Device operation and maintenance data management system and method based on multimodal data fusion

Abstract

This application proposes a device operation and maintenance data management system and method based on multi-modal data fusion. The system includes: a retrospective portrait acquisition module for obtaining a retrospective portrait of the acquisition moment of the target dimension data of the target device tag; a real-time portrait acquisition module for obtaining a real-time portrait of the target device tag; a portrait deviation analysis module for constructing a portrait deviation vector; a threshold judgment module for obtaining a data deletion ratio; and a data cleaning module for deleting operation and maintenance data according to the data deletion ratio. This application solves the technical problems in the prior art that the continuous increase of device operation and maintenance data leads to large storage pressure, and the traditional deletion method cannot intelligently judge the importance of data and requires manual intervention, resulting in low data management efficiency, and achieves the technical effect of automatically evaluating the data value based on the device service life and control parameter changes, and intelligently cleaning low-reference-value data, thereby reducing the storage pressure and improving the data management efficiency.

Description

Technical Field

[0001] The present invention relates to the field of data management, and particularly to a device operation and maintenance data management system and method based on multi-modal data fusion. Background Art

[0002] With the continuous development of industrial automation and intelligent manufacturing, the collection and analysis of device operation and maintenance data have become important means to ensure device safety, extend device life, and improve production efficiency. In a modern industrial environment, various devices generate a large amount of operation and maintenance data every moment, including but not limited to parameters such as temperature, pressure, vibration, and flow rate. These data are collected by various sensors and stored in a database for device status monitoring, fault prediction, and maintenance decision-making. However, with the increase in data collection frequency and the number of devices, the device operation and maintenance data have grown exponentially, bringing huge pressure to data storage and management. A large amount of data not only occupies server disk space but also may lead to a decline in data query and analysis efficiency, increasing the system operation cost.

[0003] Currently, for device operation and maintenance data management, common data cleaning is mainly based on time factors, that is, deleting historical data with a longer storage time, or based on usage frequency, deleting data with fewer access times. Although these data management methods alleviate the storage pressure to a certain extent, they have obvious deficiencies: First, they cannot intelligently judge the importance and reference value of data, which may lead to the accidental deletion of historical data with important reference significance; Second, it usually requires manual intervention, such as setting deletion rules, confirming the deletion range, etc., and the operation is cumbersome and inefficient; At the same time, simple deletion methods based on time or usage frequency cannot consider the dynamic changes in the device operation state and are difficult to meet the data management requirements of the entire device life cycle. Therefore, in the prior art, the continuous increase in device operation and maintenance data leads to a large storage pressure, and the traditional deletion method cannot intelligently judge the importance of data and requires manual intervention, resulting in low data management efficiency. Summary of the Invention

[0004] In view of the technical problems in the prior art that the continuous increase in device operation and maintenance data leads to a large storage pressure, and the traditional deletion method cannot intelligently judge the importance of data and requires manual intervention, resulting in low data management efficiency, the present invention provides a device operation and maintenance data management system and method based on multi-modal data fusion to solve the problems.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] In a first aspect, the present invention provides a device operation and maintenance data management system based on multi-modal data fusion, including: a retrospective portrait acquisition module, configured to obtain a retrospective portrait of the acquisition moment of the target-dimensional data of the target device number when a preset management period is satisfied, where the retrospective portrait includes a first service duration and a first control parameter; a real-time portrait acquisition module, configured to obtain a real-time portrait of the target device number, where the real-time portrait includes a second service duration and a second control parameter; a portrait deviation analysis module, configured to construct a portrait deviation vector according to the first service duration and the first control parameter, and the second service duration and the second control parameter; a threshold judgment module, configured to retrieve a set of deviation modulus values of the target-dimensional data that satisfy the portrait deviation vector, and when the variance of the set of deviation modulus values of the target-dimensional data is greater than or equal to a variance threshold, calculate the ratio of a first duration between the storage start moment and the acquisition moment to a second duration between the storage start moment and the current moment, and set it as a data deletion ratio; a data cleaning module, configured to delete the target-dimensional data from the operation and maintenance data of the target device number starting from the storage start moment according to the data deletion ratio.

[0007] Further, the retrospective portrait acquisition module includes: a parameter acquisition unit, configured to obtain the first preset control parameter to the Nth preset control parameter of the target device number; a first statistics unit, configured to statistically calculate a first abnormal service duration of the target-dimensional data based on a historical sample library with the first preset control parameter as a constraint; a multi-parameter statistics unit, configured to statistically calculate the Nth abnormal service duration of the target-dimensional data until the Nth preset control parameter is used as a constraint based on the historical sample library; a period calculation unit, configured to extract 0.5 times the minimum value of the first abnormal service duration to the Nth abnormal service duration, and set it as the preset management period.

[0008] Further, the retrospective portrait acquisition module further includes: a storage threshold unit, configured to obtain a disk space storage capacity threshold of the operation and maintenance data of the target device number; a portrait retrospective trigger unit, configured to obtain a retrospective portrait of the acquisition moment of the target-dimensional data of the target device number when the amount of operation and maintenance data of the target device number is greater than or equal to the disk space storage capacity threshold.

[0009] Further, the first statistical unit includes: a data retrieval subunit, configured to retrieve, from the historical sample library, a corresponding number of underlying perception attribute record feature values and a corresponding number of first service duration record values that meet the first preset control parameter when the target dimension data is an underlying perception attribute; a deviation calculation subunit, configured to perform pairwise deviation analysis on the corresponding number of underlying perception attribute record feature values and the corresponding number of first service duration record values to obtain a set of deviation modulus values of underlying perception attribute record feature values and a set of deviation modulus values of first service duration record values; a feature screening subunit, configured to extract, from the set of deviation modulus values of first service duration record values, a selected set of deviation modulus values of first service duration record values for which the set of deviation modulus values of underlying perception attribute record feature values is greater than or equal to a predefined record feature value deviation modulus threshold; and a mode evaluation subunit, configured to perform mode evaluation on the selected set of deviation modulus values of first service duration record values to obtain the first dissimilated service duration.

[0010] Further, the first statistical unit includes: a mapper scheduling subunit, configured to schedule a target dimension data mapper bound to the high-level mapping attribute and the first preset control parameter when the target dimension data is a high-level mapping attribute, where the target dimension data mapper has an input underlying perception attribute; an associated data retrieval subunit, configured to retrieve, from the historical sample library, a corresponding number of input underlying perception attribute record feature values and a corresponding number of second service duration record values that meet the first preset control parameter; a feature mapping processing subunit, configured to process the corresponding number of input underlying perception attribute record feature values respectively through the target dimension data mapper to obtain a corresponding number of high-level mapping attribute feature values; and a dissimilated duration calculation subunit, configured to calculate the first dissimilated service duration according to the corresponding number of high-level mapping attribute feature values and the corresponding number of second service duration record values.

[0011] Further, the system further includes: an attribute configuration module, configured to configure, through the user terminal, a preset input underlying perception attribute of a preset high-level mapping attribute; a mapper training module, configured to collect multiple sets of data and train the target dimension data mapper with a preset target device model and a preset control parameter as constraints; and a mapping binding module, configured to bind the target dimension data mapper to the preset high-level mapping attribute and the preset control parameter; where any one of the multiple sets of data includes: preset input underlying perception attribute record data and preset high-level mapping attribute record data.

[0012] Further, the threshold judgment module includes: a deviation decomposition unit, configured to decompose the portrait deviation vector into a service life deviation vector and a control parameter deviation vector; a cloud retrieval unit, configured to send a retrieval request to the cloud, obtain a secret key, encrypt the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector to obtain a ciphertext, and upload the ciphertext to the cloud to obtain retrieval feedback information, where the retrieval feedback information includes an initial target dimension data deviation modulus set; an outlier processing unit, configured to delete outliers from the initial target dimension data deviation modulus set to obtain the target dimension data deviation modulus set.

[0013] Further, the cloud retrieval unit includes: a cloud decryption subunit, configured to decrypt the ciphertext through the cloud to obtain the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector, and construct a retrieval task constraint; a node request subunit, configured to send a retrieval request to a distributed node through the cloud to obtain a distributed node secret key; an encrypted retrieval subunit, configured to encrypt the retrieval task constraint and a retrieval key according to the distributed node secret key to obtain a retrieval task ciphertext, send the ciphertext to the distributed node, and obtain retrieval feedback information, where the retrieval feedback information includes the initial target dimension data deviation modulus set.

[0014] In a second aspect, the present invention provides a device operation and maintenance data management method based on multi-modal data fusion, including: when a preset management period is satisfied, obtaining a retrospective portrait of the acquisition time of the target dimension data of the target device tag, where the retrospective portrait includes a first service life and a first control parameter; obtaining a real-time portrait of the target device tag, where the real-time portrait includes a second service life and a second control parameter; constructing a portrait deviation vector according to the first service life and the first control parameter, and the second service life and the second control parameter; retrieving a target dimension data deviation modulus set that satisfies the portrait deviation vector, and when the variance of the target dimension data deviation modulus set is greater than or equal to a variance threshold, calculating a ratio of a first duration between the storage start time and the acquisition time to a second duration between the storage start time and the current time, and setting it as a data deletion ratio; deleting the target dimension data from the operation and maintenance data of the target device tag starting from the storage start time according to the data deletion ratio.

[0015] The beneficial effects of the present invention are:

[0016] Through the retrospective image acquisition module, when the preset management cycle is met, a retrospective image of the acquisition moment of the target dimension data of the target device number is obtained, where the retrospective image includes the first service duration and the first control parameter, providing basic data for subsequent comparative analysis to ensure the current reference value of historical data can be evaluated. Through the real-time image acquisition module, a real-time image of the target device number is obtained, where the real-time image includes the second service duration and the second control parameter, so as to compare with historical data and reflect the change of the device state. Through the image deviation analysis module, according to the first service duration and the first control parameter, as well as the second service duration and the second control parameter, an image deviation vector is constructed to quantify the degree of change in the device operation state, providing a basis for judging the value of historical data. Through the threshold judgment module, a set of deviation modulus values of the target dimension data that satisfy the image deviation vector is retrieved. When the variance of the set of deviation modulus values of the target dimension data is greater than or equal to the variance threshold, the ratio of the first duration between the storage start moment and the acquisition moment to the second duration between the storage start moment and the current moment is calculated and set as the data deletion ratio. The significance of data fluctuation is determined through statistical analysis, and the data ratio that needs to be cleaned is calculated accordingly, realizing the intelligent evaluation of data value. Through the data cleaning module, according to the data deletion ratio, the target dimension data is deleted from the operation and maintenance data of the target device number starting from the storage start moment, and the actual data cleaning operation is performed, intelligently cleaning low-value data according to the calculated deletion ratio, thereby reducing the storage pressure.

[0017] Through the above technical solution, the system can automatically identify and clean historical data that has lost its reference value due to changes in the device state during the device operation and maintenance process according to the changes in the device service duration and control parameters. When the variance of the data of a certain dimension is large, it indicates that the reference value of the data of this dimension for device operation and maintenance has decreased, and the system will automatically clean these data according to a certain ratio, thus realizing the intelligent management of device operation and maintenance data, improving the data management efficiency, and effectively reducing the storage pressure. Brief Description of the Drawings

[0018] Figure 1 It is a schematic structural diagram of the device operation and maintenance data management system based on multi-modal data fusion provided by the present invention;

[0019] Figure 2 It is a schematic flow diagram of the device operation and maintenance data management method based on multi-modal data fusion provided by the present invention.

[0020] In the drawings, the components represented by the reference numerals are as follows:

[0021] Retrospective image acquisition module 11, real-time image acquisition module 12, image deviation analysis module 13, threshold judgment module 14, data cleaning module 15. Detailed Embodiment

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined.

[0024] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for the purpose of explanation. It should be understood that those skilled in the art can realize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0025] Embodiment 1, as Figure 1 shown, the embodiment of the present invention provides a device operation and maintenance data management system based on multi-modal data fusion, including:

[0026] A retrospective portrait acquisition module 11, configured to obtain a retrospective portrait of the acquisition moment of the target dimension data of the target device number when a preset management period is satisfied, where the retrospective portrait includes a first service duration and a first control parameter.

[0027] Specifically, when it is detected that the preset management period is satisfied, the retrospective portrait acquisition module 11 acquires the device state information at the data acquisition moment for the target dimension data of the target device number (for example, a specific data attribute of the device, which may be waveform data, non-waveform data, abnormal identification data, normal identification data, etc.). These state information constitute the retrospective portrait, including the first service duration and the first control parameter. Among them, the first service duration reflects the cumulative value of the running time of the target device at that time, and the first control parameter records the set value of the control parameter of the target device at that time.

[0028] By obtaining the retrospective image, a benchmark for the past operating state of the device can be established, providing a basis for subsequent comparative analysis with the real-time state.

[0029] The real-time image acquisition module 12 is used to obtain the real-time image of the target device tag number. Among them, the real-time image includes the second service duration and the second control parameter.

[0030] Specifically, the real-time image acquisition module 12 collects the second service duration and the second control parameter at the current moment by monitoring the operating state of the target device tag number in real time. Among them, the second service duration represents the cumulative value of the device service duration of the target device at present, and the second control parameter represents the set value of the control parameter of the target device at present. The second service duration and the second control parameter together constitute the real-time image reflecting the current state of the target device.

[0031] Through the real-time image acquisition module 12, it is ensured that the system can continuously obtain the latest state information of the device, providing real-time data support for subsequent comparative analysis with the retrospective image. By capturing the control parameter and service duration of the device in real time, the change of the device state can be dynamically tracked, providing a basis for realizing the management of device operation and maintenance data.

[0032] The image deviation analysis module 13 is used to construct an image deviation vector according to the first service duration, the first control parameter, the second service duration, and the second control parameter.

[0033] Specifically, the image deviation analysis module 13 is used to compare and analyze the differences between the retrospective image and the real-time image, and quantify these differences into an image deviation vector.

[0034] The image deviation analysis module 13 constructs an image deviation vector representing the change of the target device state by calculating the differences between the corresponding parameters of the retrospective image and the real-time image, that is, calculating the differences between the first service duration and the first control parameter, and the differences between the second service duration and the second control parameter. This image deviation vector contains the change amount of the service duration and the change amount of the control parameter, comprehensively reflecting the degree of state change of the target device from the historical acquisition moment to the current moment.

[0035] The image deviation vector, as a quantitative index for measuring the change of the device operating state, provides a basis for subsequent data value evaluation and storage management decision-making. In this way, the dynamic change characteristics of the device operating state can be accurately captured, providing a basis for subsequent data cleaning and realizing the intelligent management of device operation and maintenance data.

[0036] The threshold judgment module 14 is used to retrieve the set of deviation modulus values of the target dimension data that satisfy the portrait deviation vector. When the variance of the set of deviation modulus values of the target dimension data is greater than or equal to the variance threshold, calculate the ratio of the first duration from the storage start time to the acquisition time to the second duration from the storage start time to the current time, and set it as the data deletion ratio.

[0037] Specifically, first, the threshold judgment module 14 retrieves and obtains the set of deviation modulus values of the target dimension data that satisfy the portrait deviation vector based on the constructed portrait deviation vector; second, calculate the statistical variance of the set of deviation modulus values of the target dimension data; when the calculated variance value is greater than or equal to the preset variance threshold, it is determined that there is a large fluctuation in the target dimension data, and the reference value for device operation and maintenance is reduced; subsequently, the threshold judgment module 14 calculates the ratio relationship between two time periods, that is, the ratio of the first duration from the storage start time to the data acquisition time to the second duration from the storage start time to the current time, and sets this ratio as the data deletion ratio.

[0038] The threshold judgment module 14 realizes the evaluation of the data fluctuation degree through variance analysis. When the fluctuation is large, it indicates that the target device may be worn due to the increase in service life, resulting in a reduction in the reference value of the data under the same control parameters stored earlier. By dynamically calculating the data deletion ratio, the scope of data cleaning can be accurately controlled, which can not only effectively reduce the storage pressure but also retain the data with important reference value, thereby improving the intelligent level and efficiency of device operation and maintenance.

[0039] The data cleaning module 15 is used to delete the target dimension data from the operation and maintenance data of the target device number starting from the storage start time according to the data deletion ratio.

[0040] Specifically, the data cleaning module 15 selectively cleans the operation and maintenance data of the target device number according to the data deletion ratio calculated by the threshold judgment module 14. The data cleaning module 15 receives the data deletion ratio from the threshold judgment module 14 and quantitatively deletes the target dimension data with the storage start time as the reference point. The data cleaning module 15 deletes the specified proportion of the target dimension data in chronological order starting from the storage start time to ensure the accurate execution of the data deletion operation.

[0041] Through the selective deletion mechanism based on data value judgment, the operation and maintenance data of the device can be intelligently managed, the storage space can be effectively released, and the data with important reference value can be retained at the same time. Compared with the traditional method of data deletion only based on the length of time, this system realizes the accurate evaluation of data value, ensures that the data with lower reference value is deleted, thereby avoiding the misdeletion of important data, improving the intelligent level of device operation and maintenance data management and the operation and maintenance efficiency, and at the same time reducing the data storage pressure.

[0042] Further, the retrospective image acquisition module 11 includes:

[0043] A parameter acquisition unit, configured to obtain the first preset control parameter to the Nth preset control parameter of the target device tag number;

[0044] A first statistical unit, configured to, with the first preset control parameter as a constraint, statistically analyze the first dissimilar service life of the target dimension data based on the historical sample library;

[0045] A multi-parameter statistical unit, configured to, until the Nth preset control parameter is used as a constraint, statistically analyze the Nth dissimilar service life of the target dimension data based on the historical sample library;

[0046] A period calculation unit, configured to extract 0.5 times the minimum value of the first dissimilar service life to the Nth dissimilar service life, and set it as the preset management period.

[0047] In a preferred embodiment, the retrospective image acquisition module 11 includes a parameter acquisition unit, a first statistical unit, a multi-parameter statistical unit, and a period calculation unit. These units work together to determine the preset management period:

[0048] First, the parameter acquisition unit acquires all the preset control parameters of the target device tag number, from the first preset control parameter to the Nth preset control parameter, so as to comprehensively master the operation parameter settings of the device under different control modes. For example, for a pump device, these parameters may include operating parameters such as rotational speed, flow rate, and pressure. Then, under the constraint condition of the first preset control parameter, the first statistical unit statistically analyzes the dissimilar service life of the target dimension data by querying the historical sample library, and obtains the first dissimilar service life, so as to focus on the historical change law of the target dimension data under a single control parameter. For example, when the pump device operates at a fixed rotational speed of 2000 revolutions per minute, the first statistical unit analyzes the average time required for the vibration value data to deviate from the normal range, and thus obtains the first dissimilar service life, such as 1500 hours.

[0049] Subsequently, the multi-parameter statistical unit performs historical sample analysis under constraint conditions from the second preset control parameter to the Nth preset control parameter, and obtains a complete statistical result from the second dissimilar service life to the Nth dissimilar service life, ensuring a comprehensive grasp of the dissimilarity law of the target device under various preset control parameters. Then, the period calculation unit extracts 0.5 times the minimum value of all dissimilar service lives (from the first dissimilar service life to the Nth dissimilar service life) through comparative analysis, and takes it as the preset management period. In this way, the monitoring period can be personalized for the target dimension data, ensuring evaluation before the data value changes significantly, and effectively improving the accuracy and timeliness of equipment operation and maintenance.

[0050] Through the determination of a preset management cycle based on multi-control parameter analysis, a dynamic monitoring mechanism for equipment operation and maintenance data is realized, avoiding the lag problem caused by a fixed cycle, and improving the sensitivity and response speed of the system to changes in equipment status.

[0051] Furthermore, the retrospective portrait acquisition module 11 further includes:

[0052] A storage threshold unit for obtaining the disk space storage capacity threshold of the operation and maintenance data of the target device tag.

[0053] A portrait retrospective trigger unit for obtaining the retrospective portrait of the acquisition moment of the target dimension data of the target device tag when the operation and maintenance data volume of the target device tag is greater than or equal to the disk space storage capacity threshold.

[0054] In an optional implementation manner, the retrospective portrait acquisition module 11 includes a storage threshold unit and a portrait retrospective trigger unit for triggering the retrospective portrait acquisition process based on storage pressure.

[0055] The storage threshold unit is responsible for monitoring and obtaining the disk space storage capacity threshold of the operation and maintenance data of the target device tag. For example, a storage capacity threshold of 10 GB is set for a certain pump device, and a total storage capacity threshold of 50 GB is set for the valve device group on a certain production line. The portrait retrospective trigger unit implements an intelligent trigger mechanism based on storage pressure. When the portrait retrospective trigger unit detects that the operation and maintenance data volume of the target device tag reaches or exceeds the pre-set disk space storage capacity threshold, it automatically starts the retrospective portrait acquisition process to obtain the retrospective portrait of the target dimension data of the target device tag at the acquisition moment. For example, when the operation and maintenance data of a certain pump device accumulates to 10 GB and reaches its preset threshold of 10 GB, it automatically obtains the retrospective portrait of the vibration data of the pump device at the historical acquisition moment.

[0056] Through the trigger mechanism based on storage pressure, which complements the time-period-based management method, they jointly constitute a comprehensive operation and maintenance data management strategy. By real-time monitoring the storage capacity and triggering the data analysis process when the disk space storage capacity threshold is reached, it can timely address the storage pressure problem, avoid the decline in system performance caused by excessive data accumulation, and ensure that the data cleaning process is based on a scientific evaluation of data value rather than a simple chronological order.

[0057] Furthermore, the first statistical unit includes:

[0058] A data retrieval sub-unit for, when the target dimension data is a bottom-layer perception attribute, retrieving from the historical sample library a corresponding number of bottom-layer perception attribute record feature values and a corresponding number of first service duration record values that meet the first preset control parameter.

[0059] A deviation calculation subunit, configured to perform pairwise deviation analysis on a plurality of corresponding underlying perception attribute record feature values and a plurality of first service duration record values, and obtain an underlying perception attribute record feature value deviation modulus set and a first service duration record value deviation modulus set;

[0060] A feature screening subunit, configured to extract, from the first service duration record value deviation modulus set, a first service duration record value selected deviation modulus set for which the underlying perception attribute record feature value deviation modulus set is greater than or equal to a predefined record feature value deviation modulus threshold;

[0061] A mode evaluation subunit, configured to perform mode evaluation on the first service duration record value selected deviation modulus set, and obtain the first dissimilated service duration.

[0062] In a preferred embodiment, when the target dimension data is an underlying perception attribute, the first statistical unit, through the data retrieval subunit, the deviation calculation subunit, the feature screening subunit, and the mode evaluation subunit, is configured to accurately calculate the dissimilated service duration.

[0063] When the target dimension data is an underlying perception attribute (such as direct measurement values such as temperature, pressure, vibration, etc.), the data retrieval subunit retrieves data pairs that meet the conditions of the first preset control parameter from the historical sample library. These data pairs include corresponding underlying perception attribute record feature values and a plurality of first service duration record values. For example, when the first preset control parameter is "rotation speed 2000 revolutions per minute", retrieve all the vibration value records and corresponding service duration records of the devices operating at this rotation speed. Then, the deviation calculation subunit performs pairwise comparison analysis on the retrieved data pairs. Specifically, the deviation calculation subunit calculates the differences between any two sets of data, forming two deviation modulus sets, namely the underlying perception attribute record feature value deviation modulus set and the first service duration record value deviation modulus set. For example, calculate the differences between any two measured vibration values and the corresponding service duration differences.

[0064] Next, the feature screening subunit extracts from the set of deviation modulus values of the first service duration record values those items whose corresponding deviation modulus values of the underlying perception attribute record features are greater than or equal to the predefined threshold of the deviation modulus value of the record feature, forming a set of selected deviation modulus values of the first service duration record values, so as to be able to identify which service duration differences are associated with significant data feature changes. For example, when the difference in vibration values measured twice exceeds 5 mm / s, the corresponding service duration difference is recorded. After that, the mode evaluation subunit analyzes and performs a mode evaluation on the set of selected deviation modulus values of the first service duration record values by statistical methods, determines the service duration difference value with the highest occurrence frequency, and determines it as the first dissimilated service duration. For example, if in most cases, the service duration difference required for a significant change in vibration value is 1500 hours, the first dissimilated service duration is determined to be 1500 hours.

[0065] Through the above subunits, it is possible to objectively determine the dissimilated service duration of the underlying perception attribute data under specific control parameter conditions, providing an accurate basis for the subsequent calculation of the preset management cycle.

[0066] Furthermore, the first statistical unit includes:

[0067] The mapper scheduling subunit is used to, when the target dimension data is a high-level mapping attribute, schedule the target dimension data mapper bound to the high-level mapping attribute and the first preset control parameter, where the target dimension data mapper has an input of the underlying perception attribute;

[0068] The associated data retrieval subunit is used to retrieve from the historical sample library a plurality of corresponding input underlying perception attribute record feature values and a plurality of second service duration record values that satisfy the first preset control parameter;

[0069] The feature mapping processing subunit is used to respectively process the plurality of input underlying perception attribute record feature values through the target dimension data mapper to obtain a plurality of high-level mapping attribute feature values;

[0070] The dissimilated duration calculation subunit is used to calculate the first dissimilated service duration according to the plurality of high-level mapping attribute feature values and the plurality of second service duration record values.

[0071] In a preferred embodiment, when the target dimension data is a high-level mapping attribute, the first statistical unit accurately determines the first dissimilated service duration through the mapper scheduling subunit, the associated data retrieval subunit, the feature mapping processing subunit, and the dissimilated duration calculation subunit.

[0072] When the identified target dimension data is a high-level mapping attribute (such as comprehensive indicators like health index, remaining service life, etc.), the mapper scheduling subunit schedules a specific target dimension data mapper bound to the current high-level mapping attribute and the first preset control parameter. For example, when it is necessary to analyze the alienation law of the equipment health index under the condition of "rotation speed of 2000 revolutions per minute", the target dimension data mapper pre-trained for this scenario is called. Among them, the target dimension data mapper has input low-level perception attributes, that is, this mapper receives multiple low-level perception parameters (such as vibration, temperature, pressure, etc. data) as input, and is used to generate high-level comprehensive indicators. The associated data retrieval subunit is responsible for obtaining the input data required for mapping. The associated data retrieval subunit retrieves data pairs that meet the conditions of the first preset control parameter from the historical sample library, including a number of input low-level perception attribute record feature values (such as vibration values, temperature values, etc. data) and second service life record values that correspond one by one. For example, retrieve the original data such as temperature, pressure, vibration, etc. and the corresponding service life record of all equipment operating at a specific rotation speed.

[0073] The feature mapping processing subunit performs data conversion operations. The feature mapping processing subunit inputs the retrieved several low-level perception attribute record feature values into the target dimension data mapper for processing, and converts and generates several high-level mapping attribute feature values. For example, input multiple data such as temperature and vibration into the mapper to generate high-level indicators such as equipment health index values or remaining service life estimation values. The alienation duration calculation subunit conducts a systematic analysis of several high-level mapping attribute feature values and several second service life record values. Specifically, first, the alienation duration calculation subunit calculates the deviation values between any two sets of data to form a high-level mapping attribute feature value deviation modulus set and a service life record value deviation modulus set; secondly, from the service life record value deviation modulus set, extract the items where the deviation of the corresponding high-level mapping attribute feature value exceeds a predetermined threshold; then, conduct a mode evaluation on the screened service life difference data to determine the service life difference value with the highest occurrence frequency as the first alienation service life. For example, through the above method, it is calculated that the average service time required for the equipment health index to drop from 90 points to 70 points is 2000 hours.

[0074] Through the above subunits, multi-source low-level data can be mapped into high-level comprehensive indicators, and then analyze their alienation laws, which can more comprehensively evaluate the characteristics of equipment state changes and provide a more valuable reference basis for data management decisions.

[0075] Furthermore, the embodiments of the present application further include:

[0076] An attribute configuration module for configuring the preset input low-level perception attributes of the preset high-level mapping attributes through the user terminal;

[0077] A mapper training module, which is used to collect multiple groups of data with a preset target device model and preset control parameters as constraints, and train the target dimension data mapper;

[0078] A mapping binding module, which is used to bind the target dimension data mapper with the preset high-level mapping attribute and the preset control parameter;

[0079] Wherein, any one of the multiple groups of data includes: preset input low-level perception attribute record data and preset high-level mapping attribute record data.

[0080] In a preferred embodiment, the embodiment of the present application further includes an attribute configuration module, a mapper training module and a mapping binding module to construct a target dimension data mapper, so as to support data processing of high-level mapping attributes.

[0081] The attribute configuration module is used to implement user-defined mapping relationship configuration. Through the user interface, the system administrator or device engineer can configure the required preset input low-level perception attributes (such as vibration, temperature, pressure, etc.) for the preset high-level mapping attributes (such as equipment health index, remaining service life, etc.). For example, the user can specify that the "bearing health index" requires collecting three low-level perception attributes, namely "vibration value", "temperature value" and "sound characteristics", as inputs. Through the attribute configuration module, it is ensured that the system can flexibly construct mapping relationships according to the actual device characteristics and user requirements.

[0082] The mapper training module is responsible for implementing the construction of the mapper. The mapper training module first collects multiple groups of data with a preset target device model (such as a GE-T100 model gas turbine) and preset control parameters (such as operating at 75% load) as constraints. Subsequently, through machine learning algorithms such as neural networks, decision trees or support vector machines, etc., the collected multiple groups of data are subjected to model training to generate a target dimension data mapper.

[0083] The mapping binding module is used to implement the index management of the mapper. The mapping binding module binds the trained target dimension data mapper with specific preset high-level mapping attributes (such as bearing health index) and preset control parameters (such as operating at 75% load), establishes the corresponding relationship among the three, and stores this relationship in the system database. This binding mechanism ensures that the matching mapper can be quickly called during the analysis process, improving the data processing efficiency. Among them, the multiple groups of data collected for training the target dimension data mapper have a unified structure, and each group of data includes preset input low-level perception attribute record data (such as measurement data such as vibration value of 3.2 mm / s and temperature of 85 °C) and corresponding preset high-level mapping attribute record data (such as health index of 82 points).

[0084] Through the collaborative work of the above-mentioned modules, a dedicated data mapping model can be constructed according to the device characteristics and operating conditions, and high-level indicators that are difficult to directly obtain can be calculated and predicted through the underlying sensing data, improving the comprehensiveness and accuracy of device status evaluation.

[0085] Further, the threshold judgment module 13 includes:

[0086] A deviation decomposition unit, configured to include a service life deviation vector and a control parameter deviation vector in the portrait deviation vector;

[0087] A cloud retrieval unit, configured to send a retrieval request to the cloud, obtain a secret key, encrypt the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector, obtain a ciphertext, upload the ciphertext to the cloud, and obtain retrieval feedback information, where the retrieval feedback information includes an initial target dimension data deviation modulus value set;

[0088] An outlier processing unit, configured to perform outlier deletion on the initial target dimension data deviation modulus value set to obtain the target dimension data deviation modulus value set.

[0089] In a preferred embodiment, the threshold judgment module 13 includes a deviation decomposition unit, a cloud retrieval unit, and an outlier processing unit to achieve precise data deviation analysis.

[0090] The deviation decomposition unit is responsible for performing a structured decomposition process on the portrait deviation vector. The deviation decomposition unit splits the overall portrait deviation vector into two sub-vectors, namely a service life deviation vector and a control parameter deviation vector. For example, when the target device changes from the historical state to the current state, it includes a duration deviation of the service time increasing from 1000 hours to 2500 hours, and a control parameter deviation of the rotational speed changing from 2000 revolutions per minute to 2200 revolutions per minute. By decomposing the portrait deviation vector into a service life deviation vector and a control parameter deviation vector, the characteristics of the device state change can be described more precisely.

[0091] The cloud retrieval unit is used to implement an intelligent retrieval mechanism based on a distributed database. First, the cloud retrieval unit sends a retrieval request to the cloud server to obtain a secret key; then, uses the secret key to encrypt the target dimension (such as vibration value), the target device model, the service life deviation vector, and the control parameter deviation vector to generate a ciphertext; then, uploads the ciphertext to the cloud server for retrieval and matching; after that, receives the retrieval feedback information returned by the cloud, which includes an initial target dimension data deviation modulus value set. By retrieving in the cloud through the cloud retrieval unit, the sample range of data analysis is expanded, and the accuracy of deviation analysis is improved.

[0092] The outlier processing unit is responsible for data cleaning of the retrieval results. The outlier processing unit applies statistical analysis methods to the set of deviation modulus values of the initial target dimension data obtained by the cloud retrieval unit from the cloud, identifies and deletes outliers that may affect the reliability of the results, and obtains an optimized set of deviation modulus values of the target dimension data. For example, when there are obvious outliers (such as a value much higher than other values) in the retrieved vibration value deviation data, the outlier processing unit eliminates them to ensure the accuracy of subsequent variance analysis.

[0093] Through the deviation analysis mechanism based on cloud collaboration, it is possible to use a wider range of data resources for decision support, break through the limitations of single-device data analysis, and improve the accuracy and reliability of data cleaning decisions. At the same time, the encryption processing mechanism also ensures the security during the data analysis process and protects the privacy of device operation and maintenance data.

[0094] Furthermore, the cloud retrieval unit includes:

[0095] A cloud decryption subunit, configured to decrypt the ciphertext through the cloud to obtain the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector, and construct a retrieval task constraint;

[0096] A node request subunit, configured to send a retrieval request to a distributed node through the cloud to obtain a distributed node key;

[0097] An encrypted retrieval subunit, configured to encrypt the retrieval task constraint and the retrieval key according to the distributed node key, obtain a retrieval task ciphertext, send it to the distributed node, and obtain retrieval feedback information, where the retrieval feedback information includes the set of initial target dimension data deviation modulus values.

[0098] In an alternative embodiment, the cloud retrieval unit includes a cloud decryption subunit, a node request subunit, and an encrypted retrieval subunit, which are used to implement secure retrieval of distributed data resources.

[0099] The cloud decryption subunit is responsible for decrypting the ciphertext uploaded to the cloud. The cloud decryption subunit parses the ciphertext uploaded from the local computer through the decryption service of the cloud server, and recovers the original target dimension, target device model, service time deviation vector, and control parameter deviation vector. Subsequently, a standardized retrieval task constraint is constructed based on the decrypted information to prepare for the subsequent distributed retrieval. Subsequently, the node request subunit initiates a retrieval request to multiple distributed nodes (such as equipment management systems in different factories and enterprises) through the cloud platform, and obtains the encryption key returned by each distributed node. For example, the node request subunit simultaneously sends retrieval requests to equipment operation and maintenance systems in 20 different regions, and collects their respective encryption keys to prepare for the subsequent secure data exchange. After that, the encrypted retrieval subunit uses the key obtained from the distributed node to perform secondary encryption on the retrieval task constraint and the local retrieval key to generate the retrieval task ciphertext. These retrieval task ciphertexts are sent to each distributed node respectively. When the distributed node has data that meets the retrieval conditions, the corresponding retrieval feedback information will be returned. Thereafter, the encrypted retrieval subunit summarizes all the returned information to form a complete feedback containing the initial target dimension data deviation modulus value set.

[0100] Through a multi-layer encrypted distributed retrieval mechanism, it is possible to securely integrate equipment operation and maintenance experience data from multiple factories and enterprises, greatly expanding the sample base for data analysis while ensuring the security and privacy of data from all parties. This collaborative analysis model enables accurate deviation analysis based on a large amount of operating data from similar equipment, even if the historical data of a single device is limited, thus improving the accuracy and reliability of data management decisions.

[0101] Embodiment 2, the embodiment of the present invention also provides a device operation and maintenance data management method based on multi-modal data fusion, such as Figure 2 As shown, including:

[0102] When the preset management cycle is met, a retrospective portrait of the target dimension data of the target equipment number at the collection time is obtained, wherein the retrospective portrait includes the first service time and the first control parameter;

[0103] Obtaining a real-time portrait of the target device number, wherein the real-time portrait includes a second service time and a second control parameter;

[0104] constructing a portrait deviation vector according to the first service time and the first control parameter, and the second service time and the second control parameter;

[0105] Retrieve the set of target dimension data deviation modulus values that satisfy the image deviation vector. When the variance of the set of target dimension data deviation modulus values is greater than or equal to the variance threshold, calculate the ratio of the first duration between the storage start time and the acquisition time to the second duration between the storage start time and the current time, and set it as the data deletion ratio;

[0106] Delete the target dimension data from the target device tag operation and maintenance data starting from the storage start time according to the data deletion ratio.

[0107] Furthermore, when the preset management period is satisfied, obtain the retrospective image of the acquisition time of the target dimension data of the target device tag, including:

[0108] Obtain the first preset control parameter to the Nth preset control parameter of the target device tag;

[0109] With the first preset control parameter as a constraint, statistically calculate the first abnormal service duration of the target dimension data based on the historical sample library;

[0110] Until, with the Nth preset control parameter as a constraint, statistically calculate the Nth abnormal service duration of the target dimension data based on the historical sample library;

[0111] Extract 0.5 times the minimum value of the first abnormal service duration to the Nth abnormal service duration, and set it as the preset management period.

[0112] Furthermore, when the preset management period is satisfied, obtaining the retrospective image of the acquisition time of the target dimension data of the target device tag further includes:

[0113] Obtain the disk space storage capacity threshold of the target device tag operation and maintenance data;

[0114] When the amount of target device tag operation and maintenance data is greater than or equal to the disk space storage capacity threshold, obtain the retrospective image of the acquisition time of the target dimension data of the target device tag.

[0115] Furthermore, with the first preset control parameter as a constraint, statistically calculating the first abnormal service duration of the target dimension data based on the historical sample library includes:

[0116] When the target dimension data is the underlying perception attribute, retrieve from the historical sample library the corresponding several underlying perception attribute record feature values and several first service duration record values that satisfy the first preset control parameter;

[0117] Conduct pairwise deviation analysis on the corresponding several underlying perception attribute record feature values and several first service duration record values to obtain the set of underlying perception attribute record feature value deviation modulus values and the set of first service duration record value deviation modulus values;

[0118] Extract a selected deviation modulus set of the first service duration record values from the set of deviation modulus values of the first service duration record values, where the deviation modulus set of the underlying perception attribute record feature values is greater than or equal to a predefined deviation modulus threshold of the record feature values;

[0119] Perform a mode evaluation on the selected deviation modulus set of the first service duration record values to obtain the first dissimilar service duration.

[0120] Furthermore, constrained by the first preset control parameter, statistically calculate the first dissimilar service duration of the target dimension data based on the historical sample library, including:

[0121] When the target dimension data is a high-level mapped attribute, dispatch a target dimension data mapper bound to the high-level mapped attribute and the first preset control parameter, where the target dimension data mapper has input underlying perception attributes;

[0122] Retrieve a number of corresponding input underlying perception attribute record feature values and a number of second service duration record values that satisfy the first preset control parameter from the historical sample library;

[0123] Process the number of input underlying perception attribute record feature values respectively through the target dimension data mapper to obtain a number of high-level mapped attribute feature values;

[0124] Calculate the first dissimilar service duration according to the number of high-level mapped attribute feature values and the number of second service duration record values.

[0125] Furthermore, the steps for constructing the target dimension data mapper include:

[0126] Configure the preset input underlying perception attributes of the preset high-level mapped attributes through the user terminal;

[0127] Collect multiple groups of data constrained by the preset target device model and the preset control parameter to train the target dimension data mapper;

[0128] Bind the target dimension data mapper to the preset high-level mapped attribute and the preset control parameter;

[0129] Where any one of the multiple groups of data includes: preset input underlying perception attribute record data and preset high-level mapped attribute record data.

[0130] Furthermore, retrieve a set of deviation modulus values of the target dimension data that satisfies the portrait deviation vector. When the variance of the set of deviation modulus values of the target dimension data is greater than or equal to the variance threshold, delete the target dimension data from the target device position number operation and maintenance data, including:

[0131] The image deviation vector includes a service life deviation vector and a control parameter deviation vector;

[0132] Initiate a retrieval request to the cloud, obtain a secret key, encrypt the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector, obtain the ciphertext and upload it to the cloud, and obtain retrieval feedback information, where the retrieval feedback information includes a set of initial target dimension data deviation modulus values;

[0133] Perform outlier deletion on the set of initial target dimension data deviation modulus values to obtain the set of target dimension data deviation modulus values.

[0134] Further, obtaining the retrieval feedback information includes:

[0135] Through the cloud, decrypt the ciphertext to obtain the target dimension, the target device model, the service life deviation vector, and the control parameter deviation vector, and construct a retrieval task constraint;

[0136] Through the cloud, initiate a retrieval request to the distributed nodes to obtain a distributed node secret key;

[0137] Encrypt the retrieval task constraint and the retrieval key according to the distributed node secret key to obtain a retrieval task ciphertext, send it to the distributed nodes, and obtain retrieval feedback information, where the retrieval feedback information includes the set of initial target dimension data deviation modulus values.

[0138] It should be noted that in the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0139] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded computers, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0141] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0142] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0143] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concepts.

[0144] Obviously, those skilled in the art can make various changes and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. A device operation and maintenance data management system based on multimodal data fusion, characterized in that, Including: A retrospective image acquisition module, configured to obtain a retrospective image of the acquisition moment of the target dimension data of the target equipment ID when a preset management period is satisfied, where the retrospective image includes a first service life and a first control parameter; A real-time image acquisition module, configured to obtain a real-time image of the target equipment ID, where the real-time image includes a second service life and a second control parameter; An image deviation analysis module, configured to construct an image deviation vector according to the first service life and the first control parameter, and the second service life and the second control parameter; A threshold judgment module, configured to retrieve a set of target dimension data deviation modulus values that satisfy the image deviation vector. When the variance of the set of target dimension data deviation modulus values is greater than or equal to a variance threshold, calculate the ratio of a first duration between the storage start moment and the acquisition moment to a second duration between the storage start moment and the current moment, and set it as the data deletion ratio; A data cleaning module, configured to delete the target dimension data from the operation and maintenance data of the target equipment ID starting from the storage start moment according to the data deletion ratio.

2. The system according to claim 1, wherein The retrospective image acquisition module includes: A parameter acquisition unit, configured to obtain first to Nth preset control parameters of the target equipment ID; A first statistics unit, configured to statistically calculate a first abnormal service life of the target dimension data based on a historical sample library with the first preset control parameter as a constraint; A multi-parameter statistics unit, configured to statistically calculate an Nth abnormal service life of the target dimension data based on the historical sample library until the Nth preset control parameter is used as a constraint; A cycle calculation unit, configured to extract 0.5 times the minimum value of the first abnormal service life to the Nth abnormal service life, and set it as the preset management period.

3. The system according to claim 1, wherein The retrospective image acquisition module further includes: A storage threshold unit, configured to obtain a disk space storage capacity threshold of the operation and maintenance data of the target equipment ID; An image retrospective trigger unit, configured to obtain a retrospective image of the acquisition moment of the target dimension data of the target equipment ID when the amount of operation and maintenance data of the target equipment ID is greater than or equal to the disk space storage capacity threshold.

4. The system according to claim 2, wherein The first statistics unit includes: A data retrieval sub-unit, configured to, when the target dimension data is a bottom-layer perception attribute, retrieve a corresponding number of bottom-layer perception attribute record feature values and a corresponding number of first service life record values that satisfy the first preset control parameter from the historical sample library; A deviation calculation sub-unit, configured to perform pairwise deviation analysis on the corresponding number of bottom-layer perception attribute record feature values and the corresponding number of first service life record values, and obtain a bottom-layer perception attribute record feature value deviation modulus value set and a first service life record value deviation modulus value set; A feature screening sub-unit, configured to extract a first service life record value selected deviation modulus value set in which the bottom-layer perception attribute record feature value deviation modulus value set is greater than or equal to a predefined record feature value deviation modulus threshold from the first service life record value deviation modulus value set; A mode evaluation sub-unit, configured to perform mode evaluation on the first service life record value selected deviation modulus value set, and obtain the first abnormal service life.

5. The system according to claim 2, wherein The first statistical unit includes: A mapper scheduling subunit, configured to, when the target dimension data is a high-level mapping attribute, schedule a target dimension data mapper bound to the high-level mapping attribute and the first preset control parameter, where the target dimension data mapper has an input underlying perception attribute; An associated data retrieval subunit, configured to retrieve, from the historical sample library, a plurality of input underlying perception attribute record feature values and a plurality of second service duration record values that are in one-to-one correspondence and satisfy the first preset control parameter; A feature mapping processing subunit, configured to respectively process the plurality of input underlying perception attribute record feature values through the target dimension data mapper to obtain a plurality of high-level mapping attribute feature values; An alienation duration calculation subunit, configured to calculate the first alienation service duration according to the plurality of high-level mapping attribute feature values and the plurality of second service duration record values; 6. The system according to claim 5, characterized in that, The system further includes: An attribute configuration module, configured to configure a preset input underlying perception attribute of a preset high-level mapping attribute through a user terminal; A mapper training module, configured to collect multiple groups of data and train the target dimension data mapper with a preset target device model and a preset control parameter as constraints; A mapping binding module, configured to bind the target dimension data mapper to the preset high-level mapping attribute and the preset control parameter; Wherein, any one of the multiple groups of data includes: preset input underlying perception attribute record data and preset high-level mapping attribute record data.

7. The system according to claim 1, characterized in that, The threshold judgment module includes: A deviation decomposition unit, configured to the portrait deviation vector includes a service duration deviation vector and a control parameter deviation vector; A cloud retrieval unit, configured to send a retrieval request to the cloud to obtain a secret key, encrypt the target dimension, the target device model, the service duration deviation vector, and the control parameter deviation vector to obtain a ciphertext, upload the ciphertext to the cloud, and obtain retrieval feedback information, where the retrieval feedback information includes an initial target dimension data deviation modulus value set; An outlier processing unit, configured to perform outlier deletion on the initial target dimension data deviation modulus value set to obtain the target dimension data deviation modulus value set.

8. The system according to claim 7, wherein The cloud retrieval unit includes: A cloud decryption subunit, configured to decrypt the ciphertext through the cloud to obtain the target dimension, the target device model, the service duration deviation vector, and the control parameter deviation vector, and construct a retrieval task constraint; A node request subunit, configured to send a retrieval request to a distributed node through the cloud to obtain a distributed node secret key; An encrypted retrieval subunit, configured to encrypt the retrieval task constraint and a retrieval key according to the distributed node secret key to obtain a retrieval task ciphertext, send the retrieval task ciphertext to the distributed node, and obtain retrieval feedback information, where the retrieval feedback information includes the initial target dimension data deviation modulus value set.

9. A method for managing equipment operation and maintenance data based on multimodal data fusion, characterized in that, The method includes: When a preset management period is satisfied, obtain a retrospective portrait of the acquisition moment of the target dimension data of the target device number, where the retrospective portrait includes a first service duration and a first control parameter; Obtain a real-time image of the target device tag number, where the real-time image includes a second service duration and a second control parameter; Construct an image deviation vector based on the first service duration and the first control parameter, as well as the second service duration and the second control parameter; Retrieve a set of target dimension data deviation modulus values that satisfy the image deviation vector. When the variance of the set of target dimension data deviation modulus values is greater than or equal to the variance threshold, calculate the ratio of the first duration between the storage start time and the acquisition time to the second duration between the storage start time and the current time, and set it as the data deletion ratio; Delete the target dimension data from the operation and maintenance data of the target device tag number starting from the storage start time according to the data deletion ratio.

Images