On-device data collaborative analysis system and method

Abstract

The present technology discloses an on-device-based collaborative analysis system and method. According to one specific embodiment of the present technology, measurement data are learned, on the basis of a pre-built learning model, in a plurality of corresponding cloud servers assigned according to the characteristics of the measurement data, and thus learned data are outputted, and the learned data within an acceptable range, processed by the respective assigned corresponding cloud servers, are distributed to an on-device processing device, thereby enabling high-speed processing of the measurement data, and by updating the learning model by using the learned data within the acceptable range, the accuracy of the built learning model may be increased, and response to change in the measurement data may be quickly processed.

Description

Collaborative Analysis System and Method for On-Device Data

[0001] The present invention relates to a collaborative analysis system and method for on-device data, and more particularly, to a technology capable of reducing the load on on-devices and cloud servers by collecting IoT sensor data and then allocating the measurement data to a corresponding cloud server that matches the characteristics of the measurement data.

[0002] Cloud computing refers to a computing model that reduces costs by efficiently utilizing relatively large computing resources gathered in one location. Cloud service providers (CSPs) efficiently utilize their computing resources by providing them to cloud customers according to their needs.

[0003] Cloud computing allows cloud users to use as many computing resources as they want without having to build their own computing environment, and pay cloud providers based on the amount of usage.

[0004] Cloud computing is a very promising computing model. The global cloud market size reached approximately $80 billion in 2009 and is expected to grow to $109.5 billion in 2010. It is predicted that most computing models will change to the cloud computing model in the future.

[0005] However, preprocessing and analysis in cloud computing relies on cloud servers. To handle the ever-increasing volume of data, storage capacity increases, and analytical performance must also improve to handle this growing volume. Consequently, hardware costs have reached a limit, driven by the increasing volume of data.

[0006] [Prior Art Literature]

[0007] [Patent Document]

[0008] (Patent Document 1) Korean Patent Registration No. 2510650 (Published on March 20, 2023)

[0009] The technical problem to be achieved by the present invention is to provide a collaborative analysis system and method for on-device data that can reduce the load on a cloud server by collecting IoT sensor data on an on-device basis and then allocating the measurement data to a corresponding cloud server that matches the characteristics of the measurement data.

[0010] The purpose of the present invention is not limited to the aforementioned purposes, and other unmentioned purposes and advantages of the present invention can be understood through the following description and will be more clearly understood through the embodiments of the present invention. Furthermore, it will be readily apparent that the purposes and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims.

[0011] According to one embodiment of the present invention, a collaborative analysis system for on-device data according to one embodiment,

[0012] An on-device processing device that collects time series measurement data transmitted based on IoT and processes the measurement data; and

[0013] Including a cloud server that processes the received measurement data based on the size of the measurement data and the on-device resource usage rate.

[0014] The above on-device processing device is,

[0015] It is equipped to process the received time series measurement data based on the established learning model to output learning data, and to analyze specific data characteristics based on the size of the received measurement data and the on-device resource usage rate.

[0016] The above cloud server

[0017] It is characterized in that it is equipped with multiple cloud servers allocated according to each measurement data characteristic, processes measurement data allocated to the cloud server matching a specific data characteristic based on a pre-built learning model, outputs learning data, and updates the learning model with the output learning data.

[0018] Preferably, the on-device processing device is

[0019] A data collection unit that receives measurement data of the received time series; and

[0020] It may include a data learning unit that processes measurement data of the received data collection unit based on a learning model to output learning data and updates training data of the learning model with the output learning data.

[0021] Preferably, the on-device processing device is

[0022] It may further include a data analysis unit that analyzes specific data characteristics based on the size of the received measurement data and the on-device resource usage rate.

[0023] Preferably, the data learning unit

[0024] A learning module that derives learning data by learning a learning model based on the received measurement data;

[0025] A verification module that determines whether the above learning data exists within a predetermined allowable range and performs verification on the output learning data based on the determination result; and

[0026] If the above learning data is verified as being within a predetermined allowable range, an update module may be included to update the training data of the learning model with the learning data.

[0027] Preferably, the on-device processing device is

[0028] The method may further include a data communication unit that allocates measurement data to a corresponding cloud server that matches the measurement data characteristics for measurement data that is not verified successfully.

[0029] Preferably, the cloud server is:

[0030] Learning data is derived by learning based on the pre-built learning model for the received measurement data,

[0031] Determine whether the derived learning data exists within a predetermined tolerance range and perform verification on the output learning data based on the judgment result.

[0032] If the above learning data is verified as being within a predetermined allowable range, the training data of the learning model may be updated with the learning data and the learning data of the verification success may be distributed to the on-device processing device.

[0033] Preferably, the on-device processing device is

[0034] It may be provided to update the training data of the learning model with the learning data received from the cloud server.

[0035] According to another embodiment of the present invention, an on-device based collaborative analysis method is provided.

[0036] A learning step that derives learning data by learning based on a pre-built learning model for measurement data performed and received in an on-device processing unit;

[0037] A verification step that determines whether the derived learning data exists within a predetermined tolerance range and performs verification on the output learning data based on the judgment result; and

[0038] It is characterized by including a step of updating the training data of the learning model with the learning data when the verification success is that the above learning data exists within a predetermined allowable range.

[0039] Preferably, the above on-device based collaborative analysis method comprises:

[0040] A data characteristic analysis step for analyzing the measurement data characteristics based on the size of the measurement data and the on-device resource usage rate for the above-mentioned received measurement data;

[0041] If the above learning data is not verified as being within a predetermined allowable range, a data transmission step may be further included to transmit the measurement data to the corresponding cloud server allocated to match the measurement data characteristics.

[0042] Preferably, the above on-device based collaborative analysis method comprises:

[0043] A step of deriving learning data by learning based on a pre-built learning model for received measurement data, which is performed on the cloud server and for received specific data;

[0044] A step of determining whether the derived learning data exists within a predetermined tolerance range and performing verification on the output learning data based on the determination result;

[0045] A step of updating the training data of the learning model with the learning data when the verification success is that the above learning data is within a predetermined allowable range; and

[0046] It may include a step of distributing the learning data of verification success to the on-device processing device.

[0047] According to these characteristics, measurement data is learned based on a pre-built learning model from multiple cloud servers allocated according to the characteristics of the measurement data, learning data is output, and learning data within the allowable range processed by each allocated cloud server is distributed to the on-device processing device, thereby enabling high-speed processing of measurement data.

[0048] In addition, according to one embodiment, the accuracy of the constructed learning model can be improved by updating the learning model with learning data within an acceptable range, and responses to changes in measurement data can be quickly processed.

[0049] The following drawings attached to this specification illustrate preferred embodiments of the present invention, and together with the detailed description of the invention described below, serve to further understand the technical idea of ​​the present invention, and therefore, the present invention should not be interpreted as being limited to matters described in such drawings.

[0050] Figure 1 is a configuration diagram of an on-device based collaborative analysis system according to one embodiment.

[0051] Figure 2 is a detailed configuration diagram of the on-device processing unit of Figure 1.

[0052] Figure 3 is a detailed configuration diagram of the data learning unit (12) of Figure 2.

[0053] Figure 4 is a flowchart showing the operation process of the on-device-based collaborative analysis system of Figure 1.

[0054] Hereinafter, embodiments of the present invention will be described in more detail with reference to the drawings.

[0055] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. These embodiments are provided solely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined solely by the scope of the claims.

[0056] The terms used in this specification will be briefly explained, and the present invention will be described in detail.

[0057] The terms used in this invention have been selected from widely used, current terms, taking into account the functions of the invention. However, these terms may vary depending on the intentions of those skilled in the art, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, terms may be arbitrarily selected by the applicant, in which case their meanings will be described in detail in the relevant description of the invention. Therefore, the terms used in this invention should not be defined simply as names, but rather based on their inherent meanings and the overall content of the invention.

[0058] When a part of the specification is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated. Furthermore, the term "component" as used throughout the specification refers to software, hardware components such as FPGAs or ASICs, and the "component" performs certain functions. However, the "component" is not limited to software or hardware. The "component" may be configured to reside on an addressable storage medium or may be configured to execute one or more processors.

[0059] Thus, as an example, a "part" may include components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts."

[0060] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily practice them. Furthermore, in order to clearly explain the present invention, portions irrelevant to the description are omitted in the drawings.

[0061] FIG. 1 is a configuration diagram of an on-device-based collaborative analysis system according to one embodiment, FIG. 2 is a detailed configuration diagram of the on-device processing unit of FIG. 1, and FIG. 3 is a detailed configuration diagram of the data learning unit of FIG. 2.

[0062] Referring to FIGS. 1 to 3, an on-device-based collaborative analysis system according to one embodiment has a configuration that distributes collected IoT-based measurement data characteristics to a corresponding cloud server, and thus, as illustrated in FIG. 1, may include an on-device processing device (1) and a cloud server (2).

[0063] Here, IoT-based measurement data collected from sensors and various equipment represents industrial data from sectors such as machinery, robotics, electronics, automotive / aviation / shipbuilding / offshore, steel / chemical / textiles, and other core industries. Measurement data collected through processes can be obtained from data from smart factory demonstration plants and representative factories, where manufacturing industry data can be used to derive results and apply recommendations.

[0064] For example, in the smart factory manufacturing industry, measurement data may include equipment data, including total accumulated operating hours, equipment operating rate, equipment vibration values, current, voltage, power, production speed, and equipment operating position coordinates, as well as environmental data, such as workplace temperature values. In another example, in the semiconductor manufacturing process, measurement data may include input data, including chamber pressure, heater temperature, temperature rise and fall rates, gas injection amount, gas injection distance, and RF plasma power.

[0065] Meanwhile, the on-device processing device (1) and the cloud server (2) may be configured separately as shown in FIG. 1, and may be formed in a structure in which these separated components are connected through wireless or wired communication. In another example, the on-device processing device (1) and the cloud server (2) may be devices that are directly connected through wires or connectors, but are not limited thereto.

[0066] The on-device processing device (1) processes and analyzes measurement data collected from sensors, processes, and equipment, IoT-based data, such as information, photos, and videos, and the analysis results can be checked on a visual dashboard. In another example, the analysis results can be checked on a user terminal that receives the results on a web basis.

[0067] That is, the on-device processing unit (1) collects a variety of large-capacity time series measurement data that occur in units of milliseconds (ms) and nanoseconds (ns) in actual industrial sites such as smart factory semiconductor processing, analyzes the on-device resource usage rate for processing the collected measurement data of each time series, and derives measurement data characteristics based on the measurement data size and the on-device resource usage rate.

[0068] Here, the on-device resource usage rate is set to the CPU and RAM of the on-device, and the CPU and RAM capacity of the on-device set to the on-device resource usage rate can be values ​​already applied by those skilled in the art, and although the CPU and RAM capacity of the on-device applied to the on-device resource usage rate are not specifically specified in this specification, it should be understood at the level of those skilled in the art.

[0069] For example, the on-device processing device (1) can output preprocessed learning data after removing noise as a learning result of a pre-built learning model based on the derived measurement data characteristics, and as another example, the on-device processing device (1) can transmit the collected measurement data based on the measurement data characteristics to a corresponding cloud server (2-1) among the cloud servers (2) that matches the measurement data characteristics. Here, the cloud server (2) includes a predetermined number of corresponding cloud servers (2-1) to (2-3), and here, each of the corresponding cloud servers (2-1) to (2-3) is registered in the form of a lookup table and stored by matching the measurement data characteristics.

[0070] Accordingly, the on-device processing device (1) may include a data collection unit (11), a data learning unit (12), a data analysis unit (13), and a data communication unit (14), as illustrated in FIG. 2.

[0071] The data collection unit (11) collects a variety of large-capacity time series measurement data that occurs in units of milliseconds (ms) and nanoseconds (ns) in industrial sites and transmits the collected time series measurement data to the data learning unit (12).

[0072] Referring to FIG. 3, the data learning unit (12) may include a learning module (121), a verification module (122), and an update module (123).

[0073] The learning module (121) performs noise removal and preprocessing based on a pre-built learning model on the measurement data of the received time series to output learning data, and the output learning data is provided to the verification module (122).

[0074] Here, for example, the learning module (121) learns based on a pre-built learning model and performs noise removal and preprocessing on the measurement data.

[0075] Here, the learning model is constructed using statistical analysis of the measurement data and algorithms such as normal distribution and Z score, and noise removal and preprocessing can be performed on the received measurement data through learning based on the established learning model.

[0076] The verification module (122) determines whether the received learning data is within a predetermined allowable range, performs verification on the learning data, and if the verification is successful, supplies the learning data to the update module (123). Accordingly, the update module (123) updates the training data of the learning model with the generated learning data.

[0077] Although the process of constructing a learning model in this specification, performing noise removal and preprocessing on measurement data based on the constructed learning model to generate learning data, and updating the generated learning data as training data for the learning model is not specifically stated, it can be understood by those skilled in the art.

[0078] In addition, verification of noise removal and preprocessing for measurement data is determined by the allowable range for each processing item, and these allowable ranges can apply values ​​that are already being applied. Although the allowable ranges for each processing item are not specifically stated in this specification, they should be understood by those skilled in the art.

[0079] For example, in the smart factory manufacturing industry, learning data may include equipment replacement timing, equipment failure probability, and product defect rate, and in the semiconductor manufacturing process field, learning data may include, but is not limited to, wafer refractive index, wafer thickness, defect / failure prediction, and range of measurement data for yield improvement, and power consumption.

[0080] Meanwhile, the data analysis unit (13) derives measurement data characteristics based on the size of the collected various time series measurement data and the predetermined on-device resource usage rate, and the derived measurement data characteristics are transmitted to the data communication unit (14).

[0081] Accordingly, the data communication unit (14) transmits the time series measurement data exceeding the allowable range of the learning data to the corresponding cloud server (2-1) that matches the characteristics of the analyzed measurement data. At this time, the time series measurement data is transmitted in synchronization with the cloud server (2) using a time stamp generated in real time.

[0082] Meanwhile, the cloud server (2-1) performs noise removal and preprocessing on the received measurement data through learning based on a pre-built learning model for the received measurement data, and outputs learning data.

[0083] And, for example, the cloud server (2-1) determines that the verification is successful if the derived learning data is within the preset allowable range, updates the learning model with the learning data that has been verified as a result of the determination, and then updates the learning model of the data learning unit (12) with the learning data via the data communication unit (14).

[0084] Meanwhile, if the derived learning data does not exist within the above-mentioned allowable range, the cloud server (2-1) determines that the verification has failed and trains the learning model with weights set based on the error between the learning data and the actual data, thereby outputting learning data that exists within the allowable range.

[0085] Accordingly, the data learning unit (12) updates the established learning model with the received learning data. Thereafter, the data analysis unit (12) continues to perform noise removal and preprocessing on the measurement data received in time series based on the updated learning model.

[0086] FIG. 4 is a flowchart showing the operation process of the collaborative analysis system of on-device data of FIG. 1, and describes the collaborative analysis process of on-device data according to another embodiment of the present invention with reference to FIG. 3.

[0087] In one embodiment, the on-device processing device (1) outputs learning data based on the collected time series measurement data and a pre-built learning model, and verifies whether the output learning data exists within the specified allowable range (steps S11 to S13).

[0088] If the verification result of step (S13) is a success in that the output learning data is within the above allowable range, the on-device processing device (1) updates the learning model with the output learning data (step S14).

[0089] Meanwhile, if the verification is not successful in step (S13), the on-device processing device (1) derives measurement data characteristics based on the size of the measurement data and the on-device resource usage rate, and transmits specific data to the corresponding cloud server (2) allocated to match the derived measurement data characteristics (steps (S15, S16)).

[0090] The cloud server (2) learns the received specific data based on a pre-built learning model, outputs learning data, determines whether the output learning data is within a pre-determined allowable range, and if the result of the determination is a verification success that the learning data is within the allowable range, updates the learning model with the output learning data, and distributes the learning data to the on-device (S21 to S24).

[0091] In the above step (S23), if the verification success is not that the learning data is within the allowable range, the cloud server (2) trains the learning model with the weight set as the error between the derived learning data and the measured data (step S25).

[0092] And the cloud server (2) learns based on the trained learning model according to the above step (S21), outputs learning data within the output allowable range, and updates the learning model with the output learning data.

[0093] In this embodiment, measurement data is learned based on a pre-built learning model from multiple cloud servers allocated according to measurement data characteristics, learning data is output, and learning data within an allowable range processed by each allocated cloud server is distributed to an on-device processing device, thereby enabling high-speed processing of measurement data.

[0094] In addition, one embodiment can improve the accuracy of a constructed learning model by updating the learning model with learning data within an acceptable range, and can quickly process responses to changes in measurement data.

[0095] For ease of understanding, the processor is sometimes described as being used alone. However, those skilled in the art will recognize that a processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processor may include multiple processors or a single processor and a single controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0096] Here, the software may include a computer program, code, instructions or a combination of one or more of these, which may configure a processing device to perform a desired operation or may independently or collectively command a processing device.

[0097] Software and / or information, signals and data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage media or device, or transmitted signal wave for interpretation by a control unit or for providing instructions or data to a processing unit.

[0098] The software may be distributed across networked computer systems, stored and executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0099] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the medium may be those specifically designed and configured for the embodiment or may be known and usable by those skilled in the art of computer software.

[0100] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program instructions such as ROMs, RAMs, and flash memory.

[0101] Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0102] The above hardware device may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0103] Although the embodiments described above have been described by way of limited examples and drawings, those skilled in the art will appreciate that various modifications and variations can be made based on the above teachings. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0104] Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the claims described below but also by equivalents of the claims.

[0105] [Explanation of symbols]

[0106] 1: On-device processing unit

[0107] 11: Data Collection Department

[0108] 12: Data Learning Department

[0109] 121: Learning Module

[0110] 122: Verification module

[0111] 123: Renewal module

[0112] 13: Data Analysis Department

[0113] 14: Data Communication Department

Claims

1. An on-device processing device that collects time series measurement data transmitted based on IoT and processes the measurement data; and Including a cloud server that processes the received measurement data based on the size of the measurement data and the on-device resource usage rate. The above on-device processing unit, It is equipped to process the measurement data of the received time series based on the established learning model to output learning data, and analyze specific data characteristics based on the size of the received measurement data and the on-device resource usage rate. The above cloud server A collaborative analysis system based on an on-device, characterized in that it is equipped with multiple cloud servers allocated according to each measurement data characteristic, processes measurement data allocated to the cloud server matching a specific data characteristic based on a pre-built learning model, outputs learning data, and updates the learning model with the output learning data.

2. In paragraph 1, the on-device processing device, A data collection unit that receives measurement data of a received time series; and An on-device-based collaborative analysis system including a data learning unit that processes measurement data from a received data collection unit based on a learning model to output learning data and updates training data of the learning model with the output learning data.

3. In the second paragraph, the on-device processing device, An on-device based collaborative analysis system further comprising a data analysis unit that analyzes specific data characteristics based on the size of received measurement data and on-device resource usage rate.

4. In the second paragraph, the data learning unit, A learning module that derives learning data by learning a learning model based on received measurement data; A verification module that determines whether the above learning data exists within a predetermined allowable range and performs verification on the output learning data based on the determination result; and An on-device-based collaborative analysis system including an update module that updates the training data of a learning model with the learning data when the verification success is that the above learning data exists within a predetermined tolerance range.

5. In the fourth paragraph, the on-device processing device, An on-device-based collaborative analysis system further comprising a data communication unit that allocates measurement data to a corresponding cloud server that matches the measurement data characteristics for measurement data that is not verified successfully above.

6. In paragraph 5, the cloud server, Learning data is derived by learning based on the pre-built learning model for the received measurement data. Determine whether the derived learning data is within the established tolerance range and perform verification on the output learning data based on the judgment result. An on-device-based collaborative analysis system that is equipped to update the training data of a learning model with the learning data when the above learning data is verified to be within a predetermined allowable range and to distribute the learning data of the verification success to the on-device processing device.

7. In paragraph 6, the on-device processing device, An on-device-based collaborative analysis system equipped to update training data of the learning model with learning data received from the cloud server.

8. A learning step for deriving learning data by learning based on a pre-built learning model for measurement data performed and received in an on-device processing device; A verification step for determining whether the derived learning data is within a predetermined allowable range and performing verification on the output learning data based on the judgment result; and An on-device-based collaborative analysis method characterized by including a step of updating training data of a learning model with learning data when the verification success is that the learning data exists within a predetermined allowable range.

9. In the 8th paragraph, the on-device based collaborative analysis method, A data characteristic analysis step for analyzing the measurement data characteristics based on the size of the measurement data and the on-device resource usage rate for the above received measurement data; A collaborative analysis method based on an on-device, further comprising a data transmission step of transmitting measurement data to a corresponding cloud server allocated to match the measurement data characteristics if the verification success of the above learning data is not within a predetermined allowable range.

10. In the 9th paragraph, the on-device based collaborative analysis method, A step of deriving learning data by learning based on a pre-built learning model for received measurement data for specific data performed on the cloud server; A step of determining whether the derived learning data exists within a predetermined tolerance range and performing verification on the output learning data based on the determination result; If the verification success is that the above learning data is within a predetermined allowable range, a step of updating the training data of the learning model with the learning data; and An on-device based collaborative analysis method comprising a step of distributing learning data of verification success to the on-device processing device.

11. A computer-readable recording medium characterized by having recorded thereon a program for executing an on-device-based collaborative analysis method on a computer according to any one of claims 8 to 10.

12. A computer program stored on a computer-readable recording medium for executing a collaborative analysis method based on an on-device in conjunction with a computer, The above on-device based collaborative analysis method is, A learning step for deriving learning data by learning based on a pre-built learning model for measurement data performed and received in an on-device processing unit; A verification step that determines whether the derived learning data is within a predetermined tolerance range and performs verification on the output learning data based on the determination result; A step of updating the training data of the learning model with the learning data when the verification success is that the above learning data is within a predetermined tolerance range; A data characteristic analysis step that analyzes the measurement data characteristics based on the size of the measurement data and the on-device resource usage rate for the received measurement data; An operating program for an on-device-based collaborative analysis system, including a data transmission step for transmitting measurement data to a corresponding cloud server allocated to match measurement data characteristics if the verification success of the above learning data is not within a predetermined allowable range.

Images