Intelligent data burning method and system based on data stream

Abstract

The invention relates to the technical field of electric data processing, in particular to an intelligent data burning method and system based on data streams. The method comprises the following steps that format recognition and semantic analysis are carried out by collecting input source data to be burnt in real time, so that metadata information corresponding to the data to be burnt is generated; hardware storage structure information and burning protocol specifications corresponding to the target equipment are obtained, data stream description output is carried out, and a corresponding data stream description file is generated; obtaining a storage bandwidth and a cache capacity corresponding to the target equipment, and carrying out burning dynamic task scheduling and low-density redundancy coding partitioning to generate fault-tolerant coding data stream partitioning; equipment state data corresponding to target equipment is monitored in real time, burning write-in processing and block-by-block CRC check are carried out, and meanwhile online repair verification analysis is carried out, so that a data stream burning verification completion result is generated. According to the method, a repair mechanism and dynamic resource scheduling can be designed to realize efficient and reliable data burning.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical data processing, and in particular to an intelligent data burning method and system based on data stream. Background Art

[0002] In the manufacturing process of modern electronic products, data burning refers to the critical process of writing firmware or configuration information into hardware devices (such as chips, memory, etc.). With the continuous advancement of technology, especially in areas such as the Internet of Things, smart hardware, and embedded systems, the requirements for data burning accuracy, efficiency, and intelligence are increasing. For example, the selection of burning programs and the data processing process for different versions of hardware devices require real-time optimization to ensure the efficiency and accuracy of the burning process. However, most current burning technologies use a single burning process, usually manually or by semi-automated equipment according to a preset program to complete the data writing step by step. This approach has certain drawbacks. When dealing with different device types and hardware configurations, it requires continuous adjustment of burning parameters based on hardware differences, and it is unable to intelligently optimize based on real-time device status data. This leads to a high error rate during the burning process and reduces burning efficiency. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a data stream-based intelligent data burning method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a data stream-based intelligent data burning method includes the following steps: Step S1: collecting input source data to be burned in real time through a bus interface, and performing format recognition and semantic analysis on the input source data to be burned to generate metadata information corresponding to the data to be burned; obtaining hardware storage structure information corresponding to a target device and a burning protocol specification, and outputting a data flow description of the metadata information corresponding to the data to be burned based on the hardware storage structure information corresponding to the target device and the burning protocol specification, thereby generating a data flow description file containing the metadata, structure information, and burning instructions; Step S2: Obtaining the storage bandwidth and cache capacity corresponding to the target device, and performing dynamic burning task scheduling for the data stream description file containing metadata, structural information, and burning instructions based on the storage bandwidth and cache capacity corresponding to the target device, to generate a data stream burning task sequence; Step S3: performing low-density redundant coding and segmenting on the corresponding data stream to be programmed in the data stream programming task sequence to generate fault-tolerant coded data stream segments; monitoring the device status data corresponding to the target device, including storage temperature, voltage fluctuation, and erase / write counts, in real time, and programming and writing the fault-tolerant coded data stream segments based on the device status data to generate device programming data stream segments; Step S4: Perform block-by-block CRC check on the device burned data stream blocks, locate the error blocks corresponding to the device burned data stream blocks, and at the same time use the corresponding redundant information in the fault-tolerant encoded data stream blocks and combine the device status data to perform online repair verification analysis on the error blocks to generate the result of the completion of the data stream burn verification.

[0005] Further, step S1 includes the following steps: Step S11: Configure the bus interface to include SPI, I2C, and UART interface protocols to collect the corresponding input source data to be burned in real time; Step S12: Perform format recognition and parsing on the input source data to be burned to design a multi-mode data parser to support automatic recognition and parsing of various data formats including binary, text, XML, and JSON, and obtain the parsed data of the input format to be burned; Step S13: Perform semantic parsing of data fields on the parsed data of the input format to be burned to generate metadata information corresponding to the data to be burned; Step S14: Obtain the hardware storage structure information corresponding to the target device and the burn protocol specification, where the hardware storage structure information includes flash block distribution, page address mapping table, and bad block marking information; Step S15: Based on the hardware storage structure information corresponding to the target device and the burn protocol specification, perform data stream description output on the metadata information corresponding to the data to be burned, and generate a data stream description file including metadata, structure information, and burn instructions.

[0006] Further, step S15 includes the following steps: Step S151: Obtain the corresponding semantic entities of the data to be burned through the metadata information corresponding to the data to be burned, and extract the entity relationships between the semantic entities of the data to be burned to obtain the data entity relationships between the entities of the data to be burned; Step S152: Based on the data entity relationships between the entities of the data to be burned, construct an entity relationship graph for the semantic entities of the data to be burned to generate an entity relationship graph of the data to be burned; Step S153: Generate executable burn instructions for the target device based on the burn protocol specification corresponding to the target device for the entity relationship graph of the data to be burned to generate executable burn instructions for the target device corresponding to the data to be burned; Step S154: Based on the hardware storage structure information corresponding to the target device and in combination with the executable burn instructions for the target device corresponding to the data to be burned, perform data stream description output on the metadata information corresponding to the data to be burned, and generate a data stream description file including metadata, structure information, and burn instructions.

[0007] Further, step S2 includes the following steps: Step S21: Obtain the storage bandwidth corresponding to the target device through continuous write testing; Step S22: Calculate and obtain the cache capacity corresponding to the target device according to the page address mapping table corresponding to the target device; Step S23: Perform a burn-in dependency analysis on the data stream description file containing metadata, structure information, and burn-in instructions, to mine and analyze the dependencies of each data stream to be burned in the data stream description file based on the burn-in instructions, and generate the data burn-in dependency relationships between the data streams to be burned; Step S24: Perform a directed acyclic construction on the data streams to be burned in the data stream description file based on the data burn-in dependency relationships between the data streams to be burned, to generate a directed acyclic graph of the data streams to be burned; perform a topological burn-in sorting on the data streams to be burned in the data stream description file according to the directed acyclic graph of the data streams to be burned, to generate an initial burn-in sorting sequence of the data streams; Step S25: Perform a burn-in dynamic task scheduling on the initial burn-in sorting sequence of the data streams based on the storage bandwidth and cache capacity corresponding to the target device, to generate a data stream burn-in task sequence.

[0008] Further, step S25 includes the following steps: Step S251: Divide the initial burn-in sorting sequence of the data streams into each data segment, obtain the corresponding data volume size through each data segment, and at the same time construct a data stream burn-in transmission topology network based on the data transmission order and dependencies corresponding to each data segment in the initial burn-in sorting sequence of the data streams and in combination with the data volume size corresponding to each data segment, where the data segment is used as a node, the data volume size is used as a node attribute, and the data transmission order and dependencies are used as edges; Step S252: Calculate the betweenness centrality corresponding to each node through the data stream burn-in transmission topology network, and determine the burn-in transmission priority of each data segment based on the betweenness centrality corresponding to each node, to generate a data stream burn-in transmission priority sequence; Step S253: Evaluate the bandwidth burn-in transmission pressure on the data segments corresponding in the data stream burn-in transmission priority sequence based on the storage bandwidth corresponding to the target device, to obtain the storage bandwidth burn-in transmission pressure corresponding to each data segment in the target device; Step S254: Perform a cache transmission saturation quantization on the data segments corresponding in the data stream burn-in transmission priority sequence based on the cache capacity corresponding to the target device, to obtain the cache burn-in transmission saturation rate corresponding to each data segment in the target device; Step S255: Analyze the current load of the target device based on the storage bandwidth burn-in transmission pressure and cache burn-in transmission saturation rate corresponding to each data segment in the target device, so as to obtain the current burn-in load amount corresponding to the target device; dynamically adjust the burn-in transmission rate corresponding to each data segment according to the current burn-in load amount corresponding to the target device, and perform burn-in dynamic task scheduling on the initial burn-in sorting sequence of the data stream based on the burn-in transmission rate corresponding to each data segment to generate a data stream burn-in task sequence.

[0009] Further, step S3 includes the following steps: Step S31: Perform low-density redundant encoding and chunking on the data stream to be burned in the data stream burn-in task sequence to generate fault-tolerant encoded data stream chunks; Step S32: Monitor the storage temperature corresponding to the target device in real time; Step S33: Monitor the voltage fluctuation corresponding to the target device in real time; Step S34: Monitor the number of erase and write cycles corresponding to the target device in real time; Step S35: Use the storage temperature, voltage fluctuation, and number of erase and write cycles corresponding to the target device as device status data, and perform burn-in write processing on the fault-tolerant encoded data stream chunks based on the device status data to generate device burn-in data stream chunks.

[0010] Further, step S31 includes the following steps: Perform chunking on the data stream to be burned in the data stream burn-in task sequence to obtain each data stream chunk to be burned; Obtain the data content to be burned corresponding to each chunk through each data stream chunk to be burned, and perform content-related grouping on each data stream chunk to be burned based on the data content to be burned corresponding to each chunk to obtain key data stream chunks to be burned and non-key data stream chunks to be burned; Obtain the content correlation coefficient and the number of erase and write cycles of the corresponding data chunks through the key data stream chunks to be burned and the non-key data stream chunks to be burned, calculate the optimal number of parity bits of the corresponding data chunks based on the content correlation coefficient and the number of erase and write cycles, and at the same time use high redundancy encoding for the key data stream chunks to be burned and low redundancy encoding for the non-key data stream chunks to be burned to generate key encoded data stream chunks to be burned and non-key encoded data stream chunks to be burned; Generate fault-tolerant encoded data stream chunks by adding index tags containing position information and redundancy information to each key encoded data stream chunk to be burned and non-key encoded data stream chunk to be burned.

[0011] Further, the burn-in write processing of the fault-tolerant encoded data stream chunks based on the device status data in step S35 includes the following steps: Obtain the historical programming status data of the device, and train a Bayesian network model based on the historical programming status data of the device to predict the writing error probability corresponding to the data stream block of the target device; Input the current corresponding device status data into the Bayesian network model to perform writing prediction analysis of the target device on the fault-tolerant encoded data stream block, so as to obtain the writing error probability corresponding to each fault-tolerant encoded data stream block; Based on the writing error probability corresponding to each fault-tolerant encoded data stream block, perform programming writing processing on the corresponding fault-tolerant encoded data stream block. If it is determined as a high error probability block according to the writing error probability, increase the programming voltage, and the increase amplitude is less than or equal to 5% of the standard programming voltage, and reduce the clock frequency, and the reduction amplitude is less than or equal to 10% of the standard clock frequency; if it is determined as a low error probability block according to the writing error probability, increase the programming voltage, and the increase amplitude is less than 5% of the high error probability block, and increase the clock frequency, and the increase amplitude is less than 10% of the high error probability block, so as to perform the writing operation of each fault-tolerant encoded data stream block to generate the device programming data stream block.

[0012] Further, step S4 includes the following steps: Step S41: Perform block-by-block CRC check on the device programming data stream block to obtain the device programming check success block and the device programming check failure block; Step S42: Obtain the corresponding block address and error type through the device programming check failure block, and locate the error block corresponding to the device programming check failure block based on the block address and error type; Step S43: Obtain the index label of the error block in the fault-tolerant encoded data stream block based on the block address, and obtain the corresponding redundant information of the error block in the fault-tolerant encoded data stream block based on the index label. At the same time, use the corresponding redundant information of the error block in the fault-tolerant encoded data stream block to perform online repair on the error block, so as to use the redundant information to repair the error block by using the iterative decoding algorithm. If the repair fails, trigger the re-programming process, otherwise perform iterative CRC check until the check is successful; Step S44: Perform programming verification analysis on the device programming check success block corresponding to the successful CRC check based on the device status data, so as to calculate the bit error rate and the remaining erase-write life corresponding to each programming block according to the device status data, and verify the bit error rate and the remaining erase-write life corresponding to each programming block according to the preset threshold. If the verification fails, repair and check again to generate the data stream programming verification completion result.

[0013] Further, the present invention also provides an intelligent data programming system based on a data stream for executing the intelligent data programming method based on a data stream as described above. The intelligent data programming system based on a data stream includes: The data stream description output module to be burned is used to collect the input source data to be burned in real time through the bus interface, and perform format recognition and semantic parsing on the input source data to be burned, so as to generate metadata information corresponding to the data to be burned; obtain the hardware storage structure information and the burning protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the data to be burned based on the hardware storage structure information and the burning protocol specification corresponding to the target device, so as to generate a data stream description file containing metadata, structure information and burning instructions; The burning dynamic scheduling module is used to obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burning dynamic task scheduling on the data stream description file containing metadata, structure information and burning instructions based on the storage bandwidth and cache capacity corresponding to the target device, so as to generate a data stream burning task sequence; The data stream block burning and writing module is used to perform low-density redundant coding on the data stream to be burned in the data stream burning task sequence to generate a fault-tolerant coded data stream block; monitor the device state data corresponding to the target device in real time, including storage temperature, voltage fluctuation and erasure count, and perform burning and writing processing on the fault-tolerant coded data stream block based on the device state data to generate a device burning data stream block; The burning error online repair module is used to perform block-by-block CRC check on the device burning data stream block, locate the error block corresponding to the device burning data stream block, and perform online repair verification analysis on the error block by using the corresponding redundant information in the fault-tolerant coded data stream block and combining the device state data, so as to generate a data stream burning verification completion result.

[0014] The beneficial effects of the present invention: 1. Compared with the prior art, the intelligent data burning method based on data flow proposed by the present invention has the beneficial effect that the data of the input source to be burned is collected in real time through the bus interface. This process is crucial for ensuring the accuracy and timeliness of the data. By identifying the format and parsing the semantics of the input source data, the original data can be converted into metadata with structured information, which not only improves the efficiency of data processing but also provides a solid foundation for subsequent operations. During the generation of metadata, a comprehensive analysis and precise identification of the data are involved, enabling it to correspond to the hardware structure and burning specifications of the target device. Obtaining the hardware storage structure information and burning protocol specifications of the target device is the core part of this step, which can effectively guide the output of the subsequent data flow description and ensure that the burning process meets the actual requirements of the device. The generated data flow description file contains metadata, hardware structure information, and burning instructions, which is crucial for the subsequent data transmission and burning process. The greatest advantage of this step lies in its high degree of automation and flexibility, which can provide personalized burning solutions for different devices and significantly improve the accuracy and stability of the burning process. Secondly, by obtaining the storage bandwidth and cache capacity of the target device, resources are reasonably scheduled to determine the order of burning tasks. The storage bandwidth and cache capacity directly affect the speed and efficiency of data burning. Therefore, when scheduling the data flow description file, these hardware parameters need to be carefully optimized. The core goal of dynamic task scheduling is to reasonably allocate the burning process of the data flow to ensure that each burning task can be completed within the specified time and avoid performance degradation caused by problems such as insufficient bandwidth or cache overflow. By precisely scheduling the task sequence, the burning efficiency can be improved while ensuring the burning quality. This process makes the burning of the data flow no longer a single, linear task but a dynamically adjustable and flexible process that can adapt to the different performance characteristics of the target device, making the entire burning process more efficient and accurate. Then, the data flow to be burned in the data flow burning task sequence needs to be processed by low-density redundant coding and block division. The main purpose of this process is to improve the fault tolerance of the data during the burning process. Redundant coding and block division can provide a basis for repairing possible device failures by adding additional data redundancy information. During the burning process, the state of the target device, such as storage temperature, voltage fluctuation, and number of erase-write cycles, will affect the storage and burning process of the data. Real-time monitoring of these device state data helps to dynamically adjust the burning strategy to ensure that the data can be stored under the best conditions. The burning and writing process based on the device state data can prevent burning failure problems caused by excessive temperature, unstable voltage, or excessive number of erase-write cycles. Further intelligent optimization based on the real-time state data of the device. The advantage of this stage is that it fully considers various actual situations during the operation of the device, thereby reducing the error rate of the data flow during the burning process.Finally, by performing block-by-block CRC check on the device programming data stream, it is ensured that the data of each block is not damaged or in error during the programming process, and errors in the data transmission process can be quickly identified and located. When an error block is detected, the redundant information in the fault-tolerant encoded data stream block can be used to repair the error online. Combining with the device status data, this process further improves the accuracy and efficiency of error repair. At the same time, the repaired data stream can ensure that the final programming result meets the expected standard through verification and analysis. This method greatly improves the reliability of the programming process, thus ensuring that the device can stably store and read data during long-term operation.

[0015] 2. The intelligent data programming system based on data stream proposed by the present invention is generally composed of a programming data stream description and output module, a programming dynamic scheduling module, a data stream block programming and writing module, and a programming error online repair module, and can implement any intelligent data programming method based on data stream described in the present invention. It is used to realize the intelligent data programming method based on data stream through the operation between computer programs running on each module. The internal structure of the system cooperates with each other, which can greatly reduce repetitive work and manpower input, and can quickly and effectively provide a more accurate and efficient intelligent data programming process based on data stream, thus simplifying the operation process of the intelligent data programming system based on data stream. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects, and advantages of the present invention will become more apparent by reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 It is a schematic flow chart of the steps of the intelligent data programming method based on data stream of the present invention; Figure 2 is Figure 1 a detailed schematic flow chart of step S1 in Figure 3 is Figure 2 a detailed schematic flow chart of step S15 in DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0018] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides an intelligent data programming method based on data stream, and the method includes the following steps: Step S1: Real-time collect the input source data to be burned through the bus interface, and perform format recognition and semantic parsing on the input source data to be burned to generate metadata information corresponding to the data to be burned; obtain the hardware storage structure information and the burning protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the data to be burned based on the hardware storage structure information and the burning protocol specification corresponding to the target device, and generate a data stream description file containing metadata, structure information, and burning instructions; Step S2: Obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burning dynamic task scheduling on the data stream description file containing metadata, structure information, and burning instructions based on the storage bandwidth and cache capacity corresponding to the target device to generate a data stream burning task sequence; Step S3: Perform low-density redundant coding and chunking on the data stream to be burned in the data stream burning task sequence to generate fault-tolerant coded data stream chunks; monitor in real time the device status data corresponding to the target device, including storage temperature, voltage fluctuation, and erasing / writing times, and perform burning and writing processing on the fault-tolerant coded data stream chunks based on the device status data to generate device-burned data stream chunks; Step S4: Perform block-by-block CRC check on the device-burned data stream chunks, locate the error chunks corresponding to the device-burned data stream chunks, and at the same time use the redundant information corresponding to the fault-tolerant coded data stream chunks and combine with the device status data to perform online repair verification analysis on the error chunks to generate a result indicating the completion of data stream burning verification.

[0019] In the embodiment of the present invention, please refer to Figure 1 As shown in the figure, it is a schematic diagram of the step flow of the intelligent data burning method based on data stream of the present invention. In this example, the intelligent data burning method based on data stream includes the following steps: Step S1: Real-time collect the input source data to be burned through the bus interface, and perform format recognition and semantic parsing on the input source data to be burned to generate metadata information corresponding to the data to be burned; obtain the hardware storage structure information and the burning protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the data to be burned based on the hardware storage structure information and the burning protocol specification corresponding to the target device, and generate a data stream description file containing metadata, structure information, and burning instructions; In an embodiment of the present invention, the input source data to be burned is collected in real time through a PCIe 4.0 bus interface. The collection frequency is 100 MHz, and the transmission rate reaches 32 GB / s. The data collection module is built with a format recognition engine. Based on the regular expression matching algorithm, this engine can recognize more than 10 file system formats such as FAT32, NTFS, and EXT4, and perform a sector-by-sector scan on the input source data to be burned. For every 512 bytes of data scanned, the format recognition engine compares with a preset file system feature library. For example, when it detects that a data block contains the "FAT" identifier and the subsequent bytes conform to the cluster allocation table structure of the FAT32 file system, the data format can be determined as FAT32. The semantic parsing module deeply analyzes the data with the recognized format and extracts metadata such as file name, size, creation time, and permission information. For a piece of data to be burned containing 100 files, the semantic parsing module can complete the parsing within 5 seconds and generate a metadata list containing information such as file name, file type, file size, and timestamp. At the same time, by querying the hardware specification database of the target device, the corresponding hardware storage structure information is obtained, such as the storage medium type (NAND Flash, NOR Flash, etc.), page size (usually 2 KB, 4 KB, etc.), block size (usually 128 KB, 256 KB, etc.), and burn protocol specification (such as JTAG, SPI, I2C, etc.). Based on the obtained hardware storage structure information and burn protocol specification, the data stream description generator converts the metadata information into a data stream description file in a specific format. This file contains three parts: the metadata part, which records the basic information of the file; the structure information part, which divides the data into storage-suitable units according to the storage structure of the target device, such as splitting a large file into multiple data blocks matching the device page size; and the burn instruction part, which generates specific burn operation instructions according to the burn protocol specification, such as address setting, data transmission, verification, etc. instructions. The finally generated data stream description file adopts the XML format and its size is 5% of the original data, providing a structured data basis for subsequent burn task scheduling.

[0020] Step S2: Obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burn dynamic task scheduling on the data stream description file containing metadata, structure information, and burn instructions based on the storage bandwidth and cache capacity corresponding to the target device to generate a data stream burn task sequence; In an embodiment of the present invention, by querying the hardware specification parameters of the target device, its storage bandwidth is obtained as 50 MB / s and the cache capacity is 256 MB. Based on these parameters, the task scheduler uses the dynamic time window algorithm to schedule the burning tasks for the data stream description file. First, the data blocks in the data stream description file are sorted according to size and dependency relationship to form an initial task queue. Then, according to the storage bandwidth and cache capacity, the amount of data that can be processed within each time window (set to 100 ms) is calculated, that is, 50 MB / s × 0.1 s = 5 MB. The task scheduler sequentially selects data blocks from the initial task queue to form task groups with a size not exceeding 5 MB. Each task group contains 10 - 15 data blocks. During the selection process, critical data blocks and data blocks with dependency relationships are preferentially processed. For example, for a data burning task containing a system boot program and an application program, the system boot program data block will be preferentially arranged in the previous task groups. At the same time, considering the limitation of the cache capacity, ensure that the data of each task group can be fully loaded into the device cache to avoid cache overflow. When a task group is executed, the task scheduler will dynamically adjust the composition and execution order of subsequent task groups according to the real-time state of the device (such as cache usage, storage bandwidth utilization, etc.). For example, if it is found that the device cache usage is close to the upper limit, the size of the task group will be appropriately reduced; if the storage bandwidth utilization is low, the amount of data in the task group will be increased. Through this dynamic adjustment, a data stream burning task sequence containing 100 task groups is finally generated. Each task group has a clear execution order and time arrangement to ensure the efficiency and stability of the data burning process.

[0021] Step S3: Perform low-density redundant coding and chunking on the data stream to be burned corresponding to the data stream burning task sequence to generate fault-tolerant coded data stream chunks; monitor in real time the device state data of the target device corresponding to the storage temperature, voltage fluctuation, and number of erasure and write cycles, and perform burning and writing processing on the fault-tolerant coded data stream chunks based on the device state data to generate device-burned data stream chunks; In an embodiment of the present invention, the data stream to be burned in the data stream burning task sequence is block-coded by adopting a BCH (15,5) low-density redundancy coding algorithm. The size of each data block is set to 16 KB. The coding module starts from the head of the task sequence, reads data in sequence, and generates an independent block every time 16 KB is read. For each block, the coding module divides it into multiple 15-bit data blocks, where 5 bits are the original data and 10 bits are the parity bits. The parity bits are calculated through a specific generating polynomial. For example, for the original data 10110, the calculated parity bits are 1100101110, and the two are combined to form a 15-bit encoded data block. During the burning process, the storage temperature of the target device is monitored in real time by a DS18B20 temperature sensor with a sensor accuracy of ±0.5 °C and a sampling period of 200 ms; the voltage fluctuation is monitored in real time by an INA219 voltage monitoring chip with a monitoring accuracy of ±0.01 V and a sampling period of 100 ms; the number of erasure and write cycles is obtained in real time by reading the page address mapping table of the storage medium with a reading period of 500 ms. These three items of data are used as device status data and input into the burning parameter adjustment model. The model dynamically adjusts the burning parameters according to the device status data. When the storage temperature exceeds 30 °C, the burning voltage is increased by 5%, from 3.3 V to 3.465 V; when the voltage fluctuation exceeds 0.1 V, the burning clock frequency is reduced by 10%, from 50 MHz to 45 MHz; when the number of erasure and write cycles exceeds 100 times, the number of parity bits is increased, from 10 bits to 12 bits. For example, when the detected storage temperature is 32 °C, the voltage fluctuation is 0.12 V, and the number of erasure and write cycles is 120 times, the above three adjustment measures are applied simultaneously. In this way, it is ensured that the data burning process can maintain a high success rate under different device states, and finally the device burning data stream blocks are generated.

[0022] Step S4: Perform block-by-block CRC check on the device burning data stream blocks, locate the error blocks corresponding to the device burning data stream blocks, and at the same time use the redundant information corresponding in the fault-tolerant coded data stream blocks and combine the device status data to perform online repair verification analysis on the error blocks to generate the result of the completion of the data stream burning verification.

[0023] In the embodiment of the present invention, by using the CRC-32 checksum algorithm to perform block-by-block verification on the device programming data stream, a polynomial of 0x04C11DB7 is generated. The verification module sequentially reads each block from the device storage medium, reads 16KB of data each time, calculates its CRC value, and compares it with the original CRC value recorded during the programming process. If they are inconsistent, it is determined that there is an error in the block, and the physical address of the error block in the storage medium is located through the address resolution module. For example, if it is found that the CRC value of the BLOCK_005 block does not match, its physical address range is located from 0x00010000 to 0x00013FFF, and the redundant information of the fault-tolerant encoded data stream block generated in step S3 is used to perform online repair on the error block. For the BCH-encoded block, the repair module first constructs a check equation based on the check bits and the original data, and then gradually corrects the error bits through the iterative decoding algorithm. After each iteration, the CRC value is recalculated for verification. If the repair is still not successful after 5 iterations, the reprogramming process is triggered. For example, for the BLOCK_005 block, the repair module determines that there are 3 error bits through analyzing its redundant information and successfully repairs it after 3 iterations. During the repair process, the repair effect is evaluated in combination with the device status data. If the storage temperature is too high (exceeding 40°C) or the voltage fluctuation is too large (exceeding 0.2V), even if the CRC check passes, the repair is considered unreliable and the repair is performed again. After the repair is completed, the CRC check is performed again until the check is successful. For all blocks that pass the verification, the bit error rate and the remaining erase-write life are calculated based on the device status data. If the bit error rate exceeds 0.001% or the remaining erase-write life is less than 80%, the block is repaired again or marked as a potential risk block. Finally, after all blocks pass the verification, a data stream programming verification completion result including the verification result, repair record, bit error rate, and remaining erase-write life is generated.

[0024] Further, as an embodiment of the present invention, refer to Figure 2 shown in Figure 1 is a detailed step flow diagram of step S1 in Step S11: Configure the bus interface to include SPI, I2C, and UART interface protocols to collect the corresponding input source data to be programmed in real time; In an embodiment of the present invention, a dedicated data acquisition module is adopted to collect the input source data to be programmed in real time through a bus interface. This module integrates SPI, I2C, and UART interfaces, supports simultaneous parallel acquisition of multiple protocol data. The SPI interface operates in mode 3 (CPOL = 1, CPHA = 1), with a maximum clock frequency of 10 MHz, and uses a 4-wire system (SCK, MISO, MOSI, SS) for communication; the I2C interface operates in fast mode, with a transmission rate of 400 kHz, and supports 7-bit and 10-bit addressing; the UART interface has a baud rate set to 115200 bps, 8 data bits, 1 stop bit, and no parity bit. During the acquisition process, the module uses a hardware timer to achieve precise timing control to ensure the accuracy of data acquisition. For the SPI interface, when the SS signal is detected to be pulled low, data acquisition is started, and data is received synchronously according to the clock signal; for the I2C interface, the data transmission direction and target device are determined by detecting the start bit and address bit; for the UART interface, asynchronous data reception is achieved through a baud rate generator and a data buffer. The acquired data is stored in the SRAM of the module in frames, and each frame of data contains information such as a timestamp, interface type, and data length, forming the input source data to be programmed.

[0025] Step S12: Perform format recognition and parsing on the input source data to be programmed, and design a multi-mode data parser to support automatic recognition and parsing of various data formats including binary, text, XML, and JSON, to obtain the parsed data of the input format to be programmed; In an embodiment of the present invention, by performing format recognition and parsing on the input source data to be programmed, a multi-mode data parser is used. This parser integrates various format recognition algorithms and can automatically recognize data formats such as binary, text, XML, and JSON. First, the data format is recognized through file header features. For binary files, check the specific byte sequence of the file header, such as the start marker of 0xFFD8 for JPEG files; for text files, detect whether they contain ASCII printable characters; for XML files, check whether they contain "<" and ">" tag symbols; for JSON files, check whether they contain "{" and "}" or "[" and "]" symbols. If the file header features are not obvious, further recognition is performed through statistical analysis methods. For example, calculate the occurrence frequency of different characters in the data. XML and JSON files usually contain more specific symbols, while the character distribution of binary files is relatively random. For the recognized different format data, corresponding parsing algorithms are used for parsing. For XML files, a SAX parser is used to parse line by line to build a document object model; for JSON files, a recursive descent parser is used for parsing to convert it into a key-value pair structure. The parsed data is uniformly converted into an internal standard format to form the parsed data of the input format to be programmed.

[0026] Step S13: Perform semantic parsing on the data field of the data to be programmed in the input format to generate metadata information corresponding to the data to be programmed; In an embodiment of the present invention, by performing semantic parsing on the data field of the data to be programmed in the input format, a semantic parsing engine is used. Based on a predefined semantic rule library and a machine learning model, the engine can identify the meanings and relationships of data fields. First, the data is tokenized, splitting the continuous data string into meaningful words or phrases. For structured data such as XML and JSON, the tag names and attribute names are directly extracted as the tokenization results; for unstructured data, tokenization algorithms in natural language processing are used. Then named entity recognition is performed to identify the entity types in the data, such as device ID, firmware version, timestamp, etc. A trained BiLSTM-CRF model is used for entity recognition, and the F1 value of this model on a specific domain dataset reaches 0.95. Next, relationship extraction is performed to analyze the relationships between entities, such as "belong to", "associate with", "temporal precedence", etc. Finally, based on the identified entities and relationships, metadata information corresponding to the data to be programmed is generated, including information such as data source, data type, data structure, data meaning, etc., forming a structured metadata description.

[0027] Step S14: Obtain the hardware storage structure information corresponding to the target device and the programming protocol specification, where the hardware storage structure information includes flash block distribution, page address mapping table, and bad block marking information; In an embodiment of the present invention, by obtaining the hardware storage structure information corresponding to the target device and the programming protocol specification, the hardware storage structure information is stored in a dedicated hardware information database. This database adopts a distributed architecture and has high availability and fast query capabilities. Through the model number and serial number of the target device, the corresponding hardware storage structure information is retrieved from the database, including flash block distribution, page address mapping table, and bad block marking information. The flash block distribution information describes the physical structure of the flash chip, such as the total number of blocks, the number of pages per block, the number of bytes per page, etc.; the page address mapping table records the mapping relationship from logical address to physical address for address conversion; the bad block marking information records the damaged blocks in the flash to avoid writing data to these blocks. The programming protocol specification is stored in a protocol library, and the corresponding programming protocol specification is retrieved through the device model, including information such as programming command format, data transmission method, and verification algorithm, providing a basis for generating programming instructions later.

[0028] Step S15: Based on the hardware storage structure information corresponding to the target device and the programming protocol specification, perform data stream description output on the metadata information corresponding to the data to be programmed, and generate a data stream description file containing metadata, structure information, and corresponding programming instructions.

[0029] In an embodiment of the present invention, based on the hardware storage structure information corresponding to the target device and the flashing protocol specification, a data stream description of the metadata information corresponding to the data to be flashed is output. A data stream generator is used, which determines the storage location and method of the data to be flashed in the target device according to the hardware storage structure information. For example, according to the flash block distribution and the page address mapping table, the physical address where the data should be stored is calculated; according to the bad block marking information, the damaged blocks are avoided. Combining the flashing protocol specification, corresponding flashing instructions are generated. For each data block, a flashing instruction containing address information, data length, check code, etc. is generated. The metadata information, hardware storage structure information, and flashing instructions are integrated and organized in a standardized data stream description format to generate a data stream description file corresponding to the metadata, structure information, and flashing instructions. This file adopts a hierarchical structure, clearly describing the data source, storage method, and flashing process, providing complete guidance for the actual data flashing operation.

[0030] Further, as an embodiment of the present invention, refer to Figure 3 shown in Figure 2 for the detailed step flow diagram of step S15. In this embodiment, step S15 includes the following steps: Step S151: Obtain the corresponding semantic entities of each data to be flashed through the metadata information corresponding to the data to be flashed, and extract the entity relationships between the semantic entities of each data to be flashed to obtain the data entity relationships between the data entities of each data to be flashed; In an embodiment of the present invention, each semantic entity of the data to be burned is extracted from the metadata information of the data to be burned. The metadata information is stored in a dedicated metadata warehouse. The warehouse adopts a distributed file system architecture and has the ability to process 100,000 records per second. The semantic entities of the data to be burned are extracted with the help of a natural language processing tool. The tool has a built-in named entity recognition model and uses a BERT pre-trained language model. It is fine-tuned on a specific domain dataset and achieves an F1 value of 0.95. For example, for a data file containing "Device ID: D20250617, Firmware version: V2.3.1, Production batch: P202503, Verification code: 0xABCD1234", the tool can accurately identify semantic entities such as "device ID", "firmware version", "production batch", and "check code". Subsequently, it extracts entity relationships and analyzes the grammatical relationships between various semantic entities through dependency parsing technology. For the sentence "The firmware version of device D20250617 is V2.3.1", by analyzing the dependency structure of the sentence, it is determined that there is an "attribute-value" relationship between "device ID" and "firmware version". Then, using the remote supervision algorithm and predefined relationship templates, it automatically obtains training data from a large amount of unlabeled text to further identify the relationship between entities. For example, there is a "production association" relationship between "production batch" and "device ID". Finally, the data entity relationship between each data entity to be burned is obtained.

[0031] Step S152: constructing an entity relationship graph for each semantic entity of the data to be programmed based on the data entity relationship between each entity of the data to be programmed, so as to generate an entity relationship graph of the data to be programmed; In an embodiment of the present invention, an entity relationship graph for the data to be burned is constructed based on previously obtained data entity relationships. The graph database Neo4j is used as a storage and processing platform. This database supports 100,000 graph operations per second, enabling efficient storage and querying of entity relationships. Each semantic entity of the data to be burned is represented as a node in the graph, and the data entity relationships between entities are represented as edges. For example, semantic entities such as "Device ID: D20250617," "Firmware Version: V2.3.1," "Production Batch: P202503," and "Verification Code: 0xABCD1234" are converted into nodes in the graph. Each node contains attributes such as entity type, name, and value. The "attribute-value" relationship between "Device ID" and "Firmware Version" is converted into a directed edge of the "attribute-value" type. Similarly, the "production association" relationship between "Production Batch" and "Device ID" is also converted into a directed edge. In this way, all semantic entities of the data to be burned and their relationships are converted into a graph structure, generating an entity relationship graph for the data to be burned. This diagram comprehensively shows the relationship between the various parts of the data to be burned, providing a structured data basis for the subsequent generation of executable burning instructions.

[0032] Step S153: Generate executable programming instructions for the entity relationship diagram of the data to be programmed based on the programming protocol specification corresponding to the target device, so as to generate executable programming instructions for the target device corresponding to the data to be programmed; In the embodiment of the present invention, by generating executable programming instructions for the entity relationship diagram of the data to be programmed based on the programming protocol specification corresponding to the target device, the programming protocol specification of the target device is stored in a dedicated protocol library. The protocol library adopts a classified and hierarchical storage structure and contains programming protocol specifications of more than 1000 different devices. A protocol parsing engine is used to parse the protocol specification. This engine can identify information such as command formats, parameter definitions, and data transmission methods in the protocol. For example, for a certain model of target device, its programming protocol specification requires the programming instruction format to be "CMD + parameter 1 + parameter 2 + checksum", where "CMD" is the command code, "parameter 1" is the device ID, and "parameter 2" is the data block address. Extract relevant information from the entity relationship diagram of the data to be programmed, convert "device ID: D20250617" into the format required by the protocol specification, determine the data block address, calculate the checksum, and according to the command definition in the protocol specification, combine this information into an executable programming instruction, such as "CMD001 + D20250617 + 0x0000 + 0xABCD". For all data entities in the entity relationship diagram, parse and convert them according to the protocol specification to generate a series of executable programming instructions for the target device, ensuring that the instructions meet the programming requirements of the target device.

[0033] Step S154: Output a data stream description of the metadata information corresponding to the data to be programmed based on the hardware storage structure information corresponding to the target device and in combination with the executable programming instructions for the target device corresponding to the data to be programmed, and generate a data stream description file containing metadata, structure information, and the corresponding programming instructions.

[0034] In an embodiment of the present invention, by based on the hardware storage structure information corresponding to the target device and combining with the burn-in instructions executable by the target device for the data to be burned, the metadata information corresponding to the data to be burned is output in the form of a data stream description. The hardware storage structure information of the target device is stored in a dedicated hardware information library, which contains detailed information such as storage partitions, address ranges, read / write permissions of various devices. For example, the hardware storage structure of a certain target device is divided into a boot area (address range 0x0000 - 0x1000), a firmware area (address range 0x1000 - 0x8000), and a configuration area (address range 0x8000 - 0xFFFF). According to the nature of the data to be burned and the burn-in instructions, the storage location of the data in the hardware storage is determined. For firmware data, according to the burn-in instructions, it is determined that it should be stored at the position starting from the address 0x1000 in the firmware area. By combining the metadata information, information such as the source, use, and format of the data is associated with the hardware storage structure information and the burn-in instructions. Using a standardized data stream description language, these information are organized into a structured document, generating a data stream description file containing metadata, structure information, and burn-in instructions. This file clearly describes the storage method and burn-in process of the data to be burned in the target device, providing detailed guidance for the actual data burn-in operation.

[0035] Further, step S2 includes the following steps: Step S21: Obtain the storage bandwidth corresponding to the target device through continuous write testing; In an embodiment of the present invention, continuous write testing is performed by using a dedicated storage performance testing device to obtain the storage bandwidth corresponding to the target device. The testing device is equipped with a high-speed data transmission interface and is connected to the target device through an SPI bus, and the working frequency is set to 20 MHz. During the testing process, a continuous data block with a size of 128 MB is generated, and the data block is filled with a random byte sequence to ensure the randomness of the data. The data block is continuously written to the storage medium of the target device at a fixed transmission rate. After each write operation is completed, the start time and end time of the data transmission are recorded, accurate to the microsecond level. Ten consecutive write tests are performed, and the interval between each test is 5 seconds to avoid the influence of factors such as overheating of the storage device on the test results. For example, the first write of 128 MB of data takes 6.5 seconds, the second takes 6.3 seconds, and so on. The bandwidth value of each test is calculated through the formula "storage bandwidth = data volume / transmission time". The results of the ten tests are arithmetically averaged to obtain the final storage bandwidth. Assuming that the bandwidth values calculated from the ten tests are 19.69 MB / s, 20.32 MB / s, 19.87 MB / s, etc., the average storage bandwidth is calculated to be 20.05 MB / s. This value is used as an important basis for subsequent burn-in task scheduling.

[0036] Step S22: Calculate and obtain the cache capacity corresponding to the target device according to the page address mapping table corresponding to the target device; In the embodiment of the present invention, by retrieving the page address mapping table from the hardware information library of the target device, which is stored in the form of a binary file and contains detailed information such as the address mapping information of each physical page in the device storage medium, the page size, and whether it is a bad block. Assume that the target device uses a NAND Flash storage medium with a page size of 4KB and a total of 2048 pages. By traversing the page address mapping table and counting the number of valid pages, multiplying the number of valid pages by the page size can calculate the cache capacity corresponding to the target device. During the traversal process, a dedicated binary file parsing tool is used to read the content of the page address mapping table byte by byte. For each page of information read, check its "bad block flag" field. If the flag is "0", the page is considered a valid page and counted. After traversing the entire page address mapping table, assume that the number of valid pages counted is 2000 pages. By calculating "2000 pages × 4KB / page = 8MB", the cache capacity of the target device is obtained as 8MB. This cache capacity information will be used to evaluate the usage of the cache during the data burning process.

[0037] Step S23: Analyze the burning dependency of the data stream description file containing metadata, structure information, and burning instructions, so as to mine and analyze the dependency relationships corresponding to each data stream to be burned in the data stream description file based on the burning instructions, and generate the data burning dependency relationships between each data stream to be burned; In an embodiment of the present invention, by using a dedicated text parser to process a data stream description file containing metadata, structure information, and programming instructions, the parser reads the file line by line. First, it identifies the programming instruction fields in the file. The programming instructions are presented in a specific format, such as "CMD:WRITE,ADDR:0x1000,DATA:FILE1", which clearly indicates the write operation, the target address, and the data source. By analyzing the programming instructions, the dependency relationships between the data streams to be programmed are mined. If the data FILE2 in the instruction "CMD:WRITE,ADDR:0x2000,DATA:FILE2" depends on the data read by the instruction "CMD:READ,ADDR:0x1000,DATA:FILE1" for preprocessing, then it is determined that there is a data dependency relationship between FILE2 and FILE1; if the instruction "CMD:WRITE,ADDR:0x3000,DATA:FILE3" can only be performed after the instruction "CMD:WRITE,ADDR:0x2000,DATA:FILE2" is completed, then it is determined that there is a sequential dependency relationship between FILE3 and FILE2. In the above manner, all the programming instructions in the data stream description file are analyzed, and the identified dependency relationships are recorded in the form of a table. The table includes fields such as "source data stream", "target data stream", and "dependency type", and finally, the complete data programming dependency relationships between the data streams to be programmed are generated.

[0038] Step S24: Based on the data programming dependency relationships between the data streams to be programmed, a directed acyclic graph is constructed for the data streams to be programmed in the data stream description file to generate a directed acyclic graph of the data streams to be programmed; according to the directed acyclic graph of the data streams to be programmed, a topological programming sort is performed on the data streams to be programmed in the data stream description file to generate an initial programming sort sequence of the data streams. In the embodiment of the present invention, according to the data burning dependency generated in step S23, a directed acyclic graph of the data stream to be burned is constructed using a graph theory algorithm. Each data stream to be burned is used as a node in the graph, and the node attributes include information such as the data stream name, data volume size, and data type. According to the data burning dependency, a directed edge is established between nodes with a dependency relationship, and the direction of the edge points from the dependency source node to the dependency target node, and the dependency type is noted in the attributes of the edge. For example, if the data stream "FILE2" depends on "FILE1", a directed edge is drawn from node "FILE1" to node "FILE2", and the edge attribute is marked as "data dependency". During the construction process, a topological sorting algorithm is used to detect and eliminate possible circular dependencies to ensure that the generated graph is a directed acyclic graph. Suppose there is a circular dependency of "FILE4→FILE5→FILE4" in the initially constructed graph. By adjusting the dependency relationship, introducing virtual nodes, or re-planning the data stream processing order, the cycle is broken. After the directed acyclic graph is constructed, the topological sorting algorithm is used again to sort the nodes in the graph, and an initial burning sorting sequence of the data stream is generated according to the topological order of the nodes. During the sorting process, the nodes without predecessor nodes (i.e., not depending on other data streams) are preferentially arranged at the front of the sequence and arranged in turn backward to ensure that the burning order of the data streams in the sequence conforms to their dependency relationship.

[0039] Step S25: Based on the storage bandwidth and cache capacity corresponding to the target device, perform dynamic task scheduling for the initial burning sorting sequence of the data stream to generate a data stream burning task sequence.

[0040] In an embodiment of the present invention, by combining the previously obtained storage bandwidth of 20.05 MB / s and the calculated cache capacity of 8 MB, a burn-in dynamic task scheduling is performed on the initial burn-in sorting sequence of the data stream. The data in the sorting sequence is divided into task units with each 32 KB as a unit, and the theoretical transmission time of each task unit under the current storage bandwidth is calculated using the formula "transmission time = data volume / storage bandwidth". At the same time, considering the cache capacity of the target device, the usage of the cache during the data burn-in process is simulated. When the data volume of a certain task unit exceeds the remaining cache capacity, it is split into multiple subtasks to ensure that the data can be successfully written into the cache. For example, for a 64 KB data task unit with a remaining cache capacity of 48 KB, it is split into two subtasks of 48 KB and 16 KB. According to the transmission time of the task unit and the cache usage, the priority of each task unit is calculated. The priority calculation comprehensively considers factors such as the data volume size, the urgency of the dependency relationship, and the impact on the overall burn-in process. For example, a higher priority is given to the task unit related to the key system startup file; a lower priority is given to the task unit of the non-critical configuration file. The task units are rearranged in the order of priority to generate a data stream burn-in task sequence. During the burn-in process, the data is transmitted to the target device in the order of the task sequence through the DMA controller, and the storage bandwidth and cache usage are monitored in real time to dynamically adjust the task execution order and transmission rate to ensure that the data burn-in process is completed efficiently and stably.

[0041] Further, step S25 includes the following steps: Step S251: The initial burn-in sorting sequence of the data stream is divided into respective data segments, and the data volume size corresponding to each data segment is obtained. At the same time, based on the data transmission order and dependency relationship of each data segment in the initial burn-in sorting sequence of the data stream and in combination with the data volume size corresponding to each data segment, a data stream burn-in transmission topology network is constructed, where the data segment is used as a node, the data volume size is used as a node attribute, and the data transmission order and dependency relationship are used as edges; In an embodiment of the present invention, the initial burn-in sorting sequence of the data stream is divided into 2048 data segments according to a fixed-size rule, and the size of each data segment is set to 2 KB to ensure the integrity and independence of the data unit. The CRC32 checksum algorithm is used to calculate the checksum value for each data segment, and the checksum value is stored as part of the data segment for subsequent verification of the accuracy of data transmission. When constructing the data stream burn-in transmission topology network, a high-performance graph database is used as the storage and computing carrier. Each data segment is used as a node in the topology network, and the node attributes include key information such as the data volume size, checksum value, and creation timestamp. For example, the data volume size attribute of node D100 is 2048 bytes, the checksum value is 0x12345678, and the creation timestamp is 1678901234567. The edges between nodes are determined by the data transmission order and dependency relationship. If the burn-in of data segment D150 needs to be performed after D149 is completed, a directed edge is established from D149 to D150, and the attribute of the edge records the transmission order number as 149→150, and the dependency type is sequential dependency; if the burn-in of D160 depends on the completion of data preparation of D155 and D158, directed edges are established from D155 and D158 to D160 respectively, and the edge attribute indicates that the dependency type is data dependency. For a set of data segments with complex cyclic dependencies, virtual nodes are introduced for decoupling processing to ensure that the topology network is a directed acyclic graph structure. The finally constructed topology network contains 2048 nodes and 3000 directed edges, which are stored in a memory database with a capacity of 1 GB and support 150,000 graph queries and update operations per second.

[0042] Step S252: Calculate the betweenness centrality corresponding to each node through the data stream burn-in transmission topology network, and determine the burn-in transmission priority for each data segment based on the betweenness centrality corresponding to each node to generate a data stream burn-in transmission priority sequence; In the embodiment of the present invention, by using the Brandes algorithm to calculate the betweenness centrality of each node in the data stream burning transmission topology network, in order to improve the calculation efficiency, 12 CPU cores are allocated to process the calculation tasks in parallel. Each node maintains a dependency count table and a path contribution table during the calculation process. The dependency count table records the number of shortest paths passing through the node, and the path contribution table records the contribution degree of the node to each shortest path. After 40 rounds of iterative calculation, the betweenness centrality values of all nodes are obtained, and their range is between 0.0005 and 0.12. Based on the betweenness centrality values, the burning transmission priorities of data segments are determined. The radix sorting algorithm is used to sort the nodes. The time complexity of this algorithm is O(n), which can efficiently complete the sorting task of large-scale data. The sorting result is stored in a double-ended queue. The data segment corresponding to the node with a high betweenness centrality value is located at the head of the queue and is burned and transmitted first. For example, the data segment D200 with a betweenness centrality value of 0.12 is ranked first in the queue because it is in a key position in the data stream transmission path and has a greater impact on the overall burning process; while the data segment D1800 with a betweenness centrality value of 0.0005 is ranked at the rear of the queue. The finally generated data stream burning transmission priority sequence clearly ranks the data segments on the critical path (such as system core startup files, basic driver program data) at the front and the data segments on the non-critical path (such as user configuration files, log record data) at the rear, providing a clear priority guidance for subsequent burning and transmission.

[0043] Step S253: Based on the storage bandwidth corresponding to the target device, evaluate the storage bandwidth burning transmission pressure of the corresponding data segments in the data stream burning transmission priority sequence, so as to obtain the storage bandwidth burning transmission pressure corresponding to each data segment in the target device; In the embodiment of the present invention, by evaluating the bandwidth burning transmission pressure of data segments within the data stream burning transmission priority sequence according to the storage bandwidth characteristics of the target device, the target device adopts an SPI Flash storage interface, and its theoretical maximum bandwidth is 32 MB / s. Considering factors such as protocol overhead and bus competition in actual transmission, the actual available bandwidth is 26 MB / s. The data segments in the priority sequence are evaluated in groups of 32 each. The total data volume of each group of data segments is 64 KB. The sliding window algorithm is used to calculate the theoretical transmission time of each group of data segments. The calculation formula is: transmission time = data volume / storage bandwidth. For example, if the total data volume of a certain group of data segments is 64 KB, the theoretical transmission time is 64 KB / 26 MB / s ≈ 2.46 ms. However, in actual transmission, due to the existence of instruction transmission overhead, bus arbitration delay, etc., a loss coefficient of 1.3 is introduced, and the actual transmission time is about 3.2 ms. By evaluating all groups, the groups with a transmission time exceeding 5 ms are marked. For example, the data segments from group 40 to group 45 have an actual transmission time of 5.5 ms, which is determined as a high-pressure area, indicating that when these data segments are transmitted under the current bandwidth conditions, they will bring greater pressure to the storage interface and may affect the burning efficiency and stability.

[0044] Step S254: Based on the cache capacity corresponding to the target device, perform cache transmission saturation quantization on the corresponding data segments within the data stream burning transmission priority sequence to obtain the cache burning transmission saturation rate corresponding to each data segment in the target device. In the embodiment of the present invention, by performing cache transmission saturation quantization on the data segments within the data stream burning transmission priority sequence according to the cache capacity of the target device, the target device is equipped with a direct mapped cache with a capacity of 8 KB, and the cache block size is set to 32 bytes. According to the priority sequence order, the data segments are loaded into the cache in sequence, and the LRU (Least Recently Used) replacement strategy is used to manage the cache content to ensure that frequently used data is retained in the cache. By simulating the cache access process, the cache hit rate and miss rate of each data segment are accurately calculated. For example, the data segment D300 has a size of 2 KB and contains 64 cache blocks. It is accessed 120 times during the burning process, with 90 hits and 30 misses. The cache hit rate is 75%. For all data segments, calculate their cache occupancy rate. The formula is: cache occupancy rate = data segment size / cache capacity × cache hit rate. After calculation, it is found that the cache occupancy rate of data segments D500 to D550 exceeds 85%, which is marked as a high saturation area, meaning that these data segments occupy too much space in the cache, may cause frequent cache replacement operations, affect the data reading efficiency, and thus have an adverse impact on the burning process.

[0045] Step S255: Analyze the current load of the target device based on the storage bandwidth burning transmission pressure and cache burning transmission saturation rate corresponding to each data segment in the target device, so as to obtain the current burning load amount corresponding to the target device; dynamically adjust the burning transmission rate corresponding to each data segment according to the current burning load amount corresponding to the target device, and perform burning dynamic task scheduling on the initial burning sorting sequence of the data stream based on the burning transmission rate corresponding to each data segment to generate a data stream burning task sequence.

[0046] In the embodiment of the present invention, by comprehensively considering the storage bandwidth burning transmission pressure and cache burning transmission saturation rate corresponding to each data segment in the target device, the current load of the target device is analyzed, and a load calculation formula is defined: load amount = storage bandwidth pressure coefficient × 0.6 + cache saturation rate coefficient × 0.4, where both the storage bandwidth pressure coefficient and the cache saturation rate coefficient are normalized to between 0 and 1. For example, the storage bandwidth pressure of a certain data segment is 80%, and the normalized pressure coefficient is 0.8; the cache saturation rate is 90%, and the normalized saturation rate coefficient is 0.9, then its load amount = 0.8 × 0.6 + 0.9 × 0.4 = 0.84. Based on the calculated load amount, the PID controller is used to dynamically adjust the burning transmission rate corresponding to each data segment. A load amount threshold is set. When the load amount exceeds 0.8, the transmission rate is reduced by 15%; when the load amount is lower than 0.4, the transmission rate is increased by 8%. The initial transmission rate is set to 22MB / s. After multiple iterative adjustments, for the high-load data segment D600, according to its load amount of 0.88, the transmission rate is reduced from 22MB / s to 18.7MB / s; for the low-load data segment D700, according to the load amount of 0.35, the transmission rate is increased to 23.8MB / s, and finally it stabilizes at a suitable transmission rate of 19MB / s. According to the adjusted transmission rate, the initial burning sorting sequence of the data stream is rearranged, and the data segments in the high-load area are scattered and inserted into the low-load time period for transmission to generate the final data stream burning task sequence. This sequence performs data transmission through the DMA (Direct Memory Access) controller, effectively reducing the CPU load, realizing the efficient and stable operation of the data burning process, and ensuring that the data burning task can be completed quickly and accurately under the resource constraints of the target device.

[0047] Further, step S3 includes the following steps: Step S31: Perform low-density redundant coding and block division on the data stream to be burned corresponding to the data stream burning task sequence to generate fault-tolerant coded data stream blocks; In the embodiment of the present invention, the data stream to be burned in the data stream burning task sequence is block-processed by combining a fixed size with a low redundancy encoding method. Each data block is set to be 32KB in size. A dedicated data block device is used to sequentially read data from the starting position of the task sequence. Every time 32KB of data is read, an independent block is generated and assigned a unique number. For example, for a data stream to be burned of 512KB, 16 blocks can be obtained, numbered from BLOCK_001 to BLOCK_016. After block division, low-density redundant encoding is performed on each block. The parity check encoding algorithm is adopted. This algorithm generates check bits by calculating the parity of the binary bits in the data block. Taking the BLOCK_001 block as an example, it contains 32KB (i.e., 32×1024 = 32768 bytes) of data. Every 8 bits of data are taken as a group, and the number of "1"s in each group is calculated. If the number of "1"s is odd, for odd parity, the check bit is set to 0; for even parity, the check bit is set to 1; vice versa. The generated check bits are appended to the end of the block to form a fault-tolerant encoded data stream block. After this processing, each block not only has the basic error detection ability but also reduces the additional overhead of data storage and transmission due to low redundancy encoding, preparing for subsequent burning operations.

[0048] Step S32: Monitor the storage temperature corresponding to the target device in real time; In the embodiment of the present invention, the storage temperature of the target device is monitored in real time through a built-in temperature sensor. The DS18B20 model is selected for the temperature sensor, with a measurement range of -55°C to +125°C and an accuracy of up to ±0.5°C. The sensor collects temperature data at intervals of every 100 milliseconds and transmits the data to the device monitoring module through a single-wire protocol. The monitoring module processes the received temperature data. First, the data is filtered to remove abnormal values caused by factors such as electromagnetic interference. The median filtering algorithm is adopted. Every time 5 consecutive temperature data are collected, they are sorted from small to large, and the middle value is taken as the valid temperature data. For example, if the 5 consecutive data collected are 25.3°C, 25.1°C, 24.9°C, 25.5°C, 25.2°C, the middle value 25.2°C after sorting is the valid temperature data. The processed data is stored in the cache of the monitoring module and updated in real time, providing an accurate basis for subsequent adjustment of the burning strategy based on temperature data.

[0049] Step S33: Monitor the voltage fluctuation corresponding to the target device in real time; In an embodiment of the present invention, by using a high-precision voltage monitoring chip to monitor the voltage fluctuation of the target device in real time, a chip of model AD7865 is selected. Its resolution is 16 bits and the sampling rate can reach 100 kSPS. It can accurately capture the minute changes in voltage. The chip is connected to the power line of the target device through an analog front-end circuit to collect voltage signals in real time. The collected analog voltage signals are converted by A / D inside the chip and become digital signals, which are then transmitted to the data processing unit. The data processing unit analyzes the digital signals and calculates parameters such as the average value, maximum value, minimum value, and fluctuation range of the voltage. For example, 100 voltage data are collected within 1 second, and the calculated average value is 3.28 V, the maximum value is 3.32 V, the minimum value is 3.25 V, and the fluctuation range is 0.07 V. At the same time, the normal voltage range threshold is set to 3.3 V ± 0.1 V. When the monitored voltage exceeds this range, the warning mechanism is triggered, and the abnormal information is marked and stored to adjust the burning parameters in combination with the voltage fluctuation during the burning process.

[0050] Step S34: Monitor the corresponding number of erase / write cycles of the target device in real time; In an embodiment of the present invention, the number of erase / write cycles is monitored in real time by reading the page address mapping table of the storage medium of the target device. The storage medium uses NAND Flash, and its page address mapping table records the usage status information of each storage page, including the number of erase / write cycles, whether it is a bad block, etc. A dedicated storage medium management module reads the page address mapping table every 500 milliseconds. The module summarizes and statistically analyzes the read number of erase / write cycle data, calculates the average number of erase / write cycles of the entire storage medium and the distribution of pages with a higher number of erase / write cycles. For example, in a certain read, it is found that there are 2048 storage pages in the storage medium, and the total number of erase / write cycles is 102,400 times. Then the average number of erase / write cycles is 102,400 ÷ 2048 = 50 times. At the same time, it is found that the number of erase / write cycles of 10 pages numbered from PAGE_001 to PAGE_010 exceeds 80 times. These information are sorted out and recorded. By continuously monitoring the number of erase / write cycles, the wear degree of the storage medium can be understood, providing key reference data for the burning operation.

[0051] Step S35: Use the storage temperature, voltage fluctuation, and number of erase / write cycles corresponding to the target device as device status data, and perform a burning write process on the error-correcting coded data stream blocks based on the device status data to generate device burning data stream blocks.

[0052] In an embodiment of the present invention, by integrating the previously acquired storage temperature, voltage fluctuation data, and number of erase-write cycles into device status data, a model is trained using the Bayesian network learning algorithm. The block feature parameters of the burn-in data stream and device parameters in the data are used as network nodes, and the burn-in result is used as the output node. The network structure is set to three layers. The input layer contains 10 nodes corresponding to features such as data size and encoding type. The hidden layer contains 8 nodes, and the output layer is 1 node representing the write error probability. The conditional probability table between nodes is calculated using the maximum likelihood estimation method. For example, when the data block size is 16 KB, BCH encoding is used, and the number of erase-write cycles of the target device's storage page exceeds 50 times, the maximum likelihood estimation method is used to calculate the conditional probability between nodes based on the device's historical burn-in status data. For example, assuming that in the historical data, when the data block size is in the range of 16 KB - 32 KB, BCH encoding is used, and the number of erase-write cycles of the target device's storage page is in the range of 50 - 100 times, a total of 100 burn-in operations are performed, and among them, 12 are write failures. Then, under this condition combination, the conditional probability of write failure is 12÷100 = 0.12. This probability relationship is stored in the conditional probability table to obtain a Bayesian network model that can predict the write error probability of the target device for data stream blocks. Secondly, the status data of the current target device is acquired, including the device's real-time temperature (measured by a built-in temperature sensor with an accuracy of ±0.5 °C), the current number of erase-write cycles of the storage medium (obtained by reading the counter in the page address mapping table), the remaining available storage space (calculated based on the total capacity of the storage medium and the used space), and the detailed parameters of the fault-tolerant encoded data stream block (such as block number, data content summary, encoding type, number of parity bits). The above data is organized into a vector that conforms to the input format of the Bayesian network model and input into the trained model. The model performs reasoning and calculation based on the input data according to the established conditional probability relationship. For example, when the data size is 16 KB and BCH encoding is used, it corresponds to a probability value. Combining with the number of erase-write cycles of 60 times of the storage page, it corresponds to an adjusted probability. Considering the influence of the real-time temperature of 28 °C, etc., after a series of probability operations, for example, finally, the write error probability of this block is obtained as 0.15. The write error probability of each fault-tolerant encoded data stream block is calculated in turn to form a list containing the block number and the corresponding error probability. Then, a high error probability threshold of 0.1 and a low error probability threshold of 0.03 are set. The previously obtained write error probability list is traversed, and each fault-tolerant encoded data stream block is processed. For the BLOCK_001 block with a write error probability of 0.15, it is determined as a high error probability block, and its corresponding standard burn-in voltage is increased from 3.3 V to 3.3 V×(1 + 5%) = 3.465 V, and the standard clock frequency is decreased from 50 MHz to 50 MHz×(1 - 10%) = 45 MHz. For the write error probability of 0.For the BLOCK_010 block of 02, which is determined to be a block with a low error probability, the programming voltage is increased to 3.465V×(1 + 3%) = 3.569V (the increase is less than 5% of that of the block with a high error probability), and the clock frequency is increased to 45MHz×(1 + 8%) = 48.6MHz (the increase is less than 10% of that of the block with a high error probability). According to the adjusted programming voltage and clock frequency, the dedicated programming controller sequentially performs the write operation on all the fault-tolerant encoded data stream blocks, writes the data into the target device storage medium, and finally generates the device programming data stream blocks.

[0053] Further, step S31 includes the following steps: Perform block processing on the to-be-programmed data stream corresponding to the data stream programming task sequence to obtain each to-be-programmed data stream block; In the embodiment of the present invention, the to-be-programmed data stream in the data stream programming task sequence is block-processed by adopting a fixed-size block strategy. The size of each data block is set to 16KB, and the operation is performed by a dedicated data block processing device. The device starts from the head of the data stream programming task sequence, reads the data in sequence, and generates an independent data block every time 16KB of data is read. At the same time, a unique identification number is assigned to each block. For example, for a to-be-programmed data stream with a size of 256KB, after block processing, 16 data blocks can be obtained, numbered from BLOCK_001 to BLOCK_016 in sequence. During the block processing, the device performs boundary check on each block to ensure the integrity of the data. If the amount of data at the end of the data stream is less than 16KB, the remaining data is taken as a separate block and filled with specific padding bytes (such as 0xFF) afterwards to make the block size reach the standard. After the block processing is completed, all the blocks are stored in the buffer area in the order of the original data stream to form each to-be-programmed data stream block, providing a structured data unit for subsequent processing.

[0054] Preferably, obtain the to-be-programmed data content corresponding to each block through each to-be-programmed data stream block, and perform content-related grouping on each to-be-programmed data stream block based on the to-be-programmed data content corresponding to each block to obtain the key to-be-programmed data blocks and the non-key to-be-programmed data blocks; In the embodiments of the present invention, by using a content analysis tool to parse the content of each data stream block to be burned, the content of the data to be burned corresponding to each block is obtained. This tool can identify the key fields, data types, and the association relationships between data. Taking a data stream containing system configuration information and log records as an example, the data in a certain block includes key information such as system startup parameters and network configuration, while another block is the recent operation log record. Through the preset content classification rules, the block data is analyzed. The data blocks containing data crucial for system operation, such as system core parameters and driver program codes, are classified as key data blocks to be burned; the data blocks containing user operation records and temporary cache data, etc., which have less impact on the core functions of the system, are classified as non - key data blocks to be burned. For example, after analysis, among 16 blocks, 6 blocks such as BLOCK_001, BLOCK_003, and BLOCK_005 are determined as key data blocks to be burned, and the remaining 10 blocks are non - key data blocks to be burned. The classification results are stored in a table form, and the table includes fields such as block number, data content summary, and category, which is convenient for subsequent processing.

[0055] Preferably, the content correlation coefficient and the number of erase - write times of the corresponding data blocks are obtained through the key data blocks to be burned and the non - key data blocks to be burned. Based on the content correlation coefficient and the number of erase - write times, the optimal number of parity bits of the corresponding data blocks is calculated. At the same time, according to the optimal number of parity bits of the corresponding data blocks, high - redundancy coding is adopted for the key data blocks to be burned, and low - redundancy coding is adopted for the non - key data blocks to be burned, so as to generate key encoded data blocks to be burned and non - key encoded data blocks to be burned; In an embodiment of the present invention, the content correlation coefficient and the number of erase-write cycles of each data block are calculated by using a dedicated data analysis algorithm. The content correlation coefficient is obtained by calculating the similarity between the block data and the standard key data template. The cosine similarity algorithm is adopted to convert the data into a vector form for calculation, and the value range is from 0 to 1. The closer the value is to 1, the higher the correlation. The number of erase-write cycles is statistically obtained according to the page address mapping table and the historical operation records of the target device. Each storage page has a corresponding erase-write cycle counter. For example, the content correlation coefficient between the key data block to be burned, BLOCK_001, and the standard key data template is 0.95, and the number of erase-write cycles of the storage page where it is located is 50 times; the content correlation coefficient of the non-key data block to be burned, BLOCK_010, is 0.3, and the number of erase-write cycles is 10 times. According to the pre-established mathematical model, the optimal number of parity bits is calculated by combining the content correlation coefficient and the number of erase-write cycles. The model considers factors such as data importance and storage reliability. For the key data block to be burned, since its content correlation coefficient is high and the number of erase-write cycles may affect data stability, the calculated optimal number of parity bits is relatively large; for the non-key data block to be burned, it is relatively small. According to the optimal number of parity bits, a high-redundancy error correction coding algorithm, such as BCH coding, is adopted for the key data block to be burned, and more parity bits are added to improve the error correction ability of the data; a low-redundancy coding algorithm, such as parity check code, is adopted for the non-key data block to be burned to reduce the coding overhead while ensuring basic data verification. After coding, finally, the key data block to be burned and the non-key data block to be burned are generated, effectively improving the reliability and efficiency of data storage.

[0056] Preferably, index tags corresponding to the position information and the redundant information are added to each key data block to be burned and the non-key data block to be burned to generate a fault-tolerant coded data stream block.

[0057] In an embodiment of the present invention, an index tag generation tool is used to add index tags to critical to-be-burned coding blocks and non-critical to-be-burned coding blocks. The index tag contains two parts of content: location information and redundant information. The location information records the offset address and storage order of the block in the original data stream, accurate to the byte level, so as to quickly locate during the burning and reading processes. The redundant information includes the coding type of the block, the number of parity bits, and the association relationship with other blocks, etc., which are used for data verification and recovery. For example, in the index tag generated for the critical to-be-burned coding block BLOCK_001, the location information shows that its offset address in the original data stream is 0 and the storage order is the 1st; the redundant information indicates that BCH coding is adopted, the number of parity bits is 16 bits, and there is a data association with BLOCK_003 and BLOCK_005. In the index tag of the non-critical to-be-burned coding block BLOCK_010, the location information is an offset address of 96KB and the storage order is the 7th, and the redundant information shows that a parity check code is adopted, the number of parity bits is 1 bit, and there is no special association relationship. The index tag is added to the head of the block data in a specific format, and each tag occupies a fixed number of bytes (such as 64 bytes). After the addition is completed, a fault-tolerant coding data stream block containing complete index tags is generated, and these blocks will be stored and transmitted in a new order. During the data burning process, it can effectively improve the fault tolerance and recovery efficiency of the data, and ensure the integrity and accuracy of the data.

[0058] Further, the process of burning and writing the fault-tolerant coding data stream block based on the device state data in step S35 includes the following steps: Obtain the device historical burning state data, and train a Bayesian network model based on the device historical burning state data to predict the writing error probability corresponding to the data stream block of the target device; In an embodiment of the present invention, by retrieving device historical programming status data from a device historical data repository, which adopts a distributed file system architecture and stores detailed data of 10,000 programming tasks in the past three years. Each set of data includes programming time, characteristic parameters of programming data stream blocks (such as data size, coding type, number of parity bits), target device model, storage medium parameters (such as number of erase / write cycles, page address mapping table status), and programming result (write success or failure and error type). The Bayesian network learning algorithm is used to train the model. The characteristic parameters of the programming data stream blocks and device parameters in the data are used as network nodes, and the programming result is used as the output node. The network structure is set to three layers. The input layer includes 10 nodes corresponding to characteristics such as data size and coding type, the hidden layer includes 8 nodes, and the output layer is 1 node representing the write error probability. The conditional probability table between nodes is calculated using the maximum likelihood estimation method. For example, when the data block size is 16KB, BCH coding is used, and the number of erase / write cycles of the target device storage page exceeds 50 times, the maximum likelihood estimation method is used to calculate the conditional probability between nodes based on the device historical programming status data. For example, assuming that in the historical data, when the data block size is in the range of 16KB - 32KB, BCH coding is used, and the number of erase / write cycles of the target device storage page is in the range of 50 - 100 times, a total of 100 programming operations are performed, and 12 of them are write failures. Then, under this combination of conditions, the conditional probability of write failure is 12÷100 = 0.12. This probability relationship is stored in the conditional probability table. For each different state combination of the input nodes and the state of the output node, the corresponding conditional probability is calculated and stored in this way, and this probability relationship is stored in the conditional probability table. After 500 iterations of training, the prediction accuracy of the model on the test set reaches 85%, and a Bayesian network model capable of predicting the write error probability of the target device for data stream blocks is obtained.

[0059] Preferably, by inputting the currently corresponding device status data into the Bayesian network model to perform write prediction analysis on the fault-tolerant encoded data stream blocks for the target device, the write error probability corresponding to each fault-tolerant encoded data stream block can be obtained; In an embodiment of the present invention, by obtaining the status data of the current target device, including the real-time device temperature (measured by a built-in temperature sensor with an accuracy of ±0.5°C), the current number of erasures and writes of the storage medium (obtained by reading the counter in the page address mapping table), the remaining available storage space (calculated based on the total capacity of the storage medium and the used space), and the detailed parameters of the error-tolerant encoded data stream blocks (such as block number, data content summary, encoding type, number of parity bits), the above data is organized into a vector that conforms to the input format of the Bayesian network model and input into the trained model. The model performs inference calculations based on the input data according to the established conditional probability relationships. Taking the error-tolerant encoded data stream block numbered BLOCK_001 as an example, the size of this block of data is 16KB, using BCH encoding, with 16 parity bits, the current number of erasures and writes of the storage page of the target device is 60 times, and the real-time temperature is 28°C. The model will comprehensively consider the probability relationships among the states of multiple nodes such as data size, encoding type, number of erasures and writes, and real-time temperature, traverse the conditional probability table, and perform combined operations on the relevant probabilities. For example, when the data size is 16KB and BCH encoding is used, there is a corresponding probability value. Combining the number of erasures and writes of the storage page 60 times, there is another adjusted probability. Considering the influence of the real-time temperature of 28°C, etc., after a series of probability operations, for example, finally, the write error probability of this block is obtained as 0.15; the calculations are sequentially performed on all 100 error-tolerant encoded data stream blocks to obtain the write error probability corresponding to each block, forming a list containing the block number and the corresponding error probability, providing a basis for subsequent adjustment of the burning strategy.

[0060] Preferably, based on the write error probabilities corresponding to the respective error-tolerant encoded data stream blocks, the corresponding error-tolerant encoded data stream blocks are subjected to a burning and writing process. If it is determined to be a high error probability block according to the write error probability, the burning voltage is increased by less than or equal to 5% of the standard burning voltage, and the clock frequency is decreased by less than or equal to 10% of the standard clock frequency; if it is determined to be a low error probability block according to the write error probability, the burning voltage is increased by less than 5% of the high error probability block, and the clock frequency is increased by less than 10% of the high error probability block, thereby performing the write operation for each error-tolerant encoded data stream block to generate the device burning data stream block.

[0061] In an embodiment of the present invention, by setting the high error probability threshold to 0.1 and the low error probability threshold to 0.03, traversing the previously obtained write error probability list, and processing each fault-tolerant encoded data stream block. For the BLOCK_001 block with a write error probability of 0.15, it is determined as a high error probability block, and its corresponding standard programming voltage is increased from 3.3V to 3.3V×(1 + 5%) = 3.465V, and precise adjustment is achieved through a voltage regulation module with an adjustment accuracy of ±0.01V; the standard clock frequency is decreased from 50MHz to 50MHz×(1 - 10%) = 45MHz, and the clock signal frequency is adjusted using a clock generator with an adjustment accuracy of ±1MHz. For the BLOCK_010 block with a write error probability of 0.02, it is determined as a low error probability block, and the programming voltage is increased to 3.465V×(1 + 3%) = 3.569V (the increase amplitude is less than 5% of the high error probability block), and the clock frequency is increased to 45MHz×(1 + 8%) = 48.6MHz (the increase amplitude is less than 10% of the high error probability block). According to the adjusted programming voltage and clock frequency, the dedicated programming controller sequentially performs write operations on all fault-tolerant encoded data stream blocks, writes the data into the target device storage medium, and finally generates device programming data stream blocks to complete the data programming process, while ensuring the data write success rate and improving the programming efficiency.

[0062] Further, step S4 includes the following steps: Step S41: Perform block-by-block CRC check on the device programming data stream blocks to obtain device programming check success blocks and device programming check failure blocks; In an embodiment of the present invention, a dedicated CRC check module is used to perform block-by-block check on the device programming data stream blocks. This module incorporates the CRC-32 algorithm with a generating polynomial of 0x04C11DB7. The check module sequentially reads each device programming data stream block from the device storage medium, reading 64KB of data each time (consistent with the block size). Taking the block numbered BLOCK_001 as an example, the check module first calculates the CRC value of this block. The module inputs the block data bit by bit into the CRC calculation circuit, and the circuit performs shift and exclusive OR operations according to the CRC-32 algorithm, and finally outputs a 32-bit CRC value. After the calculation is completed, the obtained CRC value is compared with the original CRC value recorded during the programming process. If the two are consistent, it is determined that the block check is successful and marked as a device programming check success block; if not, it is determined that the check fails and marked as a device programming check failure block. This operation is sequentially performed on all blocks, and finally the check results are sorted into a list to clearly distinguish the blocks with successful and failed checks, providing a basis for subsequent processing.

[0063] Step S42: Obtain the corresponding block address and error type for the failed block in device programming verification, and locate the error block corresponding to the failed block in device programming verification based on the block address and error type; In the embodiment of the present invention, for the failed block in device programming verification, the address parsing module is used to obtain its corresponding block address. This module directly accesses the page address mapping table of the storage medium and extracts the start address and end address of the failed block in physical storage. For example, for the failed BLOCK_001 block, the address parsing module determines that its physical storage address range is from 0x00000000 to 0x0000FFFF. At the same time, the error type identification module is used to analyze the error characteristics of the failed block. This module determines the error type by checking information such as the integrity, continuity of the block data, and the status of the redundancy check bits. If it is found that there are consecutive multiple-bit errors in the data and the check bits do not match the data, it is determined as a burst error; if there are only individual-bit errors, it is determined as a random error. Combining the block address and the error type, the specific error block is accurately located. For example, it is determined that there is a burst error in the area from 32KB to 32.5KB of the BLOCK_001 block, providing a clear target for subsequent repair work.

[0064] Step S43: Obtain the index label of the error block in the error-corrected coding data stream block based on the block address, and obtain the corresponding redundant information in the error-corrected coding data stream block based on the index label. At the same time, use the corresponding redundant information in the error-corrected coding data stream block to perform online repair on the error block, so as to use the redundant information to repair the error block by using an iterative decoding algorithm. If the repair fails, trigger the re-programming process; otherwise, perform iterative CRC verification until the verification is successful; In the embodiment of the present invention, based on the previously obtained block address, the index label retrieval module is used to search for the corresponding index label in the error-corrected coding data stream block. Each error-corrected coding data stream block adds an index label containing location information and redundant information during generation, which is stored in a special index table. For example, for the error block BLOCK_001, the corresponding index label is found in the index table through its physical address, and the redundant information of this block is obtained from it, including the coding type used (such as BCH coding), the number of check bits, and the association relationship with other blocks. Using the obtained redundant information, an iterative decoding algorithm is used to perform online repair on the error block. Taking BCH coding as an example, the decoding algorithm first constructs a check equation based on the check bits and the original data, and then gradually corrects the error bits through iterative calculation. After each iteration, the CRC value is recalculated and compared with the original value. If they are the same, the repair is successful; if they are still different after 10 iterations, it is determined that the repair fails, and the re-programming process is triggered. If the repair is successful, the repaired block is subjected to CRC verification again, and the above process is repeated until the verification is successful to ensure the accuracy of the data.

[0065] Step S44: Based on the device status data, perform a burn verification analysis on the device burn verification successful blocks corresponding to the successful CRC verification, calculate the bit error rate and remaining erase / write life corresponding to each burn block according to the device status data, and verify the bit error rate and remaining erase / write life corresponding to each burn block according to the preset threshold. If the verification fails, repair and verify again to generate a data stream burn verification completion result.

[0066] In the embodiment of the present invention, by inputting the storage temperature, voltage fluctuation, and number of erase / write cycles in the device status data into the bit error rate calculation module, this module calculates the bit error rate of each burn block based on a preset mathematical model in combination with the device status data. For example, for the device burn verification successful block BLOCK_001, if the current storage temperature is 28°C, the voltage fluctuation range is 0.12V, and the number of erase / write cycles is 60 times, the model calculates that its bit error rate is 0.001%. At the same time, the erase / write life evaluation module calculates the remaining erase / write life of the storage medium according to the number of erase / write cycle data. Assuming that the total erase / write life of the storage medium is 1000 times and the current average number of erase / write cycles is 60 times, the remaining erase / write life is (1000 - 60) ÷ 1000 × 100% = 94%. Compare the calculated bit error rate and remaining erase / write life with the preset thresholds (the bit error rate threshold is 0.005% and the remaining erase / write life threshold is 80%). If the bit error rate and remaining erase / write life of BLOCK_001 both meet the threshold requirements, the verification passes; if any one does not meet, repair and verify again until all blocks pass the verification, and finally generate a data stream burn verification completion result.

[0067] Furthermore, the present invention also provides an intelligent data burn system based on a data stream for executing the intelligent data burn method based on a data stream as described above. The intelligent data burn system based on a data stream includes: A to-be-burned data stream description output module, configured to collect to-be-burned input source data in real time through a bus interface, perform format recognition and semantic parsing on the to-be-burned input source data to generate metadata information corresponding to the to-be-burned data; obtain the hardware storage structure information and burn protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the to-be-burned data based on the hardware storage structure information and burn protocol specification corresponding to the target device, thereby generating a data stream description file including metadata, structure information, and burn instructions; A burn dynamic scheduling module, configured to obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burn dynamic task scheduling on the data stream description file including metadata, structure information, and burn instructions based on the storage bandwidth and cache capacity corresponding to the target device, thereby generating a data stream burn task sequence; The data stream block burning and writing module is used to perform low-density redundant coding and block division on the to-be-burned data stream corresponding to the data stream burning task sequence to generate fault-tolerant coded data stream blocks; by monitoring in real time the device status data of the target device corresponding to the storage temperature, voltage fluctuation and erasing / writing times, and performing burning and writing processing on the fault-tolerant coded data stream blocks based on the device status data to generate device-burned data stream blocks; The burning error online repair module is used to perform block-by-block CRC check on the device-burned data stream blocks, locate the error blocks corresponding to the device-burned data stream blocks, and at the same time perform online repair verification and analysis on the error blocks by using the corresponding redundant information in the fault-tolerant coded data stream blocks and combining with the device status data to generate the result of completed data stream burning verification.

[0068] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An intelligent data burning method based on data flow, characterized in that Including the following steps: Step S1: Real-time collect the input source data to be burned through the bus interface, and perform format recognition and semantic parsing on the input source data to be burned, so as to generate metadata information corresponding to the data to be burned; Obtain the hardware storage structure information and the burning protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the data to be burned based on the hardware storage structure information and the burning protocol specification corresponding to the target device, so as to generate a data stream description file containing metadata, structure information, and burning instructions; Step S2: Obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burning dynamic task scheduling on the data stream description file containing metadata, structure information, and burning instructions based on the storage bandwidth and cache capacity corresponding to the target device, so as to generate a data stream burning task sequence; Step S3: Perform low-density redundant coding and chunking on the data stream to be burned in the data stream burning task sequence, so as to generate error-tolerant coded data stream chunks; Real-time monitor the device status data corresponding to the target device, including storage temperature, voltage fluctuation, and erase-write times, and perform burning and writing processing on the error-tolerant coded data stream chunks based on the device status data, so as to generate device burning data stream chunks; Step S4: Perform block-by-block CRC check on the device burning data stream chunks, locate the error chunks corresponding to the device burning data stream chunks, and at the same time use the redundant information corresponding in the error-tolerant coded data stream chunks and combine the device status data to perform online repair verification analysis on the error chunks, so as to generate a data stream burning verification completion result.

2. The intelligent data burning method based on data flow according to claim 1, wherein Step S1 includes the following steps: Step S11: Configure the SPI, I2C, and UART interface protocols through the bus interface to collect the corresponding input source data to be burned in real time; Step S12: Perform format recognition and parsing on the input source data to be burned, so as to design a multi-mode data parser to support automatic recognition and parsing of various data formats including binary, text, XML, and JSON, and obtain the input format parsing data to be burned; Step S13: Perform data field semantic parsing on the input format parsing data to be burned, so as to generate metadata information corresponding to the data to be burned; Step S14: Obtain the hardware storage structure information and the burning protocol specification corresponding to the target device, where the hardware storage structure information includes flash block distribution, page address mapping table, and bad block marking information; Step S15: Perform data stream description output on the metadata information corresponding to the data to be burned based on the hardware storage structure information and the burning protocol specification corresponding to the target device, so as to generate a data stream description file containing metadata, structure information, and burning instructions.

3. The intelligent data burning method based on data flow according to claim 2, wherein Step S15 includes the following steps: Step S151: Obtain the corresponding semantic entities of each data to be burned through the metadata information corresponding to the data to be burned, and perform entity relationship extraction between the semantic entities of each data to be burned, so as to obtain the data entity relationship between the data entities of each data to be burned; Step S152: Based on the data entity relationship between the data entities of each data to be burned, construct an entity relationship graph for the semantic entities of each data to be burned, so as to generate an entity relationship graph of the data to be burned; Step S153: Generate executable burn instructions for the entity relationship diagram of the data to be burned based on the burn protocol specification corresponding to the target device, so as to generate executable burn instructions for the target device corresponding to the data to be burned; Step S154: Output a data stream description of the metadata information corresponding to the data to be burned based on the hardware storage structure information corresponding to the target device and in combination with the executable burn instructions for the target device corresponding to the data to be burned, and generate a data stream description file including metadata, structure information, and burn instructions.

4. The intelligent data burning method based on data flow according to claim 3, characterized in that Step S2 includes the following steps: Step S21: Obtain the storage bandwidth corresponding to the target device through continuous write testing; Step S22: Calculate and obtain the cache capacity corresponding to the target device according to the page address mapping table corresponding to the target device; Step S23: Analyze the burn dependency relationship of the data stream description file including metadata, structure information, and burn instructions, so as to mine and analyze the dependency relationship of each data stream to be burned in the data stream description file based on the burn instructions, and generate the data burn dependency relationship between each data stream to be burned; Step S24: Perform a directed acyclic construction on each data stream to be burned in the data stream description file based on the data burn dependency relationship between each data stream to be burned, so as to generate a directed acyclic graph of the data stream to be burned; perform a topological burn sorting on each data stream to be burned in the data stream description file according to the directed acyclic graph of the data stream to be burned, so as to generate an initial burn sorting sequence of the data stream; Step S25: Perform a burn dynamic task scheduling on the initial burn sorting sequence of the data stream based on the storage bandwidth and cache capacity corresponding to the target device, and generate a data stream burn task sequence.

5. The intelligent data burning method based on data stream according to claim 4, wherein Step S25 includes the following steps: Step S251: Divide the initial burn sorting sequence of the data stream into each data segment, obtain the corresponding data volume size through each data segment, and at the same time construct a data stream burn transmission topology network based on the data transmission order and dependency relationship of each data segment in the initial burn sorting sequence of the data stream and in combination with the corresponding data volume size of each data segment, where the data segment is used as a node, the data volume size is used as a node attribute, and the data transmission order and dependency relationship are used as edges; Step S252: Calculate the betweenness centrality of each node through the data stream burn transmission topology network, and determine the burn transmission priority of each data segment based on the betweenness centrality of each node, so as to generate a data stream burn transmission priority sequence; Step S253: Evaluate the bandwidth burn transmission pressure of the data segments corresponding to the data stream burn transmission priority sequence based on the storage bandwidth corresponding to the target device, so as to obtain the storage bandwidth burn transmission pressure of each data segment in the target device; Step S254: Quantify the cache transmission saturation of the data segments corresponding to the data stream burn transmission priority sequence based on the cache capacity corresponding to the target device, so as to obtain the cache burn transmission saturation rate of each data segment in the target device; Step S255: Analyze the current load of the target device based on the storage bandwidth burning transmission pressure and cache burning transmission saturation rate corresponding to each data segment in the target device to obtain the current burning load of the target device; Dynamically adjust the burning transmission rate corresponding to each data segment according to the current burning load of the target device, and perform dynamic task scheduling for the initial burning sorting sequence of the data stream based on the burning transmission rate corresponding to each data segment to generate a data stream burning task sequence.

6. The intelligent data burning method based on data stream according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Perform low-density redundant coding and chunking on the data stream to be burned in the data stream burning task sequence to generate fault-tolerant coded data stream chunks; Step S32: Monitor the storage temperature corresponding to the target device in real time; Step S33: Monitor the voltage fluctuation corresponding to the target device in real time; Step S34: Monitor the number of erase / write cycles corresponding to the target device in real time; Step S35: Use the storage temperature, voltage fluctuation, and number of erase / write cycles corresponding to the target device as device status data, and perform burning and writing processing on the fault-tolerant coded data stream chunks based on the device status data to generate device burning data stream chunks.

7. The intelligent data burning method based on data stream according to claim 6, wherein Step S31 includes the following steps: Perform chunking on the data stream to be burned in the data stream burning task sequence to obtain each data stream chunk to be burned; Obtain the content to be burned corresponding to each chunk through each data stream chunk to be burned, and perform content-related grouping on each data stream chunk to be burned based on the content to be burned corresponding to each chunk to obtain key data stream chunks to be burned and non-key data stream chunks to be burned; Obtain the content correlation coefficient and the number of erase / write cycles of the corresponding data chunks through the key data stream chunks to be burned and the non-key data stream chunks to be burned, calculate the optimal number of parity bits of the corresponding data chunks based on the content correlation coefficient and the number of erase / write cycles, and at the same time use high redundancy coding for the key data stream chunks to be burned and low redundancy coding for the non-key data stream chunks to be burned to generate key coded data stream chunks to be burned and non-key coded data stream chunks to be burned; Generate fault-tolerant coded data stream chunks by adding index tags containing position information and redundancy information to each key coded data stream chunk to be burned and non-key coded data stream chunk to be burned.

8. The intelligent data burning method based on data flow according to claim 6, wherein The burning and writing processing of the fault-tolerant coded data stream chunks based on the device status data in Step S35 includes the following steps: Obtain the device historical burning status data, and train a Bayesian network model based on the device historical burning status data to predict the writing error probability of the target device corresponding to the data stream chunks; Perform writing prediction analysis of the target device on the fault-tolerant coded data stream chunks by inputting the current device status data into the Bayesian network model to obtain the writing error probability of each fault-tolerant coded data stream chunk. Based on the write error probability corresponding to each error-tolerant encoded data stream block, perform a burn-in write process on the corresponding error-tolerant encoded data stream block. If it is determined to be a high error probability block according to the write error probability, increase the burn-in voltage, with the increase amplitude less than or equal to 5% of the standard burn-in voltage, and decrease the clock frequency, with the decrease amplitude less than or equal to 10% of the standard clock frequency. If it is determined to be a low error probability block according to the write error probability, increase the burn-in voltage, with the increase amplitude less than 5% of the high error probability block, and increase the clock frequency, with the increase amplitude less than 10% of the high error probability block, so as to perform the write operation of each error-tolerant encoded data stream block to generate the device burn-in data stream block.

9. The intelligent data burning method based on data flow according to claim 1, wherein Step S4 includes the following steps: Step S41: Perform block-by-block CRC check on the device burn-in data stream block to obtain the device burn-in check success blocks and the device burn-in check failure blocks; Step S42: Obtain the corresponding block address and error type through the device burn-in check failure blocks, and locate the error blocks corresponding to the device burn-in check failure blocks based on the block address and error type; Step S43: Obtain the index label of the error block in the error-tolerant encoded data stream block based on the block address, and obtain the corresponding redundant information in the error-tolerant encoded data stream block based on the index label. At the same time, use the corresponding redundant information in the error-tolerant encoded data stream block to perform online repair on the error block, so as to use the redundant information to repair the error block by using the iterative decoding algorithm. If the repair fails, trigger the re-burn-in process, otherwise perform iterative CRC check until the check is successful; Step S44: Perform burn-in verification analysis on the device burn-in check success blocks corresponding to the successful CRC check based on the device status data, so as to calculate the bit error rate and remaining erase-write life corresponding to each burn-in block according to the device status data, and verify the bit error rate and remaining erase-write life corresponding to each burn-in block according to the preset threshold. If the verification fails, repair and check again to generate the data stream burn-in verification completion result.

10. An intelligent data burning system based on data flow, characterized in that, For implementing the intelligent data burn-in method based on data stream as described in claim 1, the intelligent data burn-in system based on data stream includes: A to-be-burned data stream description output module, which is used to collect the to-be-burned input source data in real time through the bus interface, perform format recognition and semantic parsing on the to-be-burned input source data to generate the metadata information corresponding to the to-be-burned data; obtain the hardware storage structure information and burn-in protocol specification corresponding to the target device, and perform data stream description output on the metadata information corresponding to the to-be-burned data based on the hardware storage structure information and burn-in protocol specification corresponding to the target device, so as to generate a data stream description file including metadata, structure information and burn-in instructions; A burn-in dynamic scheduling module, which is used to obtain the storage bandwidth and cache capacity corresponding to the target device, and perform burn-in dynamic task scheduling on the data stream description file including metadata, structure information and burn-in instructions based on the storage bandwidth and cache capacity corresponding to the target device, so as to generate a data stream burn-in task sequence; The data stream block burning and writing module is used to perform low-density redundant coding and blocking on the corresponding data stream to be burned in the data stream burning task sequence to generate fault-tolerant coded data stream blocks; by real-time monitoring of the target device's corresponding device status data including storage temperature, voltage fluctuation and erase and write count, and burning and writing the fault-tolerant coded data stream blocks based on the device status data, to generate device burning data stream blocks; The online repair module for programming errors is used to perform CRC check on each block of the device programming data stream, locate the error block corresponding to the device programming data stream block, and use the corresponding redundant information in the fault-tolerant coded data stream block and combine it with the device status data to perform online repair verification analysis on the error block to generate a data stream programming verification completion result.

Images