An airbag tray pressure data analysis management system and method

Abstract

The application relates to the technical field of data processing, and discloses an air bag tray pressure data analysis management system and method, which comprises an access module, a decoupling module, a tracking module, a compensation module and a splicing module. The method comprises the following steps: obtaining a sensing mark and original pressure of a tray with an air bag by analyzing a data packet and writing the original pressure into a time sequence database; updating a mapping table by analyzing a node binding instruction to obtain a time window; outputting a real-time state corresponding to the tray according to a scheduling message; when the state is stable and the time window is greater than zero, calculating an offset by using original pressure before and after unbinding to generate compensation pressure, and generating an abnormal mark when a transient gradient deviates; and extracting numerical values according to time boundaries to splice complete data flow and push the data flow. The application compensates for step errors, guarantees the continuity of pressure data of the air bag tray during circulation, and improves the accuracy of time sequence state analysis and underlying business management.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an analysis and management system and method for airbag tray pressure data. Background Technology

[0002] In the industrial production process of power battery pack manufacturing, the power battery pack is usually supported by a dedicated tray with built-in inflatable airbags to maintain the consistency of stress state under different processes. The pressure data inside the airbag can reflect the tray's load-bearing state and the airbag's sealing state. High-frequency acquisition, state classification, and numerical calculation of this data constitute the basic business requirement for the analysis and management of airbag tray pressure data.

[0003] Considering that airbag trays are usually powered by no external cables during in-plant logistics, existing industrial scenarios mostly use wireless sensors with built-in micro batteries to monitor air pressure and convert discrete measurement values ​​into digital messages to be transmitted to the background for persistent storage. Then, the real-time received readings are compared with preset static thresholds to assist the control terminal in anomaly investigation or material flow processing.

[0004] The existing processing flow has shortcomings in multi-node concurrent writing and time-series continuity maintenance. High-frequency synchronous reporting can easily cause high-frequency time-series data writing blockage. At the same time, due to hardware battery life limitations, sensors need to be replaced, and manufacturing tolerances and calibration errors between old and new nodes can cause abrupt changes in the collected values. Due to the lack of dynamic benchmark compensation and cross-cycle stitching mechanisms for node binding relationships, it is difficult to maintain the continuity of stress data formed in different hardware cycles on the time axis, which can easily lead to anomaly identification errors during monitoring and judgment.

[0005] Therefore, this invention proposes an analysis and management system and method for airbag tray pressure data to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an analysis and management system and method for airbag tray pressure data, which solves the problems of pressure data writing blockage caused by multi-node concurrency and misjudgment of airbag status caused by reference step when replacing sensors on an airbag tray.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an analysis and management system for airbag tray pressure data, comprising: The access module is used to parse data packets to obtain sensor identifiers, the original pressure of the airbag tray, and timestamps, and write the sensor identifiers, the original pressure, and the timestamps into a high-frequency time series database. The decoupling module is used to parse the node binding update instruction to obtain the tray code and the new sensor identifier. In the preset mapping table, the end time of the old sensor identifier associated with the tray code is updated to the unbinding time, the start time of the new sensor identifier is set as the binding time, and the time window is obtained by subtracting the unbinding time from the binding time. The tracking module is used to output the real-time status corresponding to the tray code based on the external scheduling message. The real-time status is either transient, steady-state, or an unidentified default abnormal state. The compensation module is used to calculate the historical gradient and predicted value using the original pressure before the unbinding time when the real-time state is the steady state and the time window is greater than zero, subtract the mean of the original pressure after the binding time from the predicted value to generate an offset, add the offset to the subsequent original pressure to generate a compensation pressure, and generate an anomaly mark when the instantaneous gradient of the compensation pressure deviates from a preset threshold. The splicing module is used to look up the time boundary in the preset mapping table according to the tray code, extract the original pressure and the compensation pressure according to the time boundary to splice them into a complete data stream, and push the anomaly mark to the external scheduling terminal.

[0008] Preferably, in the access module, the step of writing the sensor identifier, the original pressure, and the timestamp into the high-frequency time series database includes: A double-buffered queue is constructed by allocating foreground and background memory blocks in the memory pool. The sensor identifier, the original pressure, and the timestamp are pushed into the foreground memory block; When the storage pointer of the foreground memory block reaches the set capacity, the read / write pointer is switched and the background thread is woken up. The sensor identifier, the original pressure and the timestamp in the background memory block are then written in batches into the high-frequency time series database.

[0009] Preferably, in the decoupling module, the steps of updating the end time of the old sensor identifier associated with the tray code to the unbinding time in a preset mapping table, setting the start time of the new sensor identifier to the binding time, and subtracting the unbinding time from the binding time to obtain the time window include: Determine whether the old sensor identifier is empty. If the old sensor identifier is not empty, read the timestamp value of the moment when the node binding update instruction is received as the unbinding time, perform an update operation in the database where the preset mapping table is located, locate the record row containing the old sensor identifier, and set the end time of the record row as the unbinding time. A new data row containing the new sensor identifier is generated in the preset mapping table, and the start time of the new data row is set to the binding time; If the old sensor identifier is empty, the time window is set to zero; if the old sensor identifier is not empty, the binding time and the unbinding time are extracted and the difference is calculated to obtain the time window; if the time window is less than zero, the time window is assigned a value of zero.

[0010] Preferably, in the tracking module, the step of outputting the real-time status corresponding to the tray code based on the external scheduling message includes: Deserialize the external scheduling message to obtain context features; The context features that allow pressure fluctuations are classified as the transient state, the context features that require pressure to maintain a constant trend are classified as the steady state, and the context features that do not match the preset process classification rules are classified as the unidentified default abnormal state. The enumeration value bound to the tray code is refreshed based on the classification result of the context features, and the enumeration value is output as the real-time status.

[0011] Preferably, in the compensation module, the step of calculating the historical gradient and predicted value using the original pressure before the unbinding time includes: Extract the last original pressure corresponding to the unbinding time and the old original pressure corresponding to the preset historical collection period. The difference is obtained by subtracting the last original pressure from the old original pressure, and the difference is divided by the time span corresponding to the preset historical acquisition cycle to generate the historical gradient. The predicted value is obtained by multiplying the historical gradient with the time window, and the predicted value is generated by adding the last original pressure with the predicted value.

[0012] Preferably, in the compensation module, the step of generating an offset by subtracting the binding time from the predicted value and taking the mean of the original pressure includes: The original pressure is continuously collected from the low-power wireless sensor corresponding to the new sensor identifier for a preset number of periods after the binding time. The average value is generated by summing the original pressure over the preset number of cycles and dividing by the number of cycles. The offset is generated by subtracting the mean from the predicted value.

[0013] Preferably, in the compensation module, the step of generating an anomaly marker when the instantaneous gradient of the compensation pressure deviates from a preset threshold includes: The instantaneous gradient is obtained by calculating the rate of change of the compensation pressure in adjacent acquisition cycles. Determine whether the instantaneous gradient of multiple consecutive acquisition cycles is less than the preset threshold. If the instantaneous gradient of multiple consecutive acquisition cycles is less than the preset threshold, then generate the abnormal marker.

[0014] Preferably, in the splicing module, the step of looking up the time boundary in the preset mapping table based on the tray code includes: Retrieve historical binding records and current binding records associated with the tray code from the preset mapping table; Extract the earliest start time of all records in the search results as the lower time limit; Determine whether the end time of the currently bound record is null. If it is null, obtain the current time value of the central computing server as the time limit. If it is not null, extract the latest end time of all records in the search results as the time limit. The time boundary is determined based on the lower time limit and the upper time limit.

[0015] Preferably, in the splicing module, the step of extracting the original pressure and the compensated pressure according to the time boundary and splicing them into a complete data stream includes: Using the interval join operation, the original pressure and the compensated pressure are extracted from the high-frequency time series database in ascending order of timestamps, based on the old sensor identifier associated with the tray code, the new sensor identifier, and the time boundary. The extracted original pressure and the compensated pressure are written into an array buffer in the order they are stored, and then time-series merging is performed to form the complete data stream.

[0016] This invention also provides a method for analyzing and managing airbag tray pressure data, comprising the following steps: Parse the data packet to obtain the sensor identifier, the original pressure of the airbag tray, and the timestamp, and write the sensor identifier, the original pressure, and the timestamp into the high-frequency time series database; Parse the node binding update command to obtain the tray code and the new sensor identifier. In the preset mapping table, update the end time of the old sensor identifier associated with the tray code to the unbinding time, set the start time of the new sensor identifier to the binding time, and subtract the unbinding time from the binding time to obtain the time window. The real-time status corresponding to the tray code is output based on the external scheduling message. The real-time status is either transient, steady-state, or an unidentified default abnormal state. When the real-time state is the steady state and the time window is greater than zero, the historical gradient and predicted value are calculated using the original pressure before the unbinding time. The predicted value is subtracted from the mean of the original pressure after the binding time to generate an offset. The offset is added to the subsequent original pressure to generate a compensation pressure. When the instantaneous gradient of the compensation pressure deviates from a preset threshold, an anomaly marker is generated. The time boundary is retrieved from the preset mapping table based on the tray code. The original pressure and the compensation pressure are extracted according to the time boundary and combined into a complete data stream. The anomaly marker is then pushed to the external scheduling terminal.

[0017] This invention provides an analysis, management, and method for airbag tray pressure data. It offers the following advantages: 1. This invention constructs a dual-buffered queue by allocating front-end and back-end memory blocks in the memory pool, and combines read-write pointer switching logic to convert high-concurrency discrete measurement values ​​into batch dump mode, reducing database input and output blocking caused by multi-node synchronous reporting, thereby maintaining continuous throughput in the high-frequency pressure data access stage and realizing the caching and persistent management of airbag tray pressure data.

[0018] 2. This invention divides time windows by parsing node binding instructions and defines steady-state ranges by combining process characteristics. It uses historical gradients before unbinding time to calculate predicted values ​​and generates an offset based on the predicted values ​​and the original average pressure after binding time. Then, it adds the subsequent collected values ​​to the offset to form the compensation pressure, reducing measurement step errors caused by hardware replacement. While maintaining the temporal continuity of pressure data before and after sensor label changes, it improves the accuracy of calculation and analysis for physical leaks.

[0019] 3. This invention retrieves the corresponding time boundary based on the tray code in the mapping table, and extracts the original pressure and compensation pressure from the high-frequency time series database in ascending order of timestamp using interval connection operations. It merges the associated record time series scattered in different hardware cycles into a complete data stream, enabling the application layer to carry out backtracking analysis based on continuous data links, and providing a data foundation for airbag status judgment under complex working conditions. Attached Figure Description

[0020] Figure 1 This is a diagram illustrating the architecture of the airbag tray pressure data analysis and management system of the present invention. Figure 2 This is a flowchart of the airbag tray pressure data analysis and management method of the present invention; Figure 3 This is a flowchart of the data double-buffering access process of the present invention; Figure 4 This is the sensor binding update and time window calculation logic of the present invention; Figure 5This is a flowchart of the tray status classification and determination process of the present invention; Figure 6 This is a flowchart of the pressure numerical compensation and early warning process of the present invention; Figure 7 This is a flowchart of the timing data boundary splicing process of the present invention; Figure 8 The data index statistical distribution comparison bar chart of the present invention is shown below; (a) is a comparison bar chart of high frequency time series database write delay; (b) is a comparison bar chart of anomaly mark output false alarm rate.

[0021] Among them: 100, access module; 200, decoupling module; 300, tracking module; 400, compensation module; 500, splicing module. Detailed Implementation

[0022] 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.

[0023] Reference Figure 1 This invention provides an analysis and management system for airbag tray pressure data, including: an access module 100, a decoupling module 200, a tracking module 300, a compensation module 400, and a splicing module 500.

[0024] The airbag tray pressure data analysis and management system is deployed on a central computing server, which is used to build a computing processing environment for pressure data access, cache scheduling, database retrieval, status classification, numerical compensation, and data stitching. The central computing server is configured with a high-frequency time-series database and a mapping table for maintaining business attributes, and establishes message interaction channels with external dispatching terminals and underlying data acquisition nodes through network communication ports. In a preferred embodiment, the high-frequency time-series database is InfluxDB, TimescaleDB, or other time-series databases capable of batch writing, time window querying, and time-stamped ascending order retrieval of pressure data records. In one embodiment, the central computing server runs a server, a client, and an interface service module. The server receives and stores online device data, the client maintains the binding relationship between tray codes and sensor identifiers, and the interface service module provides tray pressure query and anomaly marking interaction to the external dispatching terminal.

[0025] The access module 100 is used to parse data packets, obtain sensor identifiers, the original pressure of the airbag tray and timestamp, and write the sensor identifiers, original pressure and timestamps into the high-frequency time series database.

[0026] The decoupling module 200 is communicatively connected to the access module 100 and is used to parse the node binding update command and obtain the tray code and the new sensor identifier. In the preset mapping table, the end time of the old sensor identifier associated with the tray code is updated to the unbinding time, the start time of the new sensor identifier is set to the binding time, and the time window is obtained by subtracting the binding time from the unbinding time.

[0027] The tracking module 300 is used to output the real-time status corresponding to the tray code based on the external scheduling message; the real-time status is either transient, steady state, or an unidentified default abnormal state.

[0028] The compensation module 400 is connected to the decoupling module 200 and the tracking module 300 respectively. It is used to calculate the historical gradient using the original pressure before the unbinding time when the real-time state is steady and the time window is greater than zero; calculate the predicted value based on the historical gradient and the time window; subtract the mean of the original pressure after the binding time from the predicted value to generate the offset; add the offset to the subsequently acquired original pressure to generate the compensation pressure; and generate an anomaly mark when the instantaneous gradient of the compensation pressure deviates from the preset threshold.

[0029] The splicing module 500 is connected to the compensation module 400. It is used to look up the corresponding time boundary in the mapping table according to the pallet code, and extract the original pressure and compensation pressure from the high-frequency time series database according to the time boundary. The splicing module 500 splices the original pressure and compensation pressure into a complete data stream and pushes the abnormal marker to the external scheduling terminal.

[0030] Reference Figure 2 This invention provides a method for analyzing and managing airbag tray pressure data. The analysis and management method is executed by the aforementioned analysis and management system and includes the following steps: S1, parse the data packet, obtain the sensor identifier, the original pressure of the tray with airbag and the timestamp, and write the sensor identifier, the original pressure and the timestamp into the high-frequency time series database; S2, parse the node binding update instruction, obtain the tray code and the new sensor identifier; in the preset mapping table, update the end time of the old sensor identifier associated with the tray code to the unbinding time, set the start time of the new sensor identifier to the binding time, and subtract the binding time from the unbinding time to obtain the time window. S3, based on the external scheduling message, outputs the real-time status corresponding to the tray code; the real-time status is either transient, steady state, or an unidentified default abnormal state; S4. When the real-time state is steady and the time window is greater than zero, calculate the historical gradient using the original pressure before the unbinding time; calculate the predicted value based on the historical gradient and the time window; subtract the mean of the original pressure after the binding time from the predicted value to generate the offset; add the offset to the subsequently acquired original pressure to generate the compensation pressure; generate an anomaly marker when the instantaneous gradient of the compensation pressure deviates from the preset threshold. S5 retrieves the corresponding time boundary from the mapping table based on the tray code, and extracts the original pressure and compensation pressure from the high-frequency time series database according to the time boundary; it then assembles the original pressure and compensation pressure into a complete data stream and pushes the anomaly marker to the external scheduling terminal.

[0031] To facilitate the explanation of the execution process of each technical step, the following describes the corresponding processing flow in conjunction with the access module 100, decoupling module 200, tracking module 300, compensation module 400, and splicing module 500.

[0032] Reference Figure 3 The access module 100 is used to execute step S1, which includes the following steps: S101 receives the binary data stream transmitted by the underlying data acquisition node and obtains data packets from the binary data stream.

[0033] In this embodiment, a low-power wireless sensor is installed inside the airbag tray. The low-power wireless sensor measures the air pressure inside the airbag according to a set sampling frequency, for example, 1 to 10 times per second.

[0034] To accommodate the limited installation space of the airbag tray, the low-power wireless sensor adopts a compact design with miniaturized dimensions. It integrates a wireless transmission antenna and a low-frequency receiving antenna, and uses a miniature battery, such as a 130mAh CR1632 battery, to power the ultra-low power operation of the hardware and software.

[0035] After measuring physical parameters, the low-power wireless sensor encapsulates the data into binary format and transmits it to the signal receiver via a repeater using multi-frequency communication technology. The signal receiver receives the data relayed by the repeater and transmits the corresponding binary data stream to the central computing server. The repeater employs multi-frequency automatic frequency hopping communication technology in the data link to avoid co-channel interference between the radio frequency signals emitted by the low-power wireless sensor and the repeater's own relayed signals.

[0036] For the principle of data transmission between the sensor and the repeater using radio frequency signals and the frequency hopping communication of the repeater, those skilled in the art can refer to existing wireless radio frequency transmission specifications for implementation. The signal transmission at the physical layer is a well-known technology in this field and will not be elaborated here.

[0037] Access module 100 receives a binary data stream from the server and performs frame header and frame tail matching on the binary data stream to obtain the data packets to be processed. For example, the binary data stream can be a continuously arriving byte stream. Access module 100 identifies preset fixed bytes, such as 0xAA 0x55, as frame header identifiers to extract a single complete data packet. The data packet obtained in this way is the application layer basic data set formed by the low-power wireless sensor after being transmitted sequentially through the repeater, signal receiver, and server.

[0038] S102, parse the data packet to obtain the sensor identifier, the original pressure of the airbag tray, and the timestamp.

[0039] In the underlying processing of data packets, the access module 100 uses Manchester encoding rules to decode the data packets, extracts the level transition edges in the received binary data stream, maps the falling edge transition at the center of the bit period to the first logic bit, and maps the rising edge transition at the center of the bit period to the second logic bit, thereby completing the restoration of the timing level to the basic bit stream.

[0040] After decoding, the access module 100 performs integrity verification on the decoded data packet using a cyclic redundancy check (CRC) algorithm. Specifically, the access module 100 extracts the additional checksum fragment at the end of the data packet, uses a pre-configured generator polynomial (e.g., the standard CRC-16 polynomial 0x8005) to perform modulo-2 division on the payload containing physical parameters within the data packet, and determines whether the final remainder matches the additional checksum fragment to determine whether bit flipping has occurred during wireless radio frequency transmission due to interference from the industrial environment. After successful verification, the access module 100 extracts the corresponding byte fragment according to a pre-defined frame structure protocol.

[0041] Specifically, the access module 100 extracts the device serial number field of a preset fixed length from the header of the data packet to obtain the sensor identifier. For example, the preset fixed length is 4 bytes. The sensor identifier is used to uniquely identify a specific low-power wireless sensor hardware, such as a 32-bit integer ID.

[0042] The access module 100 extracts the physical parameter fields from the data packet payload to obtain the original pressure of the airbag tray. During this process, the underlying hardware converts the analog pressure signal into a digital value using a high-precision analog-to-digital converter (ADC) circuit and encapsulates the digital value into the physical parameter fields. The access module 100 reads the digital value and performs numerical calculations according to a preset quantization resolution (e.g., the full-scale average division ratio corresponding to a 12-bit ADC) to obtain the original pressure of the airbag tray. The original pressure of the airbag tray is a pressure measurement value, for example, 30 kPa, acquired by the low-power wireless sensor within the current measurement cycle, characterizing the inflation state of the airbag in the airbag tray.

[0043] The access module 100 extracts values ​​from the time information segments of the data packet, or the receiving program generates the current time parameter at the moment of reception to obtain a timestamp. The timestamp is used to record the specific moment when the original pressure of the airbag tray is measured or reaches the central computing server.

[0044] S103 allocates a foreground memory block and a background memory block in the memory pool to build a double-buffered queue, and pushes the sensor identifier, the original pressure of the airbag tray, and the timestamp into the foreground memory block.

[0045] Industrial production lines typically involve a large number of inflatable pallets operating simultaneously. Directly executing high-frequency time-series database disk writes for multi-node, high-frequency data reporting can easily lead to input / output congestion.

[0046] As a preferred approach, to handle high-concurrency access, the access module 100 requests a contiguous memory space from the memory pool reserved by the central computing server. The access module 100 divides the contiguous memory space into a foreground memory block and a background memory block, and uses the two memory blocks to construct a double-buffered queue.

[0047] The double-buffered queue works by setting up two areas in memory. One area is used to receive new data, and the other area is used to dump data to the backend. The two areas take turns to handle the receiving and dumping tasks, so as to reduce the high-frequency time-series database input and output waiting caused by writing a single piece of data.

[0048] In this embodiment, the access module 100 combines the sensor identifier obtained in the previous step, the original pressure of the airbag tray, and the timestamp into a standard format data record. The access module 100 pushes the data record into the foreground memory block in the receiving state for temporary caching, following the auto-incrementing memory address order. When pushing the data record, the access module 100 increments the corresponding cumulative value in a global variable. The cumulative value records the total number of data currently received and temporarily stored in the foreground memory block.

[0049] S104 executes read / write pointer switching when the storage pointer of the foreground memory block reaches the set capacity, and wakes up the background thread to batch write the sensor identifier, the original pressure of the airbag tray and the timestamp in the background memory block into the high-frequency time series database.

[0050] The access module 100 monitors the cumulative status of the amount of data written to the foreground memory block, that is, it extracts the cumulative value maintained in the previous step and determines whether the storage pointer of the foreground memory block has reached the set capacity. The determination rule is as follows: ; in, This represents the cumulative number of data records that have been received and written to the foreground memory block. This indicates the maximum amount of data that can be pre-allocated to a single memory block, i.e., the set capacity. Set Capacity The value can be determined based on the total available memory size of the central computing server and the byte length of a single data record, for example, the range is 1000 to 5000 records.

[0051] When detected Reaching or exceeding the set capacity At that time, the access module 100 performs a read / write pointer switch. If it detects... Less than the set capacity If the access module 100 does not trigger a switching operation, it will maintain the current write state and wait for the data record after the next data packet is parsed to be pushed into the front-end memory block and continue to synchronously update the cumulative value after being pushed in.

[0052] When performing a read / write pointer switch, the access module 100 changes the target address pointer of the current data input stream from the original foreground memory block to the original background memory block, thus turning the original background memory block into a new foreground memory block for receiving the data records obtained from subsequent parsing; at the same time, it marks the original foreground memory block, which is already full, as a new background memory block, so that it stops receiving input data.

[0053] After the read / write pointer switch is completed, the access module 100 wakes up the background thread that is in a dormant waiting state. The awakened background thread reads all data records stored in the new background memory block and uses the batch insert interface of the high-frequency time series database to write the sensor identifiers, the original pressure of the airbag tray, and the timestamp from the background memory block into the high-frequency time series database in batches. After the persistent writing is completed, the background memory block is cleared and used for the next read / write pointer switch.

[0054] Therefore, the access module 100 transforms the high-frequency discrete single-entry storage of sensor identifiers, the original pressure of the airbag tray, and timestamps into a batch writing to a high-frequency time-series database for storage.

[0055] Reference Figure 4 The decoupling module 200 is used to execute step S2, which includes the following steps: S201, parse the node binding update command, obtain the tray code and the new sensor identifier, and query the preset mapping table to obtain the old sensor identifier associated with the tray code.

[0056] In this embodiment, during the operation of the industrial production line, the tray with airbag may trigger a replacement request due to the replacement of the low-power wireless sensor or a binding error.

[0057] The decoupling module 200 listens to the front-end interface network packets, receives control commands containing modified parameters, and performs deserialization operations according to a preset data communication protocol. In one embodiment, the front-end interface network packets are generated by the client when performing batch binding, binding modification, or binding error correction operations between the tray code and the sensor identifier. The client encapsulates the tray code corresponding to the binding operation and the new sensor identifier into a node binding update command and sends it to the decoupling module 200. During deserialization, the decoupling module 200 uses a character stream parse tree to map the received byte sequence into an entity object structure in memory, and extracts the corresponding string fields from the specified key-value pairs of the entity object structure to complete the parsing of the node binding update command and obtain the tray code and the new sensor identifier.

[0058] The node binding update command triggers a change in the data association between the pallet code and the sensor identifier. The pallet code uniquely identifies an air-cushioned pallet carrying materials on the production line. The new sensor identifier is the identification code for the low-power wireless sensor currently preparing to establish a connection with the pallet code.

[0059] To retrieve historical associated records, the decoupling module 200 generates a database query statement using the tray code as the primary key and initiates a query request to a preset mapping table to retrieve records currently in effect, thereby obtaining the old sensor identifier associated with the tray code. The old sensor identifier is the identification code of the low-power wireless sensor that the tray code was bound to before this change operation.

[0060] The pre-defined mapping table is constructed by the relational database engine based on business entity relationships during the system initialization phase, and is a two-dimensional relational data table. For example, the pre-defined mapping table includes a tray code column for storing tray codes, a sensor identifier column for storing sensor identifiers, a start time column for defining the effective start time, and an end time column for defining the failure end time.

[0061] S202, perform the update operation, and set the end time of the row containing the old sensor identifier to the unbinding time corresponding to the node binding update command trigger time.

[0062] Directly overwriting old records can affect the traceability relationship between historical data and underlying hardware. As a preferred approach, this embodiment adopts a time-axis-based lossless decoupling mechanism.

[0063] The decoupling module 200 reads the timestamp value of the central computing server at the moment it receives the node binding update command, and uses the timestamp value as the unbinding time corresponding to the trigger time of the node binding update command. The unbinding time is used to characterize the critical time point at which the low-power wireless sensor corresponding to the old sensor identifier stops providing effective pressure monitoring data for the airbag-equipped tray corresponding to the tray code.

[0064] Subsequently, the decoupling module 200 constructs an update statement with conditional constraints, locating the record row in the preset mapping table that contains the old sensor identifier and is currently in effect. The decoupling module 200 performs an update operation in the database containing the preset mapping table, setting the end time of the row containing the old sensor identifier to the unbinding time corresponding to the node binding update instruction trigger time. Through this operation, the data lifecycle of the old sensor identifier is closed and defined at the database level, rather than being physically erased.

[0065] If, during actual business execution, the query reveals that the old sensor identifier corresponding to the tray code is empty, indicating an initial binding scenario, then the decoupling module 200 skips the update operation in this step, proceeds to the subsequent new binding process, and sets the time window to zero during subsequent time window calculations.

[0066] S203, perform the insertion operation to generate a new data row containing the new sensor identifier, and set the start time of the new data row to the binding time.

[0067] After closing the historical association records, the decoupling module 200 establishes a new correspondence to maintain the continuity of business flow. The decoupling module 200 obtains the time value of the currently executing program of the central computing server and uses the time value as the binding time. The binding time is used to characterize the effective start time point when the low-power wireless sensor corresponding to the new sensor identifier begins to provide pressure monitoring data for the airbag-equipped tray corresponding to the tray code.

[0068] The decoupling module 200 constructs a data insertion statement and executes the insertion operation, generating a new data row containing the new sensor identifier in a preset mapping table. When generating the new data row, the decoupling module 200 pairs the tray code with the new sensor identifier and writes it into the corresponding column of the preset mapping table, and sets the start time of the new data row to the binding time. The end time field of the new data row is left blank to indicate that the new data row is currently in effect.

[0069] S204, subtract the binding time from the unbinding time to obtain the time window.

[0070] The decoupling module 200 determines whether the old sensor identifier associated with the tray code is empty. If the old sensor identifier is empty, the time window is set to zero; if the old sensor identifier is not empty, the binding time and unbinding time are extracted, and the difference between the binding time and unbinding time is calculated to obtain the time window. The specific time span calculation rules are as follows: ; in, Indicates a time window; Indicates the binding time; Indicates the time of unbinding.

[0071] The time window reflects the interruption span of stress data recording on the time axis caused by low-power wireless sensor replacement operations, network transmission delays, or manual barcode scanning. If the recorded binding time is less than the unbinding time under high network concurrency or clock bias conditions, the calculated time window may be negative.

[0072] After completing the subtraction operation, the decoupling module 200 performs a boundary check: if If the value is less than zero, the time window is assigned a value of zero to indicate that no valid data acquisition interruption has occurred. After calculation, the decoupling module 200 temporarily stores the time window as an environmental constraint variable in the cache memory of the central computing server. By pre-calculating and caching the time window, subsequent processing modules do not need to repeatedly perform mapping table retrieval and time difference calculation when performing data stream merging and anomaly detection.

[0073] Reference Figure 5 The tracking module 300 is used to perform step S3, which includes the following steps: S301 receives network data transmitted through an external communication network, deserializes external scheduling messages, and obtains context features.

[0074] In this embodiment, the external scheduling terminal in the industrial production line is used to control the movement trajectory and process flow nodes of the airbag-equipped pallet, and communicates with the central computing server through the existing interface service program.

[0075] The tracking module 300 listens to the designated communication port and receives external scheduling messages in byte stream format from the external scheduling terminal. These external scheduling messages are used to synchronize the current workstation or movement status of the airbag-equipped tray.

[0076] Upon receiving an external scheduling message, the tracing module 300 uses a JSON parsing component to perform deserialization, mapping the continuous network byte stream into a hierarchical business object structure in memory. Subsequently, the tracing module 300 extracts specific field variables representing the current process stage from the business object structure, such as a string variable named Current_Station, thereby obtaining contextual features.

[0077] Contextual features are used to indicate the specific process in which the airbag-equipped pallet corresponding to the pallet code is located in the current production stage, such as the inflation station, conveyor belt transport area, or static buffer area.

[0078] S302, based on the preset process classification rules, performs logical judgment, classifies the context features that allow pressure fluctuations into transient state, classifies the context features that require pressure to maintain a constant trend into steady state, and classifies the context features that do not match the preset process classification rules into unidentified default abnormal state.

[0079] The changes in airbag pressure vary across different industrial production processes, necessitating data-driven state classification of contextual features based on a process state dictionary. The tracking module 300 matches and compares the contextual features with the preset process state dictionary. This preset process state dictionary, loaded into the system memory during system initialization, is a key-value mapping data structure. It uses the process string as the lookup key and the corresponding state category identifier as the mapping value.

[0080] When the context feature indicating the tray code corresponding to the airbag is in a process that actively deforms and compresses the airbag, such as inflation, deflation, or robotic arm gripping, the tracking module 300 determines that a large range of airbag pressure changes during the process is a normal physical phenomenon and classifies the context feature that allows pressure fluctuations as transient. Transient is used to characterize the working mode where the airbag pressure is within a reasonable dynamic change range.

[0081] When the context feature indicating the pallet code corresponds to an airbag-equipped pallet in a stable conveyor belt transport or static buffer area, the tracking module 300 determines that the external force on the airbag during the process should remain constant. If the airbag pressure drops significantly, there is a possibility of airbag rupture and leakage. Therefore, the context feature requiring a constant pressure trend is classified as steady state. Steady state is used to characterize a static operating mode where the airbag pressure should not drift significantly and needs to be monitored.

[0082] If the context features cannot be matched in the preset process status dictionary, it indicates that the external scheduling terminal has issued unknown process information. The tracking module 300 classifies the unrecognized context features into the unrecognized default abnormal state as a backup state output for unknown processes.

[0083] The tracing module 300 internally calls the state mapping function to perform the above classification process. The logical expression of the state mapping function is shown in the following equation: ; in, The classification result represents the contextual features; This indicates the extracted contextual features; This represents a pre-configured set of features that allow pressure fluctuations, such as string identifiers for inflatable stations, assembly areas, etc. This represents a pre-configured set of features that require a constant pressure trend, such as a feature set containing string identifiers like conveyor belts and finished product areas. Indicates a transient state; Indicates steady state; Indicates an unrecognized default exception state. (Settings) As a fallback decision, it ensures that when an unknown procedure is sent from an external scheduling terminal, the program logic can still output a definite result without crashing.

[0084] S303, based on the classification results of context features, refresh the enumeration value bound to the tray code, and output the enumeration value as the real-time status.

[0085] After completing the state classification, the tracking module 300 converts the classification results into standardized computer variables for use by downstream computing modules. As a preferred method, the tracking module 300 locates the global state object in the memory runtime space, which is uniquely indexed by the tray code, and assigns values ​​to the state attributes defined within the global state object based on the classification results of context features.

[0086] Specifically, the tracking module 300 assigns the first enumeration value to the result mapping classified as transient, for example, the integer value 0 in the program; assigns the second enumeration value to the result mapping classified as steady-state, for example, the integer value 1 in the program; and assigns the third enumeration value to the result mapping classified as unrecognized default abnormal state, for example, the integer value -1 in the program. Thus, the tracking module 300 completes the update of the enumeration values ​​bound to the tray code.

[0087] After the assignment and refresh are completed, the tracking module 300 outputs the enumerated value as the real-time status corresponding to the tray code to the subsequent compensation module 400. The compensation module 400 then determines whether to trigger data compensation calculation based on the real-time status. The real-time status represents the business condition classification result corresponding to the tray code in the current time slice.

[0088] Reference Figure 6 The compensation module 400 is used to perform step S4, which includes the following steps: S401: When the real-time state is steady and the time window is greater than zero, the compensation operation logic is triggered; when the real-time state is transient, an unidentified default abnormal state, or the time window is less than or equal to zero, the compensation operation is terminated and the original pressure is directly used.

[0089] In this embodiment, the compensation module 400 reads the real-time status output by the tracking module 300 and the time window output by the decoupling module 200 to determine whether data compensation should be triggered. Specifically, the compensation module 400 checks the enumerated value corresponding to the real-time status and reads the value of the time window.

[0090] When the enumeration value corresponding to the real-time state indicates that the real-time state is in a steady state, and the time window is greater than zero due to the replacement of the low-power wireless sensor or network latency, the compensation module 400 triggers the compensation operation logic consisting of subsequent steps S402 to S405.

[0091] When the enumeration value corresponding to the real-time state indicates that the real-time state is transient, an unidentified default abnormal state, or the time window is less than or equal to zero, the compensation module 400 terminates the compensation operation logic. Instead of performing numerical transformation operations on the data collected by the low-power wireless sensor corresponding to the new sensor identifier, it directly uses the original pressure of the airbag tray obtained by parsing for subsequent business processing and saves the original pressure of the airbag tray as the compensation pressure to the high-frequency time series database.

[0092] S402 calculates the historical gradient using the last original pressure corresponding to the unbinding time and the old original pressure corresponding to the preset historical collection cycle.

[0093] After entering the compensation calculation logic, as a preferred method, the compensation module 400 reads the cache memory of the central computing server and extracts the last original pressure corresponding to the unbinding time. The last original pressure is the valid pressure measurement value successfully reported by the low-power wireless sensor corresponding to the old sensor identifier before it became unbinded and failed.

[0094] The compensation module 400 also extracts the old original pressure corresponding to a preset historical collection period based on a pre-set historical tracing range. The preset historical collection period can be set according to the stable pressure fluctuation characteristics of the airbag under steady state, for example, taking the time node corresponding to 10 seconds before the untying time.

[0095] After extraction, the compensation module 400 subtracts the last original pressure from the old original pressure to obtain the difference, and divides the difference by the time span corresponding to the preset historical acquisition cycle to generate the historical gradient. The historical gradient is used to represent the natural decay rate or change rate of the airbag pressure in the period before the replacement action occurs. The specific calculation rules are shown in the following formula: ; in, Indicates historical gradient; Indicates the initial pressure at the last position; Indicates the old, original pressure; This indicates the time span corresponding to the preset historical collection period, specifically the difference between the timestamp corresponding to the last original pressure and the timestamp corresponding to the old original pressure, preferably 10 seconds.

[0096] S403 is a predicted value calculated from historical gradients and time windows.

[0097] The compensation module 400 extracts historical gradients and time windows, and multiplies the historical gradients with the time windows to obtain the predicted change value. The predicted change value represents the theoretical pressure change obtained by linear extrapolation based on historical trends within the time window when data acquisition is interrupted.

[0098] Subsequently, the compensation module 400 adds the original pressure corresponding to the unbinding time to the predicted change value to generate a predicted value. The predicted value represents the theoretical pressure inside the airbag when the low-power wireless sensor corresponding to the new sensor identifier establishes its binding and becomes effective. The specific calculation rules are shown in the following formula: ; in, Indicates the predicted value; Indicates the initial pressure at the last position; Indicates historical gradient; Indicates a time window.

[0099] S404 generates the offset by subtracting the mean of the original pressure after the binding time from the predicted value.

[0100] Because the newly replaced low-power wireless sensor is affected by hardware manufacturing tolerances and differences in internal calibration benchmarks, its initial measurement readings may have a step-like measurement deviation from the theoretical expectation. Based on this, after the binding time, the compensation module 400 continuously collects the original pressure reported by the low-power wireless sensor corresponding to the new sensor identifier for a preset number of cycles.

[0101] The preset number of cycles is set based on the sampling frequency and smoothing / denoising requirements of the underlying hardware; the number of cycles in the preset number of cycles ranges from 3 to 10, preferably 5. The compensation module 400 sums the original pressure of the above preset number of cycles and divides it by the number of cycles in the preset number of cycles to generate an average value. The average value is used to represent the smoothed reference reading after filtering out high-frequency noise measured during the initial power-up phase of the low-power wireless sensor corresponding to the new sensing identifier. The calculation rule for the average value is as follows: ; in, This represents the mean; Indicates the number of cycles for the preset number of cycles; This indicates the first report from the low-power wireless sensor corresponding to the new sensor identifier after the binding time. The original pressure of the air-filled tray in each cycle.

[0102] After obtaining the mean, the compensation module 400 subtracts the mean from the predicted value to generate the offset. The offset represents the systematic difference parameter between the low-power wireless sensor corresponding to the new sensor identifier and the low-power wireless sensor corresponding to the old sensor identifier on the physical measurement reference plane, and is used for subsequent reference calibration. The formula for calculating the offset is as follows: ; in, Indicates the offset; Indicates the predicted value; This represents the mean.

[0103] S405, add the offset to the subsequently acquired original pressure to generate compensation pressure; if the instantaneous gradient of the compensation pressure deviates from the preset threshold, an anomaly marker is generated.

[0104] After the offset calculation is completed, the compensation module 400 enters steady-state monitoring processing. For each airbag tray with original pressure continuously reported by the low-power wireless sensor corresponding to the new sensor identifier, the compensation module 400 adds the offset to the original pressure to generate the compensation pressure, and associates the compensation pressure with the corresponding tray code, new sensor identifier and timestamp to save it to the high-frequency time series database.

[0105] Compensation pressure is used to correct for the differences in low-power wireless sensor hardware benchmarks corresponding to the old and new sensor identifiers at the data level, forming continuous pressure values ​​on the time axis.

[0106] To detect anomalies, the compensation module 400 calculates the rate of change of the compensation pressure over adjacent acquisition cycles, obtaining the instantaneous gradient. The instantaneous gradient reflects the rate of decrease in airbag pressure at the current moment. The specific calculation rules are shown in the following formula: ; in, Indicates the instantaneous gradient; This indicates the compensation pressure during the current data collection cycle; This indicates the compensation pressure from the previous data collection cycle; This indicates the time interval between adjacent acquisition cycles.

[0107] After calculating the instantaneous gradient, the compensation module 400 determines whether the instantaneous gradient deviates from a preset threshold. The preset threshold is determined based on the natural micro-leakage rate of the airbag under normal physical conditions in an industrial setting, and serves as a negative reference limit; the value range of the preset threshold is -5.0 kPa / s to -0.5 kPa / s, preferably -2.0 kPa / s.

[0108] To reduce misjudgments caused by occasional single-point data jitter, the compensation module 400 determines whether the instantaneous gradient of M consecutive acquisition cycles is less than a preset threshold, where M ranges from 2 to 5, preferably 3.

[0109] If the instantaneous gradient is less than a preset threshold for M consecutive sampling cycles, it is determined that the current pressure drop rate exceeds the drop rate range corresponding to the preset threshold, indicating a possible physical rupture of the airbag leading to abnormal leakage. The compensation module 400 generates an anomaly marker. The anomaly marker is used to trigger an audible and visual alarm on the industrial production line, or to instruct the programmable logic controller to automatically reject the workstation containing the tray with the airbag.

[0110] If the instantaneous gradient is greater than or equal to the preset threshold, the current air pressure fluctuation is determined to be within a reasonable range of natural consumption. The compensation module 400 does not generate an abnormality marker and continues to receive the original pressure of the tray with airbag reported in the next collection cycle to continue executing the judgment process based on the instantaneous gradient.

[0111] Reference Figure 7 The splicing module 500 is used to perform step S5, which includes the following steps: S501: Based on the tray code, find the corresponding time boundary in the mapping table, and define the time axis retrieval range according to the time boundary.

[0112] In this embodiment, when the process flow progresses to the data verification node, the splicing module 500 integrates data spanning hardware change cycles. The splicing module 500 constructs a query statement using the tray code as the primary key and retrieves historical binding records and current binding records associated with the tray code from a preset mapping table.

[0113] After obtaining the above records, the stitching module 500 extracts the earliest start time of all records in the search results as the lower time limit. For the upper time limit, the stitching module 500 determines whether the end time of the currently bound record is null; if it is null, it means that the low-power wireless sensor corresponding to the sensor identifier is in the current active monitoring state, and the stitching module 500 obtains the current time value of the central computing server as the upper time limit; if it is not null, it means that the service corresponding to the tray code has ended, and the stitching module 500 extracts the latest end time of all records as the upper time limit.

[0114] The splicing module 500 determines the time boundary based on the lower and upper time limits. The time boundary represents the start and end span of the data activation of the pallet code within a specific business cycle and serves as the retrieval condition for range queries, timestamp sorting, and data merging in the high-frequency time-series database. Subsequently, the splicing module 500 uses the time boundary as a filtering condition to define the time axis retrieval range in memory. The time axis retrieval range is used to filter data unrelated to the current pallet service with airbags at the high-frequency time-series database level. The calculation rule for the time boundary is shown in the following formula: ; ; in, Indicates time boundaries; Indicates the lower limit of the time; A mathematical function that represents the minimum value; This indicates the first one retrieved from the mapping table. The start time of the associated record; Indicates the upper limit of time.

[0115] S502 uses interval connection operations to extract the original pressure and compensated pressure from the high-frequency time series database in ascending order of timestamps, based on the old sensor identifier, new sensor identifier and time boundary associated with the tray code.

[0116] After defining the timeline retrieval range, the stitching module 500 begins extracting pressure data. As a preferred method, the stitching module 500 utilizes the interval join operation at the underlying level of the high-frequency time series database to perform a relational query between the timeline retrieval range and the data table storing the data.

[0117] When executing a query, the splicing module 500 extracts the original pressure of the airbag tray without mathematical transformation for the data storage range corresponding to the old sensor identifier before the unbinding time; and extracts the compensation pressure generated by the compensation module 400 for the data storage range corresponding to the new sensor identifier after the binding time.

[0118] For data records whose timestamps fall outside the timeline retrieval range, the splicing module 500 does not load them at the query level. To maintain the data time sequence, the splicing module 500 adds a timestamp ascending sorting condition to the query command, and extracts the sorted original pressure and compensated pressure sequentially from the high-frequency time series database according to the time boundaries.

[0119] S503 merges the original pressure and the compensation pressure in a time sequence to form a complete data stream, and pushes the anomaly marker to the external scheduling terminal.

[0120] After extracting the pressure data, the stitching module 500 allocates a contiguous array buffer in its running memory and writes the raw and compensated pressures, arranged in ascending order of timestamps, into the array buffer in the order they were stored. Subsequently, the stitching module 500 performs time-series merging of the raw and compensated pressures to form a complete data stream. The complete data stream is a collection of airbag pressure monitoring data with temporal continuity, formed after baseline step error processing.

[0121] While performing data splicing, the splicing module 500 also monitors the calculation output of the compensation module 400 and determines whether to generate an anomaly flag.

[0122] If the compensation module 400 outputs an anomaly marker, the splicing module 500 extracts the anomaly marker and the pallet code associated with it, and serializes and encapsulates the anomaly marker and pallet code into a standard format network message payload. Subsequently, the splicing module 500 pushes the network message payload to the external dispatching terminal via the interface service module through the underlying network communication protocol. The interface service module is also used by the external dispatching terminal to query the pallet pressure value corresponding to the pallet code. After receiving the network message payload, the external dispatching terminal controls the corresponding sorting equipment to remove the pallet with airbag that has abnormal air pressure leakage from the production line.

[0123] If the compensation module 400 does not output an abnormality flag, the splicing module 500 outputs and archives the complete data stream for subsequent data backtracking analysis; at this time, the splicing module 500 does not trigger alarm or intervention commands to the external dispatch terminal.

[0124] To further illustrate the implementation process and technical effects of this invention, the following description is provided in conjunction with specific application scenarios and experimental data. The specific numerical values, scenario parameters, and comparative experiments described below are only used to explain the implementation principle of this invention and do not limit the scope of protection of this invention.

[0125] Taking an industrial production line data processing scenario for the manufacturing and circulation of a power battery pack as an example, the central computing server executes the calculation and analysis management process for the pressure data of the airbag tray. In this scenario, the airbag tray is used to carry the power battery pack and flows with the process. The internal pressure data of the airbag is used to reflect the tray's carrying status and the airbag's sealing status. The central computing server is pre-configured with a high-frequency time-series database and a preset mapping table for maintaining business attributes, and establishes a double-buffered queue in the memory pool. The central computing server runs server-side, client-side, and interface service modules. The server-side is used to receive data uploaded by the signal receiver, the client-side is used to maintain the binding relationship between the tray code and the sensor identifier, and the interface service module is used to interact with the external scheduling terminal. The tray code for target tracking is set as TP-100, and the old sensor identifier is set as ID_01.

[0126] After the power battery pack enters the corresponding process along with the airbag-equipped tray, a low-power wireless sensor collects the internal pressure data of the airbag and forwards it to the signal receiver via a repeater. The signal receiver transmits the corresponding binary data stream to the server of the central computing server. The access module 100 obtains the data packet by matching the frame header and frame tail. It decodes the data using Manchester encoding rules and performs cyclic redundancy check. It parses the data packet to obtain the sensor identifier ID_01, the original pressure of the airbag-equipped tray (30.1 kPa), and the corresponding timestamp (09:59:50). It allocates a foreground memory block and a background memory block in the memory pool and pushes the sensor identifier ID_01, the original pressure of the airbag-equipped tray, and the timestamp into the foreground memory block. When the accumulated number reaches the set capacity of 1000 records, it performs a read / write pointer switch, wakes up the background thread, and writes the data records in the background memory block into the high-frequency time-series database in batches.

[0127] When the client executes the binding modification operation between tray code TP-100 and the new sensor identifier ID_02, it generates a corresponding node binding update instruction. The decoupling module 200 listens to the front-end interface network packets, parses the node binding update instruction, and obtains the tray code TP-100 and the new sensor identifier ID_02. It performs an update operation in the preset mapping table, updating the end time of the row containing the old sensor identifier ID_01 associated with tray code TP-100 to the unbinding time 10:00:00. It then performs an insertion operation to generate a new data row, setting the start time of the new sensor identifier ID_02 to the binding time 10:00:50. Subtracting the unbinding time from the binding time yields the time window. ; The deserialization of external scheduling messages reveals the contextual feature of the power battery pack moving with an airbag-equipped pallet during the smooth conveyor transport process. Logical judgment is performed based on preset process classification rules. This contextual feature matches the feature set requiring a constant pressure trend, classifying it as a steady state. The enumeration value bound to the pallet code TP-100 is refreshed, and the corresponding real-time state is output as steady state.

[0128] When the real-time state is steady and the time window of 50 seconds is greater than zero, data compensation calculation is triggered. The last original pressure of 30.0 kPa corresponding to the unbinding time is extracted from the cache memory of the central computing server. This value represents the original pressure of the airbag-equipped tray that was last successfully reported by sensor ID_01 before the unbinding time of 10:00:00. The old original pressure of 30.1 kPa corresponding to the time node 10 seconds before the preset historical acquisition period is also extracted. The historical gradient is calculated using the original pressure of the airbag-equipped tray before the unbinding time. ; Predicted values ​​are calculated based on historical gradients and time windows: ; The initial pressure of the airbag-equipped tray reported by the new sensor ID_02 for five consecutive data acquisition cycles after the binding time was continuously extracted. The values ​​were 31.4 kPa, 31.5 kPa, 31.6 kPa, 31.5 kPa, and 31.5 kPa, respectively. The initial pressures of the airbag-equipped tray were summed and the average value was generated. ; The offset is generated by subtracting the mean of the original pressure of the airbag-equipped tray after the binding time from the predicted value: ; The offset of -2.0 kPa is added to the original pressure of 31.4 kPa of the airbag tray in the next cycle, which is then obtained to generate a compensation pressure of 29.4 kPa, and the result is associated and saved to the high-frequency time series database.

[0129] When the time interval between adjacent acquisition cycles is 1 second, the original pressure of the tray with airbag reported by the new sensor ID_02 in a subsequent acquisition cycle is obtained as 27.4 kPa. The offset of -2.0 kPa is added to it to obtain the compensation pressure of the current acquisition cycle as 25.4 kPa. The compensation pressure of the previous acquisition cycle as 29.4 kPa is extracted, and the instantaneous gradient is calculated. ; The instantaneous gradient of -4.0 kPa / s was compared with the preset threshold of -2.0 kPa / s. The instantaneous gradient deviated from the preset threshold, and the instantaneous gradients of the three consecutive acquisition cycles were -3.6 kPa / s, -4.0 kPa / s, and -3.8 kPa / s, respectively, all of which were less than the preset threshold of -2.0 kPa / s, thus generating an anomaly marker.

[0130] The time boundary is determined by retrieving the corresponding lower and upper time limits from the mapping table based on the tray code TP-100. The time axis retrieval range is defined, and using interval join operations, the original and compensated pressures of the airbag-equipped tray are extracted from the high-frequency time-series database in ascending order of timestamps, based on the old sensor identifier ID_01 and new sensor identifier ID_02 associated with the tray code TP-100 and the time boundary. In the array buffer of the running memory, the original and compensated pressures of the airbag-equipped tray are concatenated into a complete data stream, and the anomaly marker is pushed to the external scheduling terminal via network message payload through the interface service module. This completes the analysis and management process for the pressure data of the airbag-equipped tray during the manufacturing process of the power battery pack.

[0131] Data processing performance verification and anomaly labeling accuracy comparison: This embodiment further constructs a data test set consisting of data nodes corresponding to 1000 airbag-equipped trays. This test set covers 100 consecutive hours of steady-state and transient interactive operational data, and extracts 500,000 timestamped high-frequency concurrent raw pressure data records from the corresponding operational data as test samples. Application verification is conducted based on this data test set, and a comparison method is set as a benchmark.

[0132] The comparison method used for verification is the traditional direct write method. This method is based on the existing basic data pass-through process, which only performs data packet parsing and directly calls the database's single persistent interface for disk write operations. In hardware replacement scenarios, this method performs hard concatenation of the original values ​​for the old and new hardware data and determines pressure exceedance based on a single fixed global threshold. This method does not include a batch dumping step with double buffer queues, does not perform difference and mean offset compensation calculations based on time windows and historical gradients, and does not extract external scheduling messages for status classification of the operating context.

[0133] The airbag tray pressure data analysis and management method of the present invention applies the complete operation process of access module 100, decoupling module 200, tracking module 300, compensation module 400 and splicing module 500, covering memory-level double buffer queue read and write pointer switching, data extraction based on time boundary and interval connection operation, state classification combined with context features, and compensation pressure calculation and anomaly judgment based on predicted value, offset and instantaneous gradient.

[0134] The test sets two core data metrics. The first metric is write latency, determined by the combined single-cycle waiting time from receiving the underlying binary data stream to persistently storing the corresponding sensor identifier and the original pressure of the airbag tray into a high-frequency time-series database. The second metric is the false alarm rate for air leakage, determined by extracting the abnormal markers output by the system and comparing them with the actual airbag rupture records manually verified on the production line. The percentage of times reasonable pressure fluctuations caused by normal process interventions (such as transient conditions caused by robotic arm gripping or inflation stations) are incorrectly output as abnormal markers is calculated.

[0135] 500 high-concurrency monitoring and anomaly diagnosis application tasks were executed on this data set. The statistical distribution of the test data is as follows: Figure 8 As shown.

[0136] Under the test conditions of this embodiment, combined with Figure 8 The statistical results shown indicate that, Figure 8 As shown in (a), when applying this comparison method, the average write latency is 450 milliseconds; when applying the airbag tray pressure data analysis and management method of the present invention, the average write latency is 15 milliseconds. Figure 8 As shown in (b), when the comparison method is applied, the false alarm rate for air leakage is 18.5%; when the airbag tray pressure data analysis and management method of the present invention is applied, the false alarm rate for air leakage is 0.2%.

[0137] The above comparison shows that the previous comparison method, lacking a memory buffer management mechanism, frequently consumes disk input / output resources for single discrete data entries, resulting in high write latency. Furthermore, this method does not introduce time-axis decoupling and dynamic compensation across hardware cycles, nor does it distinguish between transient operating conditions that allow pressure fluctuations before an alarm, easily misjudging pressure fluctuations caused by normal process extrusion as abnormal leaks. The airbag tray pressure data analysis and management method of this invention constructs a dual-buffered queue using front-end and back-end memory blocks, performing batch writes when the storage pointer reaches a set capacity, reducing frequent single-entry write calls to the underlying database. Simultaneously, it identifies transient characteristics through external scheduling messages and uses offsets in steady state to reduce step benchmark differences caused by hardware replacements, improving the timing consistency of the compensated pressure data and reducing the impact of transient operating conditions and hardware benchmark differences on the anomaly marker generation results.

[0138] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for analyzing and managing airbag tray pressure data, characterized in that, include: The access module is used to parse data packets to obtain sensor identifiers, the original pressure of the airbag tray, and timestamps, and write the sensor identifiers, the original pressure, and the timestamps into a high-frequency time series database. The decoupling module is used to parse the node binding update instruction to obtain the tray code and the new sensor identifier. In the preset mapping table, the end time of the old sensor identifier associated with the tray code is updated to the unbinding time, the start time of the new sensor identifier is set as the binding time, and the time window is obtained by subtracting the unbinding time from the binding time. The tracking module is used to output the real-time status corresponding to the tray code based on the external scheduling message. The real-time status is either transient, steady-state, or an unidentified default abnormal state. The compensation module is used to calculate the historical gradient and predicted value using the original pressure before the unbinding time when the real-time state is the steady state and the time window is greater than zero, subtract the mean of the original pressure after the binding time from the predicted value to generate an offset, add the offset to the subsequent original pressure to generate a compensation pressure, and generate an anomaly mark when the instantaneous gradient of the compensation pressure deviates from a preset threshold. The splicing module is used to look up the time boundary in the preset mapping table according to the tray code, extract the original pressure and the compensation pressure according to the time boundary to splice them into a complete data stream, and push the anomaly mark to the external scheduling terminal.

2. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the access module, the step of writing the sensor identifier, the original pressure, and the timestamp into the high-frequency time series database includes: A double-buffered queue is constructed by allocating foreground and background memory blocks in the memory pool. The sensor identifier, the original pressure, and the timestamp are pushed into the foreground memory block; When the storage pointer of the foreground memory block reaches the set capacity, the read / write pointer is switched and the background thread is woken up. The sensor identifier, the original pressure and the timestamp in the background memory block are then written in batches into the high-frequency time series database.

3. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the decoupling module, the steps of updating the end time of the old sensor identifier associated with the tray code to the unbinding time in a preset mapping table, setting the start time of the new sensor identifier to the binding time, and subtracting the unbinding time from the binding time to obtain the time window include: Determine whether the old sensor identifier is empty. If the old sensor identifier is not empty, read the timestamp value of the moment when the node binding update instruction is received as the unbinding time, perform an update operation in the database where the preset mapping table is located, locate the record row containing the old sensor identifier, and set the end time of the record row as the unbinding time. A new data row containing the new sensor identifier is generated in the preset mapping table, and the start time of the new data row is set to the binding time; If the old sensor identifier is empty, the time window is set to zero; if the old sensor identifier is not empty, the binding time and the unbinding time are extracted and the difference is calculated to obtain the time window; if the time window is less than zero, the time window is assigned a value of zero.

4. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the tracking module, the step of outputting the real-time status corresponding to the tray code based on the external scheduling message includes: Deserialize the external scheduling message to obtain context features; The context features that allow pressure fluctuations are classified as the transient state, the context features that require pressure to maintain a constant trend are classified as the steady state, and the context features that do not match the preset process classification rules are classified as the unidentified default abnormal state. The enumeration value bound to the tray code is refreshed based on the classification result of the context features, and the enumeration value is output as the real-time status.

5. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the compensation module, the step of calculating the historical gradient and predicted value using the original pressure before the unbinding time includes: Extract the last original pressure corresponding to the unbinding time and the old original pressure corresponding to the preset historical collection period. The difference is obtained by subtracting the last original pressure from the old original pressure, and the difference is divided by the time span corresponding to the preset historical acquisition cycle to generate the historical gradient. The predicted value is obtained by multiplying the historical gradient with the time window, and the predicted value is generated by adding the last original pressure with the predicted value.

6. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the compensation module, the step of generating an offset by subtracting the binding time from the predicted value and taking the mean of the original pressure includes: The original pressure is continuously collected from the low-power wireless sensor corresponding to the new sensor identifier for a preset number of periods after the binding time. The average value is generated by summing the original pressure over the preset number of cycles and dividing by the number of cycles. The offset is generated by subtracting the mean from the predicted value.

7. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the compensation module, the step of generating an anomaly marker when the instantaneous gradient of the compensation pressure deviates from a preset threshold includes: The instantaneous gradient is obtained by calculating the rate of change of the compensation pressure in adjacent acquisition cycles. Determine whether the instantaneous gradient of multiple consecutive acquisition cycles is less than the preset threshold. If the instantaneous gradient of multiple consecutive acquisition cycles is less than the preset threshold, then generate the abnormal marker.

8. The airbag tray pressure data analysis and management system according to claim 1, characterized in that, In the splicing module, the step of looking up the time boundary in the preset mapping table based on the tray code includes: Retrieve historical binding records and current binding records associated with the tray code from the preset mapping table; Extract the earliest start time of all records in the search results as the lower time limit; Determine whether the end time of the currently bound record is null. If it is null, obtain the current time value of the central computing server as the time limit. If it is not null, extract the latest end time of all records in the search results as the time limit. The time boundary is determined based on the lower time limit and the upper time limit.

9. The airbag tray pressure data analysis and management system according to claim 8, characterized in that, In the stitching module, the step of extracting the original pressure and the compensated pressure according to the time boundary and stitching them into a complete data stream includes: Using the interval join operation, the original pressure and the compensated pressure are extracted from the high-frequency time series database in ascending order of timestamps, based on the old sensor identifier associated with the tray code, the new sensor identifier, and the time boundary. The extracted original pressure and the compensated pressure are written into an array buffer in the order they are stored, and then time-series merging is performed to form the complete data stream.

10. A method for analyzing and managing airbag tray pressure data, applied to the airbag tray pressure data analysis and management system as described in any one of claims 1-9, characterized in that, Includes the following steps: Parse the data packet to obtain the sensor identifier, the original pressure of the airbag tray, and the timestamp, and write the sensor identifier, the original pressure, and the timestamp into the high-frequency time series database; Parse the node binding update command to obtain the tray code and the new sensor identifier. In the preset mapping table, update the end time of the old sensor identifier associated with the tray code to the unbinding time, set the start time of the new sensor identifier to the binding time, and subtract the unbinding time from the binding time to obtain the time window. The real-time status corresponding to the tray code is output based on the external scheduling message. The real-time status is either transient, steady-state, or an unidentified default abnormal state. When the real-time state is the steady state and the time window is greater than zero, the historical gradient and predicted value are calculated using the original pressure before the unbinding time. The predicted value is subtracted from the mean of the original pressure after the binding time to generate an offset. The offset is added to the subsequent original pressure to generate a compensation pressure. When the instantaneous gradient of the compensation pressure deviates from a preset threshold, an anomaly marker is generated. The time boundary is retrieved from the preset mapping table based on the tray code. The original pressure and the compensation pressure are extracted according to the time boundary and combined into a complete data stream. The anomaly marker is then pushed to the external scheduling terminal.

Images