Industrial report management method and system based on distributed storage
By performing local encrypted calculations and deviation corrections in a distributed storage environment, the problem of data aggregation deviations across factory areas was solved, enabling the generation of high-precision industrial reports and enhancing the reliability and usability of the reports.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-28
- Publication Date
- 2026-03-24
AI Technical Summary
In cross-plant industrial report management based on distributed storage, existing technologies cannot effectively solve the problem of data aggregation deviation caused by scattered data storage, which affects the accuracy of reports and decision-making effectiveness.
By performing local encrypted calculations in each factory area, local encrypted aggregation results are generated. These results are then decrypted and weighted at the central processing node. The initial aggregation values are corrected using a deviation correction function, and finally, corrected aggregation values are generated to produce industrial reports.
It achieves high-precision aggregation of data across plant areas, ensuring the accuracy and reliability of reports and improving the reliability of industrial management and decision-making.
Smart Images
Figure CN121722751A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of distributed storage, in particular to an industrial report management method and system based on distributed storage. BACKGROUND
[0002] In the cross-factory industrial report management based on distributed storage, the data of each factory is stored in different nodes and belongs to different data sovereign subjects. Limited by the privacy protection mechanism, the original data cannot be directly transmitted and aggregated across factories. Report data aggregation relies on privacy computing technology (such as partial homomorphic encryption, approximate homomorphic encryption, etc.), rather than directly transmitting the original data.
[0003] Industrial scenarios have strict restrictions on the real-time performance and computing power of report generation. Fully homomorphic encryption is difficult to apply due to the high demand for computing power. The simplified version of the privacy computing scheme actually used (such as only processing integers and discretizing small numbers) will cause small deviations in single-factory data calculation. When aggregated across factories, these single-factory deviations will be amplified. Industrial reports have very high requirements for aggregation accuracy (such as cost accounting and production planning scenarios that require high-precision data support), and accumulated deviations will directly affect the accuracy of the report, leading to deviations in decision-making based on the report. SUMMARY
[0004] The purpose of the present application is to provide an industrial report management method and system based on distributed storage to solve the above technical problems.
[0005] The purpose of the present application can be achieved by the following technical solutions: An industrial report management method based on distributed storage, comprising the following steps: Obtain industrial data from distributed storage nodes of multiple factories; In each factory, use a partial homomorphic encryption algorithm to perform local calculation on the industrial data to generate a local encrypted aggregation result; Transmit the local encrypted aggregation result from each factory to a central processing node; At the central processing node, decrypt the local encrypted aggregation result to obtain a preliminary aggregation value; Apply a bias correction function to process the preliminary aggregation value to generate a corrected aggregation value; Generate an industrial report based on the corrected aggregation value.
[0006] As a further scheme of the present application, obtaining industrial data comprises: Access the distributed storage nodes of each factory and read the industrial data; The industrial data includes production indexes and operation parameters; the industrial data is subjected to data cleaning and normalization before storage; the data cleaning removes invalid data points, and the normalization adjusts data values to a common range; The read industrial data is subjected to access control through a security protocol, and the security protocol verifies identity and an encrypted communication channel.
[0007] As a further scheme of the present application, the local calculation on the industrial data includes: A homomorphic encryption function is applied to the industrial data to generate an encrypted data value, and the homomorphic encryption function is implemented based on a partial homomorphic encryption algorithm; The encrypted data value retains the arithmetic properties of the industrial data; An aggregation operation is performed on the encrypted data value to generate the local encrypted aggregation result, and the aggregation operation includes accumulation and / or weighted average.
[0008] As a further scheme of the present application, the local encrypted aggregation result is transmitted from each plant area to a central processing node, including: The local encrypted aggregation result is encapsulated using a secure communication protocol to generate encapsulated data, and the encapsulated data is subjected to serialization processing to generate serialized data; The serialized data is transmitted to the central processing node through a network; The received serialized data is subjected to deserialization processing at the central processing node to recover the local encrypted aggregation result; The recovered local encrypted aggregation result is subjected to verification, and the verification includes data verification and error checking; The local encrypted aggregation result that passes the verification is decrypted using a decryption key to generate a preliminary aggregation value; wherein the decryption key is stored and managed by the central processing node; The preliminary aggregation value is subjected to data verification, and the data verification is implemented by comparing a checksum.
[0009] As a further scheme of the present application, the preliminary aggregation value is processed using a bias correction function, including: Historical industrial data is obtained from distributed storage nodes of each plant area, and the historical industrial data is subjected to statistical analysis to calculate distribution parameters of the historical industrial data; A partial homomorphic encryption algorithm is applied to the historical industrial data to perform an encryption operation, and the encrypted historical data is subjected to a decryption operation to generate decrypted historical data; A difference between the decrypted historical data and the historical industrial data is calculated to obtain an approximate error; Calculate the weight coefficient of each factory data in the aggregation process based on the variance value in the distribution parameter, wherein the factory area with larger variance value is allocated smaller weight coefficient; Multiply the approximate error of each factory area with the corresponding weight coefficient to obtain the weighted error; Sum all the weighted errors of the factory areas to obtain the error accumulation mode; Use the error accumulation mode as the independent variable and the corresponding actual aggregation value as the dependent variable to establish a linear regression model using a linear regression algorithm; Define the linear regression model as a deviation correction function.
[0010] As a further scheme of the present application: applying the deviation correction function to process the preliminary aggregation value further includes: Calculate the target difference value by subtracting the reference value from the preliminary aggregation value, wherein the reference value is obtained from model prediction; Substitute the target difference value into the deviation correction function, wherein the deviation correction function outputs a deviation estimate value; Adjust the preliminary aggregation value using the deviation estimate value, wherein the adjustment process includes subtracting the deviation estimate value from the preliminary aggregation value to generate a corrected aggregation value.
[0011] As a further scheme of the present application: generating an industrial report based on the corrected aggregation value includes: Combine the corrected aggregation value with additional metadata to generate processed data, wherein the additional metadata includes a timestamp and a factory area identifier; Perform data fusion operation on the processed data to generate a unified view; Perform format conversion operation on the unified view to generate format conversion data; Fill the format conversion data into a report template to generate a filled template, wherein the report template defines a data structure and a display format; Output the filled template as a readable document; Perform rendering operation on the readable document to generate an industrial report.
[0012] An industrial report management system based on distributed storage includes: An acquisition module: obtains industrial data from distributed storage nodes of multiple factory areas; An aggregation module: uses a partially homomorphic encryption algorithm to perform local calculation on the industrial data in each factory area to generate a local encrypted aggregation result; A processing module: transmits the local encrypted aggregation result from each factory area to a central processing node; In the central processing node, decrypt the local encrypted aggregation result to obtain a preliminary aggregation value; A correction module: applying a bias correction function to process the preliminary aggregated value to generate a corrected aggregated value; A visualization module: generating an industrial report based on the corrected aggregated value.
[0013] The beneficial effects of the present application are as follows: The present application realizes the technical effects of not taking data out of the factory area and not leaking privacy in the cross-factory area report aggregation process by completing local aggregation calculation under the encryption condition in the distributed nodes of each factory area and uniformly decrypting and weighting the verified aggregation result in the central processing node. On the premise of maintaining data sovereignty independence, the bias correction function constructed by the present application can correct the approximation error in the aggregation process based on the statistical characteristics of historical data, so that the bias caused by privacy calculation is effectively compensated in the final aggregation stage. Through the weighted processing of the differentiated error influence of each factory area and the dynamic correction of the preliminary aggregated value, the present application can inhibit the aggregation inaccuracy problem caused by the accumulation of cross-factory area bias, so that the final aggregated result is closer to the real business level. At the same time, the report generation stage fuses, formats, and template renders the corrected data to make the industrial report output in a unified structure and readable form, thereby improving the consistency and usability of report display. Overall, the present application realizes high-precision aggregation of cross-factory area data and reliable report output under privacy constraints, enhances the credibility of the report result, and provides stable data support for fine management and decision-making in industrial scenarios. BRIEF DESCRIPTION OF DRAWINGS
[0014] The present application will be further described below in conjunction with the accompanying drawings.
[0015] Figure 1 is a flowchart of an industrial report management method based on distributed storage according to the present application. DETAILED DESCRIPTION
[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0017] Please refer to Figure 1 The present application is an industrial report management method based on distributed storage, which comprises the following steps: Obtaining industrial data from distributed storage nodes of multiple factory areas; In a preferred embodiment of the present application, obtaining industrial data comprises: Accessing the distributed storage nodes of each factory area to read industrial data; The industrial data includes production indicators and operation parameters, and the industrial data is cleaned and normalized before storage, wherein the data cleaning removes invalid data points, and the normalization adjusts data values to a common range. The read industrial data realizes access control through a security protocol, and the security protocol verifies identity and encrypted communication channels.
[0018] It can be understood that when obtaining industrial data from distributed storage nodes of multiple plants, data reading is realized by accessing each node. Before the industrial data enters node storage, the cleaning step removes abnormal points that do not conform to business characteristics or do not have analysis significance, so that the stored data remains structurally complete and reliable in origin. Subsequently, through normalization processing, data from different sources are numerically mapped according to the same scale, so that they can participate in aggregation in a unified manner in subsequent calculations, and the analysis results are not affected by dimensional differences. When reading these preprocessed data, a controlled access channel is established through a security protocol, and the protocol verifies the identity of the access subject and encrypts the communication link to ensure that the data reading process is not disturbed or intercepted by unauthorized parties. The entire acquisition process relies on the independent storage properties of distributed nodes, allowing each plant to provide standardized data that can participate in calculations while maintaining data boundaries, thereby providing a consistent and reliable input basis for subsequent encrypted aggregation operations.
[0019] In each plant, a partially homomorphic encryption algorithm is used to perform local calculation on the industrial data to generate a local encrypted aggregation result. Another preferred embodiment of the application includes performing local calculation on the industrial data, which includes: Applying a homomorphic encryption function to the industrial data to generate an encrypted data value, wherein the homomorphic encryption function is implemented based on a partially homomorphic encryption algorithm. The encrypted data value retains the arithmetic properties of the industrial data. Performing an aggregation operation on the encrypted data value to generate the local encrypted aggregation result, wherein the aggregation operation includes accumulation and / or weighted average.
[0020] It should be noted that when local calculation of industrial data is performed in each factory area, the industrial data is input into a homomorphic encryption function to generate ciphertext, which is constructed based on a partial homomorphic encryption algorithm, so that the encrypted data no longer presents the original numerical value, but the internal coding structure still retains the ability to participate in specific mathematical operations. The so-called arithmetic property refers to the operation performed between ciphertexts that can correspond to the same operation relationship between plaintexts after decryption. For example, addition or weighted processing with addition linearity is completed at the ciphertext level, and the result after decryption is consistent with the result obtained by directly performing these operations on the original plaintext. This property is derived from the preservation of numerical relationship structure by the encryption function during mapping, which enables the data content to be securely shielded while not destroying the mathematical operability between data.
[0021] Since the ciphertext can still perform addition chain operations, the factory area can accumulate or perform weighted processing on local industrial data according to a predetermined weight without exposing the data, thereby forming a local encrypted aggregation result. The entire processing flow utilizes the preservation of the arithmetic structure of the ciphertext to enable aggregation calculation to be completed in an encrypted state, avoiding plaintext leakage, and enabling the value obtained after decryption at the central node to correctly reflect the true aggregation relationship within the factory area, thereby fundamentally ensuring that aggregation calculation based on encrypted data can maintain logical consistency and usability, providing reliable input for the cross-factory area aggregation link.
[0022] Transmitting the local encrypted aggregation result from each factory area to a central processing node; Decrypting the local encrypted aggregation result at the central processing node to obtain a preliminary aggregation value; In another preferred embodiment of the present application, transmitting the local encrypted aggregation result from each factory area to a central processing node comprises: Encapsulating the local encrypted aggregation result using a secure communication protocol to generate encapsulated data, and serializing the encapsulated data to generate serialized data; Transmitting the serialized data to the central processing node through a network; Deserializing the received serialized data at the central processing node to recover the local encrypted aggregation result; Verifying the recovered local encrypted aggregation result, which includes data verification and error checking; Decrypting the local encrypted aggregation result that passes the verification using a decryption key to generate a preliminary factory area aggregation value; wherein the decryption key is stored and managed by the central processing node; Weighted summing all preliminary factory area aggregation values to obtain a preliminary aggregation value, and performing data verification on the preliminary aggregation value by comparing the checksum.
[0023] It is worth noting that in the process of transmitting the local encrypted aggregation result to the central processing node, the encrypted aggregation value is encapsulated by a secure communication protocol before leaving the factory, and the encapsulation process organizes the data structure and communication meta-information to enable the encrypted result to maintain integrity in a controlled environment.
[0024] The encapsulated content is serialized into a form that can be transmitted in the network and sent to the central processing node through the established encrypted communication link. After receiving, the central node recovers the original encrypted aggregation result by deserialization, and uses the data verification and error detection mechanism to confirm that there is no tampering or damage in the transmission process.
[0025] The verified data is decrypted inside the central node using a pre-managed decryption key to obtain preliminary plaintext values reflecting the local calculation results of each factory. The preliminary aggregation values generated by multiple factories need to form a unified cross-factory aggregation result. In the weighted summation process, the weight size is determined according to the distribution characteristics in the historical data, so that the data source with higher variance bears lower weight in aggregation, thereby reducing the impact of potential errors on the final aggregation value.
[0026] This weight setting based on statistical distribution characteristics can structurally offset the deviation caused by local calculation of each factory, making the cross-factory summary closer to the real business level. The preliminary aggregation value obtained after weighting is further checked for consistency through verification and comparison to further confirm the reliability of the summary result, so that the entire cross-factory data aggregation link maintains stable, reliable and verifiable characteristics in the process of encrypted transmission, decryption processing and cross-node fusion.
[0027] applying a bias correction function to the preliminary aggregation value to generate a corrected aggregation value; In a preferred embodiment of the present application, applying a bias correction function to the preliminary aggregation value includes: obtaining historical industrial data from the distributed storage nodes of each factory, statistically analyzing the historical industrial data, and calculating the distribution parameters of the historical industrial data; applying a partial homomorphic encryption algorithm to the historical industrial data to generate encrypted historical data, and applying a decryption operation to the encrypted historical data to generate decrypted historical data; calculating the difference between the decrypted historical data and the historical industrial data to obtain an approximate error; calculating the weight coefficient of each factory data in the aggregation process based on the variance value in the distribution parameters, wherein the factory with a larger variance value is assigned a smaller weight coefficient; multiply the approximate error of each factory by the corresponding weight coefficient to obtain a weighted error; sum all the weighted errors of the factories to obtain an error accumulation mode; a linear regression model is established using a linear regression algorithm, taking the error accumulation mode as the independent variable and the corresponding actual aggregation value as the dependent variable; The linear regression model is defined as a bias correction function.
[0028] It should be noted that when correcting the preliminary aggregation value, a model that can describe the formation of bias is established by means of historical industrial data. After the historical industrial data is obtained from the distributed nodes of each plant, statistical analysis is performed to obtain distribution parameters that describe the characteristics of the data, including but not limited to variance for indicating the degree of data dispersion.
[0029] In constructing the bias structure, the historical industrial data is encrypted by a partial homomorphic encryption algorithm, and then decrypted to obtain decrypted historical data corresponding to the original historical data. Since privacy calculation produces certain approximation errors when operating locally, by comparing the decrypted values with the original historical data without encryption, the deviation introduced by the encryption calculation process in different plants can be revealed. After understanding the error formation mechanism, the weight of each plant in the subsequent aggregation process is determined according to the variance. The higher the variance, the more unstable the data in volatility, so a lower weight is allocated in aggregation to reduce the amplification effect of high volatility data on the overall bias. The approximation error of each plant is adjusted according to the corresponding weight and the sum is obtained, which can obtain the error accumulation mode reflecting the overall bias trend. In establishing the bias correction model, the error accumulation mode is taken as the independent variable, and the actual aggregation value corresponding thereto is taken as the dependent variable, and the mapping relationship between the two is constructed by linear regression method. This model can abstract the error formation law in history into an operable correction function, which is used as a bias correction function for error compensation in subsequent real-time aggregation; the actual aggregation value can be obtained by performing mathematical operations consistent with the encryption aggregation process on the historical industrial data in plaintext conditions.
[0030] In another preferred embodiment of the present application, the application of the bias correction function to the preliminary aggregation value further includes: The preliminary aggregation value is difference calculated with a reference value to obtain a target difference value, the reference value being from model prediction; The target difference value is substituted into the bias correction function, and the bias correction function outputs a bias estimate value; The bias estimate value is used to adjust the preliminary aggregation value, and the adjustment process includes subtracting the bias estimate value from the preliminary aggregation value to generate a corrected aggregation value.
[0031] It should be noted that when performing bias correction on the preliminary aggregation value, the degree of deviation generated in the current aggregation process needs to be identified through the difference between it and the reference value generated by the model prediction. The reference value reflects the trend that the aggregation result should present under the condition of no deviation. When the preliminary aggregation value is subtracted from the reference value, the result obtained can reveal the deviation direction and amplitude of the preliminary aggregation value relative to the ideal output, and this difference is defined as the target difference. After the target difference is put into the bias correction function, the bias correction function combines the cumulative pattern of historical errors and the sensitivity formed by the weight to generate a bias estimate value that describes the size of the current bias, which reflects the amount that needs to be corrected according to the target difference. The essence of bias correction is to separate the deviation caused by privacy calculation and cross-node aggregation from the preliminary aggregation value, so that the result returns to a level that is more consistent with historical rules and industrial reality, and therefore the bias estimate value is subtracted from the preliminary aggregation value in the final adjustment process.
[0032] The method of generating the reference value can be completed based on the prediction of the sequence rule formed from historical industrial data. The process can start by extracting time series features from historical industrial data of each plant area, and then arranging these historical data in chronological order and dividing them into several adjacent sliding windows. The length of each window can be set according to the update frequency of industrial data, for example, ten to twenty cycles can be used as a window in daily production data, so that the window can cover the common components of periodic fluctuations and random fluctuations.
[0033] Within each window, the basic statistical features are formed by calculating the mean and median, and then combined with the short-term change trend of the data within the window to construct the input sequence for prediction. These trends can be represented by the difference between adjacent data points, the sign of the difference, the directionality of the change, etc., so that the prediction model not only receives static numerical information, but also obtains clues about the rate of change.
[0034] On this basis, a simple time series prediction model can be constructed, for example, a linear trend extrapolation method can be used to extend the change trend within the window to the current time point in a linear manner. The slope and intercept are obtained by least squares fitting of several data points at the end of the window, thereby generating a prediction value representing the future data level. When processing data from different plant areas, each plant area can independently generate a prediction value. In order to obtain the overall reference value across plant areas, the prediction results need to be synthesized. In the synthesis process, the stability indicators in the historical data can be referred to to assign weights to different plant areas. These stability indicators can be represented by the variance level of the historical time series. Generally, plant areas with smaller variance are assigned higher weight values. The weight can be roughly set between zero point one and zero point nine, so that the prediction values of different plant areas are synthesized according to their data stability.
[0035] The weighted prediction value is added to obtain a reference value, which can reflect the natural trend of the industrial data in the current period without relying on the bias correction model, and provide a reliable basis for subsequent comparison of the preliminary aggregated value and the reference value.
[0036] It is worth noting that the reference value serves to provide a trend level inferred based on historical rules, which is a prediction value and does not represent the industrial data actually occurring in the current period. The generation of the industrial report must be based on the latest production situation of each factory area, and the preliminary aggregated value is the result of decryption and aggregation of the real encrypted data in the current period, which reflects the actual production and operation state in the current period. If the report is directly generated based on the reference value, the report will lose the reflection ability of the data in the current period, and can only reflect the trend inferred by the model and cannot represent the real situation. The significance of the bias correction process is to establish a connection between the predicted trend and the actual data, and to make the real data in the current period closer to the explainable level after correction, so as to ensure that the report is based on the real collected data and has higher aggregation accuracy.
[0037] Generating an industrial report based on the corrected aggregated value.
[0038] In another preferred embodiment of the present application, generating an industrial report based on the corrected aggregated value comprises: Combining the corrected aggregated value with additional metadata to generate to-be-processed data, the additional metadata including a timestamp and a factory area identifier; Performing a data fusion operation on the to-be-processed data to generate a unified view; Performing a format conversion operation on the unified view to generate format conversion data; Filling the format conversion data into a report template to generate a filled template, the report template defining a data structure and a display format; Outputting the filled template as a readable document; Performing a rendering operation on the readable document to generate an industrial report.
[0039] It can be understood that in the process of generating an industrial report using the corrected aggregated value, the value needs to be combined with information that can identify the time and the factory area source, so that the corrected aggregated value can be accurately located in the corresponding business background in the subsequent steps. The timestamp is used to represent the specific time point at which the data is formed, so that the report can distinguish the business performance of different periods, and the factory area identifier is used to indicate the area where the aggregated value is sourced, so that the presentation of cross-factory area data has a structured dimension.
[0040] After these contents are combined into the data to be processed, the scattered data elements need to be integrated into a unified view through a data fusion method. The logic of data fusion is to align data of different sources and structures so that they can be uniformly processed in the same semantic framework. Usually, the fields need to be standardized mapped so that all items can be expressed with consistent labels.
[0041] After the unified view is formed, it needs to be converted into a format matching the structure of the report template. This process is achieved through format conversion. The principle is to reorganize the content of the unified view according to the pre-defined data field structure of the template, so that it can be arranged in the order, type and level required by the template. The template predefines the fields, layout, title and table structure that need to be displayed. Therefore, when the data after format conversion is filled into the template, the template can automatically present the complete content according to the pre-set structure.
[0042] The filled template is output as a document format that can be directly read, so as to enter the rendering stage. The rendering process presents the data as the final industrial report according to the layout instructions, structure labels and visual styles in the document format, so that the report has clear hierarchy and readability in vision. In this way, the final transformation from numerical results to business report is completed.
[0043] The present application combines the encrypted data processing flow under the distributed storage condition with the bias correction mechanism, so that the generation of cross-factory industrial report can still obtain high-precision aggregation results in the environment limited by privacy. Cross-factory data cannot be directly aggregated in plaintext, but can only rely on partial homomorphic encryption and other privacy computing methods. Privacy computing will inevitably introduce local bias when performing local arithmetic at each node. The bias will be amplified after multi-factory aggregation, resulting in the final aggregated value deviating from the true level.
[0044] The application realizes the computability of cross-factory data without exposing the plaintext by completing the local aggregation of data under the encryption condition in each factory area and uniformly decrypting and fusing the encrypted aggregation results at the central node, thereby establishing an executable cross-area report aggregation link in an industrial environment with limited data sovereignty. Based on the partial homomorphic encryption operation, an uneven approximation error is introduced in different factory areas. The application uses historical data to construct a bias accumulation mode, and forms a bias correction function based on the statistical relationship between the mode and the actual aggregation value, so that the error caused by privacy calculation can be quantified and corrected in the cross-factory aggregation stage. By weighting the errors of each factory area according to their statistical fluctuation degree, the influence of unstable data in aggregation is automatically converged, thereby suppressing the cascade amplification of errors in the cross-node aggregation link. Through the combination of bias modeling and weight adjustment, the encrypted data across factory areas can still approach the real business level after aggregation, and high-precision report generation is realized under the condition of limited privacy, solving the contradiction between privacy protection and aggregation precision in traditional encryption aggregation, and realizing a breakthrough in the credibility and usability of cross-factory reports.
[0045] An industrial report management system based on distributed storage, comprising: An acquisition module: acquiring industrial data from distributed storage nodes of multiple factory areas; An aggregation module: locally calculating the industrial data using a partial homomorphic encryption algorithm in each factory area to generate a local encrypted aggregation result; A processing module: transmitting the local encrypted aggregation result from each factory area to a central processing node; At the central processing node, a decryption operation is performed on the local encrypted aggregation result to obtain a preliminary aggregation value; A correction module: processing the preliminary aggregation value using a bias correction function to generate a corrected aggregation value; A visualization module: generating an industrial report based on the corrected aggregation value.
[0046] The above describes one embodiment of the application in detail, but the content is only the preferred embodiment of the application and cannot be considered as limiting the scope of the application. Any equivalent changes and improvements made within the scope of the application should still be within the scope of the application.
Claims
1. An industrial report management method based on distributed storage, characterized in that, Includes the following steps: Industrial data is obtained from distributed storage nodes across multiple factory areas; In each factory area, a partially homomorphic encryption algorithm is used to perform local calculations on the industrial data to generate locally encrypted aggregated results; The local encrypted aggregation results are transmitted from each factory area to the central processing node; At the central processing node, the local encrypted aggregation result is decrypted to obtain a preliminary aggregation value; The initial aggregate value is processed using a deviation correction function to generate a corrected aggregate value; An industrial report is generated based on the corrected aggregate value.
2. The industrial report management method based on distributed storage according to claim 1, characterized in that, Acquiring industrial data includes: Access the distributed storage nodes in each factory area to read industrial data; The industrial data includes production indicators and operating parameters; the industrial data undergoes data cleaning and normalization before storage; data cleaning removes invalid data points, and normalization adjusts data values to a common range; The read industrial data is access controlled through a security protocol that verifies identity and encrypted communication channels.
3. The industrial report management method based on distributed storage according to claim 1, characterized in that, The partial calculations performed on the industrial data include: The industrial data is processed by applying a homomorphic encryption function to generate encrypted data values, wherein the homomorphic encryption function is implemented based on a partial homomorphic encryption algorithm. The encrypted data value retains the arithmetic properties of the industrial data; An aggregation operation is performed on the encrypted data value to generate the local encrypted aggregation result, wherein the aggregation operation includes summation and / or weighted average.
4. The industrial report management method based on distributed storage according to claim 1, characterized in that, The transmission of the local encrypted aggregation results from each plant area to the central processing node includes: The local encrypted aggregation result is encapsulated using a secure communication protocol to generate encapsulated data, and the encapsulated data is serialized to generate serialized data. The serialized data is transmitted to the central processing node via the network; The central processing node performs deserialization processing on the received serialized data to restore the local encrypted aggregation result; The recovered partially encrypted aggregated results are verified, including data verification and error checking. The decryption key is used to decrypt the verified local encrypted aggregation result to generate a preliminary plant area aggregation value; wherein, the decryption key is stored and managed by the central processing node; A weighted summation is performed on all the preliminary aggregate values of the plant area to obtain a preliminary aggregate value. The preliminary aggregate value is then validated by comparing a checksum.
5. The industrial report management method based on distributed storage according to claim 1, characterized in that, The application of a deviation correction function to process the initial aggregate value includes: Historical industrial data is obtained from distributed storage nodes in each factory area, and statistical analysis is performed on the historical industrial data to calculate the distribution parameters of the historical industrial data. The historical industrial data is encrypted using a partial homomorphic encryption algorithm to generate encrypted historical data; the encrypted historical data is then decrypted to generate decrypted historical data. The difference between the decrypted historical data and the historical industrial data is calculated to obtain the approximate error; The weighting coefficients of each plant area's data in the aggregation process are calculated based on the variance values in the distribution parameters, with plants having larger variance values assigned smaller weighting coefficients. The approximate error of each plant area is multiplied by the corresponding weighting coefficient to obtain the weighted error; The weighted errors of all plant areas are summed to obtain the error accumulation pattern. Using the error accumulation pattern as the independent variable and the corresponding actual aggregate value as the dependent variable, a linear regression model is established using a linear regression algorithm. The linear regression model is defined as a bias correction function.
6. The industrial report management method based on distributed storage according to claim 1, characterized in that, The process of applying the deviation correction function to the preliminary aggregate value also includes: The target difference is obtained by calculating the difference between the preliminary aggregated value and the reference value, where the reference value comes from the model prediction; Substitute the target difference into the deviation correction function, and the deviation correction function outputs a deviation estimate. The initial aggregate value is adjusted using the deviation estimate, and the adjustment process includes subtracting the deviation estimate from the initial aggregate value to generate a corrected aggregate value.
7. The industrial report management method based on distributed storage according to claim 1, characterized in that, Generating industrial reports based on the corrected aggregate value includes: The corrected aggregate value is combined with additional metadata to generate data to be processed, the additional metadata including timestamp and plant area identifier; Perform a data fusion operation on the data to be processed to generate a unified view; Perform a format conversion operation on the unified view to generate format conversion data; The format-converted data is populated using a report template to generate a populated template, wherein the report template has a predefined data structure and display format; The filled template is then output as a readable document. The readable document is rendered to generate an industrial report.
8. An industrial report management system based on distributed storage, characterized in that, include: Acquisition module: Acquires industrial data from distributed storage nodes across multiple factory areas; Aggregation module: In each factory area, a partially homomorphic encryption algorithm is used to perform local calculations on the industrial data to generate locally encrypted aggregation results; Processing module: Transmits the local encrypted aggregation results from each plant area to the central processing node; At the central processing node, the local encrypted aggregation result is decrypted to obtain a preliminary aggregation value; Correction module: Applies a deviation correction function to process the initial aggregate value to generate a corrected aggregate value; Visualization module: Generates industrial reports based on the corrected aggregated values.