adding the latest batch data. Specifically, when multiple parameter characteristic values ​​enter the warning zone between the corresponding upper and lower warning limits and show a continuous deviation trend, the integrated deviation of the current production line operating status from the historical health benchmark is determined, including: For each parameter feature value, determine whether the real-time acquired parameter feature value has entered the warning zone between its corresponding upper and lower warning limits; For the parameter characteristic values ​​that enter the warning zone, determine whether they show a continuous deviation trend; If multiple parameter feature values ​​are determined to show a continuous deviation trend, the integrated deviation degree of the current production line operation state relative to the historical health benchmark is determined based on the current state of all parameter feature values ​​determined to show a continuous deviation trend and the statistical benchmark of the historical health operation data. The determination of whether the parameter exhibits a continuous deviation trend includes: extracting time-series data points of the parameter's feature value over a recent period, fitting the time-series data points using linear regression analysis, calculating a trend line and obtaining its slope; if the absolute value of the slope is greater than a preset trend sensitivity threshold, and its direction of change points away from the center of the warning upper and lower limits, and the trend is statistically significant, then the parameter's feature value is determined to exhibit a continuous deviation trend. The determination of the integrated deviation of the current production line operating status relative to the historical health benchmark includes: when two or more parameter feature values ​​are detected to be showing a continuous deviation trend, extracting the historical data sequence of these parameter feature values ​​showing a continuous deviation trend under the health state from the historical health operation data, calculating their mean to form a historical mean vector, and calculating the covariance between them to form a historical covariance matrix; constructing the current observation vector from the real-time observation values ​​of these parameter feature values ​​showing a continuous deviation trend at the current moment; and calculating the Mahalanobis distance of the current observation vector relative to the historical mean vector and the historical covariance matrix as the integrated deviation.

2. The method as described in claim 1, characterized in that, The specific steps for timestamp alignment of all parameter feature values ​​include: Retrieve structured data packets with local timestamps corresponding to all parameter feature values; The local timestamps in all structured data packets are aligned using a precision clock synchronization protocol to obtain all parameter feature values ​​after timestamp alignment.

3. The method as described in claim 1, characterized in that, Pushing all timestamp-aligned parameter feature values ​​to the corresponding digital twin virtual model of the production line to achieve the mapping and visualization of the production line's operating status in virtual space specifically includes: All timestamp-aligned parameter feature values ​​are sent to the data access layer of the corresponding digital twin virtual model of the production line via industrial communication protocol; Based on the workstation identifier and data content of all parameter feature values, update the state attributes of the corresponding virtual workstation entity in the digital twin virtual model; The visualization engine that drives the digital twin virtual model renders and displays a three-dimensional virtual mapping of the production line's operating status based on the real-time changes in the stated state attributes.

4. The method as described in claim 1, characterized in that, Based on the historical health operation data accumulated by the aforementioned digital twin virtual model, a sliding time window mechanism is used to determine the exponentially weighted moving mean and standard deviation of each parameter's characteristic value, specifically including: Historical health operation data within a specified time range is extracted from the time-series database associated with the digital twin virtual model; For each parameter feature value, based on the set window length and sliding step size, the corresponding sliding time window data sequence is extracted from the historical health operation data; Calculate the exponentially weighted moving mean and exponentially weighted moving standard deviation for each sliding time window data series.

5. The method as described in claim 1, characterized in that, The specific upper and lower limits for determining the adaptive warning values ​​of each parameter characteristic value based on production batch and equipment uptime include: For each parameter feature value, the corresponding upper and lower warning limits are determined based on the preset confidence coefficient and the exponentially weighted moving mean and exponentially weighted moving standard deviation of the parameter feature value. By associating the upper and lower warning limits with the current production batch identifier and the cumulative equipment running time, we obtain upper and lower warning limits that adapt the parameter feature values ​​to the production batch and equipment running time.

6. The method as described in claim 1, characterized in that, Real-time monitoring of the parameter characteristic values ​​of each key process in the production line and the corresponding upper and lower warning limits specifically includes: Real-time acquisition of parameter characteristic values ​​of each key process generated during production line operation, aligned with timestamps; Query and retrieve the latest upper and lower warning limits for each real-time parameter feature value associated with the current production batch and equipment uptime.

7. The method as described in claim 1, characterized in that, When the integration deviation exceeds a preset deviation threshold or any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit within a single batch, a fault warning signal for the new energy vehicle battery assembly production line is generated, specifically including: Continuously monitor whether the integration deviation exceeds the preset deviation threshold, and monitor whether any parameter feature value continuously exceeds its corresponding warning upper and lower limits within a single batch; When any monitoring condition is met, a fault warning signal containing complete warning information is generated based on the specific triggering condition and relevant data. The fault warning signal is sent to the designated warning receiver and the digital twin virtual model.

8. The method as described in claim 1, characterized in that, The parameter characteristic values ​​include the effective value of welding current, the peak value of tightening torque, and the sealing pressure attenuation rate.

9. A fault early warning system for a new energy vehicle battery assembly line, which uses the fault early warning method for a new energy vehicle battery assembly line according to any one of claims 1 to 8 for fault early warning, characterized in that, The system includes: The data acquisition module is used to collect parameter characteristic values ​​of each key process in the new energy vehicle battery assembly production line; The processing module is used to align all parameter feature values ​​with timestamps, and then push all the time-stamped parameter feature values ​​to the digital twin virtual model corresponding to the production line, so as to realize the mapping and visualization of the production line operation status in the virtual space. The processing module is also used to determine the exponentially weighted moving average and standard deviation of each parameter feature value based on the historical health operation data accumulated by the digital twin virtual model, using a sliding time window mechanism, and then determine the upper and lower limits of the early warning for each parameter feature value that are adaptive with production batch and equipment running time. The processing module is also used to monitor the parameter characteristic values ​​of each key process in the production line and the corresponding warning upper and lower limits in real time. When multiple parameter characteristic values ​​enter the warning zone between the corresponding warning upper and lower limits and show a continuous deviation trend, the integrated deviation of the current production line operation status relative to the historical health benchmark is determined. The execution module is used to generate a fault warning signal for the new energy vehicle battery assembly production line when the integration deviation exceeds a preset deviation threshold or when any parameter characteristic value continuously exceeds the corresponding warning upper or lower limit in a single batch.

Images