Method for adjusting quality criteria of model mounted on edge device by using management situation data

Abstract

The present invention relates to optimizing a management environment and a quality control, by adjusting quality criteria of an abnormality detection model applied to an edge device according to a management situation. In addition, an objective of the present invention is to establish optimal quality criteria by comprehensively calculating a production disruption caused by increasing quality criteria of a model, a loss amount of a sales profit according to a missed sales opportunity due to the production disruption, and a risk of quality costs that may occur when the quality criteria is lowered.

Description

How to adjust the quality standards of models installed on edge devices using business situation data.

[0001] The present invention relates to a method for adjusting the quality criteria of a model installed in an edge device using business situation data.

[0002] The judgment criterion of an artificial intelligence model refers to the standard by which the artificial intelligence model makes a final judgment, and is called a cut-off value.

[0003] For example, in the case of vision inspection, the judgment criterion means the reference point at which the AI ​​model will judge a product as good or bad if it is more than 90% confident in judging each product as good or bad.

[0004] Basically, there is a trade-off relationship between quality standards, which are one of the judgment criteria, and product production volume.

[0005] When product production volume is more important than quality standards, it may be advantageous to increase production volume by relaxing the criteria for filtering out excessively defective products.

[0006] Additionally, in cases where quality standards are more important than product production volume, it may be advantageous to increase product quality by strengthening the criteria for filtering out defective products.

[0007] Therefore, finding the optimal balance between quality standards and product production volume is one of the important roles of managers.

[0008] In this process, genuine quality (quality that directly affects the use of the product's core functions) is an absolute quality area that should not be adjusted according to the management environment.

[0009] With the exception of genuine quality, there is a need to raise or lower the quality standards for emotional quality (quality that can vary depending on customer preferences and preferences, independent of the product's function, such as noise, vibration, smell, and color) depending on the business situation.

[0010] The quality standards for models installed in conventional edge devices did not take into account business conditions such as product sales trends, forecasted sales, forecasted inventory, and forecasted returns.

[0011] In other words, if the quality standards of the model are raised in a situation where sales demand is expected to increase, it is detrimental to management profits, and if the quality standards of the model are lowered in a situation where quality control is required, there is a problem of risk in quality costs.

[0012] The purpose of the present invention is to optimize the management environment and quality management by adjusting the quality standards of an anomaly detection model applied to an edge device according to the management situation.

[0013] The purpose of the present invention is to establish an optimal quality standard by comprehensively calculating the risk of loss of operating profit due to production disruption caused by raising the quality standard of a model, the resulting sales failure, and the quality cost that may occur when the quality standard is lowered.

[0014] An artificial intelligence system for adjusting a quality criterion of a model installed in an edge device in consideration of a business situation according to one embodiment of the present disclosure includes: a first edge device installed with a first data anomaly detection model having a first quality criterion, the first data anomaly detection model being an artificial intelligence model for determining whether a product is good or defective; a second edge device for collecting a first business situation data set and transmitting the collected first business situation data set to an edge conductor server; an edge conductor server for transmitting a request for generating the first business situation data set and a quality criterion prediction model received from the second edge device; and an artificial intelligence conductor server for generating a quality criterion prediction model that infers a quality criterion value of a data anomaly detection model based on the first business situation data set, obtaining a second quality criterion value inferred by the quality criterion prediction model, and transmitting the obtained second quality criterion value to the edge conductor server, wherein the edge conductor server transmits the second quality criterion value to the first edge device, and the first edge device can update the first data anomaly detection model to a second data anomaly detection model based on the second quality criterion value.

[0015] According to an embodiment of the present invention, a method for operating an artificial intelligence system for adjusting a quality criterion of a model installed in an edge device in consideration of a business situation comprises the steps of: installing, by a first edge device, a first data anomaly detection model having a first quality criterion, wherein the first data anomaly detection model is an artificial intelligence model that determines whether a product is good or defective; collecting, by a second edge device, a first business situation data set and transmitting the collected first business situation data set to an edge conductor server; transmitting, by the edge conductor server, a request for generating the first business situation data set and a quality criterion prediction model received from the second edge device; generating, by the artificial intelligence conductor server, a quality criterion prediction model that infers a quality criterion value of a data anomaly detection model based on the first business situation data set; obtaining, by the artificial intelligence conductor server, a second quality criterion value inferred by the quality criterion prediction model; transmitting, by the artificial intelligence conductor server, the obtained second quality criterion value to the edge conductor server; transmitting, by the edge conductor server, the second quality criterion value to the first edge device; And the first edge device may include a step of updating the first data anomaly detection model to a second data anomaly detection model based on the second quality criterion value.

[0016] According to an embodiment of the present invention, by adjusting the quality standards of a model applied to an edge device according to a business situation, business profits and quality management can be optimized.

[0017] According to an embodiment of the present invention, the quality standards of a model applied to an edge device are optimized, thereby preventing a decrease in operating profit due to actual sales and reducing the risk of quality costs.

[0018] According to an embodiment of the present invention, even if the prediction of the quality standard fails and the inventory amount increases, flexible response is possible as the quality standard is readjusted.

[0019] Figure 1 illustrates an AI device according to one embodiment of the present disclosure.

[0020] FIG. 2 illustrates an AI server according to one embodiment of the present disclosure.

[0021] FIG. 3 is a diagram for explaining the configuration of an AI advisor system according to one embodiment of the present disclosure.

[0022] FIG. 4 is a flowchart illustrating an operation method of an AI conductor according to an embodiment of the present disclosure.

[0023] FIG. 5 is a drawing illustrating a specific configuration of an artificial intelligence system according to an embodiment of the present invention.

[0024] FIG. 6 is a diagram illustrating a process of measuring the prediction accuracy of each of a plurality of models learned by each of a plurality of algorithms according to one embodiment of the present disclosure.

[0025] FIG. 7 is a drawing showing an execution screen of an edge UI application according to one embodiment of the present invention.

[0026] FIG. 8 is a ladder diagram for explaining an operation method of an artificial intelligence system according to one embodiment of the present invention.

[0027] FIG. 9a is a diagram illustrating a process of adjusting a quality criterion value of a data anomaly detection model installed in a first edge device based on a first management situation data set.

[0028] FIG. 9b is a flowchart illustrating a process of inferring a second quality criterion value from a first management situation data set using the quality criterion prediction model of the present disclosure.

[0029] Figures 10 and 11 are drawings explaining a process for updating a data anomaly detection model mounted on a first edge device.

[0030] FIG. 11 is a diagram illustrating a re-learning process of a data anomaly detection model according to an embodiment of the present disclosure.

[0031] Figure 12 is a diagram illustrating a process of updating the quality criteria values ​​of a data anomaly detection model installed on an edge device according to re-collection of a business situation data set.

[0032] FIG. 1 illustrates an AI device (100) according to one embodiment of the present disclosure.

[0033] The AI ​​device (100) can be implemented as a fixed device or a movable device, such as a TV, a projector, a mobile phone, a smart phone, a desktop computer, a laptop, an edge device for digital broadcasting, a PDA (personal digital assistant), a PMP (portable multimedia player), a navigation device, a tablet PC, a wearable device, a set-top box (STB), a DMB receiver, a radio, a washing machine, a refrigerator, a desktop computer, digital signage, a robot, a vehicle, etc.

[0034] Referring to FIG. 1, an edge device (100) may include a communication unit (110), an input unit (120), a learning processor (130), a sensing unit (140), an output unit (150), a memory (170), and a processor (180).

[0035] The communication unit (110) can transmit and receive data with external devices such as other AI devices (100a to 100e) or AI servers (200) using wired or wireless communication technology. For example, the communication unit (110) can transmit and receive sensor information, user input, learning models, control signals, etc. with external devices.

[0036] The communication unit (110) may be named a communication interface.

[0037] The communication technologies used by the communication unit (110) include GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), LTE (Long Term Evolution), 5G, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Bluetooth (Bluetooth™), RFID (Radio Frequency Identification), Infrared Data Association (IrDA), ZigBee, NFC (Near Field Communication), etc.

[0038] The input unit (120) can obtain various types of data.

[0039] The input unit (120) may include a camera for inputting a video signal, a microphone for receiving an audio signal, a user input unit for receiving information from a user, etc. Here, the camera or microphone may be treated as a sensor, and a signal obtained from the camera or microphone may be referred to as sensing data or sensor information.

[0040] The input unit (120) can obtain input data to be used when obtaining output using learning data and learning models for model learning. The input unit (120) can also obtain unprocessed input data, in which case the processor (180) or learning processor (130) can extract input features as preprocessing for the input data.

[0041] The learning processor (130) can train a model composed of an artificial neural network using learning data. Here, the trained artificial neural network may be referred to as a learning model. The learning model can be used to infer result values ​​for new input data other than the learning data, and the inferred values ​​can be used as a basis for making decisions regarding certain actions.

[0042] The running processor (130) can perform AI processing together with the running processor (240) of the AI ​​server (200).

[0043] The running processor (130) may include memory integrated or implemented in the AI ​​device (100). Alternatively, the running processor (130) may be implemented using memory (170), external memory directly coupled to the AI ​​device (100), or memory maintained in an external device.

[0044] The sensing unit (140) can obtain at least one of internal information of the AI ​​device (100), information about the surrounding environment of the AI ​​device (100), and user information using various sensors.

[0045] Sensors included in the sensing unit (140) include a proximity sensor, a light sensor, an acceleration sensor, a magnetic sensor, a gyro sensor, an inertial sensor, an RGB sensor, an IR sensor, a fingerprint recognition sensor, an ultrasonic sensor, a light sensor, a microphone, a lidar, a radar, etc.

[0046] The output unit (150) can generate output related to vision, hearing, or touch.

[0047] The output unit (150) may include a display unit that outputs visual information, a speaker that outputs auditory information, a haptic module that outputs tactile information, etc.

[0048] The memory (170) can store data that supports various functions of the AI ​​device (100). For example, the memory (170) can store input data, learning data, learning models, learning history, etc. obtained from the input unit (120).

[0049] The processor (180) may determine at least one executable operation of the AI ​​device (100) based on information determined or generated using a data analysis algorithm or a machine learning algorithm. Then, the processor (180) may control components of the AI ​​device (100) to perform the determined operation.

[0050] To this end, the processor (180) can request, retrieve, receive or utilize data from the running processor (130) or memory (170), and control components of the AI ​​device (100) to execute at least one of the executable operations, a predicted operation or an operation determined to be desirable.

[0051] When the processor (180) requires connection to an external device to perform a determined operation, it can generate a control signal for controlling the external device and transmit the generated control signal to the external device.

[0052] The processor (180) can obtain intention information for user input and determine the user's requirement based on the obtained intention information.

[0053] The processor (180) can obtain intent information corresponding to the user input by using at least one of a STT (Speech To Text) engine for converting voice input into a string or a natural language processing (NLP) engine for obtaining intent information of natural language.

[0054] At least one of the STT engine or the NLP engine may be configured with an artificial neural network, at least in part, trained according to a machine learning algorithm. Furthermore, at least one of the STT engine or the NLP engine may be trained by the learning processor (130), the learning processor (240) of the AI ​​server (200), or through distributed processing thereof.

[0055] The processor (180) can collect history information including the operation details of the AI ​​device (100) or the user's feedback on the operation, and store the information in the memory (170) or the learning processor (130), or transmit the information to an external device such as an AI server (200). The collected history information can be used to update the learning model.

[0056] The processor (180) can control at least some of the components of the AI ​​device (100) to drive an application program stored in the memory (170). Furthermore, the processor (180) can operate two or more of the components included in the AI ​​device (100) in combination to drive the application program.

[0057] FIG. 2 illustrates an AI server according to one embodiment of the present disclosure.

[0058] Referring to FIG. 2, the AI ​​server (200) may refer to a device that trains an artificial neural network using a machine learning algorithm or utilizes a trained artificial neural network. Here, the AI ​​server (200) may be composed of multiple servers to perform distributed processing, and may be defined as a 5G network. The AI ​​server (200) may be included as part of the AI ​​device (100) and may perform at least a portion of the AI ​​processing.

[0059] The AI ​​server (200) may include a communication unit (210), memory (230), a learning processor (240), and a processor (260).

[0060] The communication unit (210) can transmit and receive data with an external device such as an AI device (100).

[0061] The memory (230) may include a model storage unit (231). The model storage unit (231) may store a model (or artificial neural network, 231a) being learned or learned through the learning processor (240).

[0062] A learning processor (240) can train an artificial neural network (231a) using learning data. The learning model can be used while mounted on the AI ​​server (200) of the artificial neural network, or can be mounted on an external device such as an AI device (100).

[0063] The learning model may be implemented in hardware, software, or a combination of hardware and software. If part or all of the learning model is implemented in software, one or more instructions constituting the learning model may be stored in memory (230).

[0064] The processor (260) can use a learning model to infer a result value for new input data and generate a response or control command based on the inferred result value.

[0065] FIG. 3 is a diagram for explaining the configuration of an artificial intelligence system according to one embodiment of the present disclosure.

[0066] Referring to FIG. 3, the AI ​​advisor system (30) may include a plurality of edge devices (100-1 embedded 100-n), an edge conductor server (200-1), and an AI conductor server (200-2).

[0067] In FIG. 3, the edge conductor server (200-1) or AI conductor server (200-2) is illustrated as one, but may be configured as two or more.

[0068] As the edge conductor server (200-1) is connected to multiple edge devices, an overall pyramid structure can be formed.

[0069] Each of the plurality of edge devices (100-1 to 100-n) may be an example of the AI ​​device (100) of FIG. 1. Each of the plurality of edge devices (100-1 to 100-n) may include all of the components of the AI ​​device (100) of FIG. 1.

[0070] Each of the plurality of edge devices (100-1 to 100-n) may be an inspection device placed on a production line at a manufacturing site. Each of the plurality of edge devices (100-1 to 100-n) may be a device for acquiring sensing data during the manufacturing process of a single product or component.

[0071] The sensing data may include one or more of vibration data detected by a vibration sensor and noise data detected by a noise sensor.

[0072] Each of the multiple edge devices (100-1 to 100-n) may have an edge application installed. The edge application can collect data suitable for creating a learning model and cleanse the collected data. This can improve data quality.

[0073] An edge application can transmit device information of an edge device, such as an edge device or inspection PC, to the edge conductor server (200-1). The device information of the edge device may include one or more of hardware specification information (CPU, memory capacity, storage capacity) and operating system information.

[0074] The edge conductor server (200-1) can receive multiple data sets and multiple device information from multiple edge devices.

[0075] The edge conductor server (200-1) can generate a plurality of modeling codes based on each of a plurality of data sets, and transmit the generated plurality of modeling codes and the plurality of data sets to the AI ​​conductor server (200-2).

[0076] The edge conductor server (200-1) can store Docker images in a Docker registry server from which they can be downloaded, or store Docker images itself.

[0077] The edge conductor server (200-1) can secure a Docker image that matches the hardware specification information of the edge device and transfer the Docker image to the edge device.

[0078] The Docker Registry server can be included in the edge conductor server (200-1).

[0079] The edge conductor server (200-1) can communicate with multiple edge devices (100-1 to 100-n) via a short-range wireless communication standard or Internet communication.

[0080] Each of the plurality of edge devices (100-1 to 100-n) can collect one or more of structured data or unstructured data.

[0081] Each of the multiple edge devices (100-1 to 100-n) may have one or more sensors connected.

[0082] Each of the plurality of edge devices (100-1 to 100-n) can receive one or more of structured data or unstructured data from a connected sensor and transmit the received data to the edge conductor server (200-1).

[0083] The edge conductor server (200-1) can transmit structured data and unstructured data received from each edge device to the AI ​​conductor server (200-2).

[0084] As another example, the edge conductor server (200-1) can generate transformed data by converting unstructured data into structured data. For example, the edge conductor server (200-1) can convert an acoustic signal into a power spectrum in the frequency domain through a Fourier transform.

[0085] In another embodiment, a data conversion unit for converting unstructured data into structured data may be provided in the AI ​​conductor server (200-2) instead of the edge conductor server (200-1).

[0086] The edge conductor server (200-1) can create a Docker image suitable for each edge device to support various types of hardware and operating systems of the edge device.

[0087] The edge conductor server (200-1) can generate a docker image that can be run on each edge device based on device information received from each of the plurality of edge devices (100-1 to 100-n).

[0088] That is, the edge conductor server (200-1) can generate a Docker image list including a plurality of Docker images corresponding to a plurality of edge devices (100-1 to 100-n).

[0089] The edge conductor server (200-1) and the AI ​​conductor server (200-2) can perform wireless communication.

[0090] The AI ​​conductor server (200-2) can generate a plurality of different models corresponding to each of a plurality of edge devices based on a plurality of modeling codes received from the edge conductor server (200-1).

[0091] The AI ​​conductor server (200-2) can transmit multiple generated models to the edge conductor server (200-1).

[0092] The edge conductor server (200-1) can transmit each of the plurality of models received from the AI ​​conductor server (200-2) to each of the plurality of edge devices (100-1 to 100-n).

[0093] Each of the plurality of edge devices (100-1 to 100-n) may be equipped with a different model received from the edge conductor server (200-1).

[0094] According to an embodiment of the present invention, a plurality of streams corresponding to each of a plurality of edge devices (100-1 to 100-n) may be constructed. That is, there may be a number of streams for each number of edges.

[0095] Each stream can be a series of processes for learning a base model and loading it onto an edge device through a base model applied to each edge device and parameters used in the base model.

[0096] If the base model applied to the edge device changes, the stream can be rebuilt with the changed base model and parameters.

[0097] Once the stream is set up, relabeling can be performed on the data collected by the edge device, and the base model can be retrained using the data and relabeled data.

[0098] FIG. 4 is a ladder diagram for explaining the operation method of an AI advisor system that generates and applies a model suitable for each edge device in a production site according to one embodiment of the present invention.

[0099] The model described below may be, but need not be limited to, an artificial neural network-based judgment model that determines whether a product or part is good or bad, learned using a deep learning or machine learning algorithm.

[0100] Referring to FIG. 4, the first edge device (100-1) can transmit the first device information and the first data set to the edge conductor server (200-1) (S401).

[0101] The first device information may include one or more of the operating system, memory capacity, CPU, and hardware information of the first edge device (100-1).

[0102] The first data set may be data collected by the first edge device (100-1).

[0103] In one embodiment, the processor (180) of the first edge device (100-1) can collect sensing data through the sensing unit (140).

[0104] In another embodiment, the processor (180) of the first edge device (100-1) can collect sensing data through a sensor that communicates with the first edge device (100-1).

[0105] The processor (180) of the first edge device (100-1) can transmit the first data set, which is sensing data collected through the communication unit (210), to the edge conductor server (200-1).

[0106] The first data set may include one or more of structured data or unstructured data.

[0107] Structured data can be a type of data that has a label that can determine pass (or OK) or fail (or NG), and unstructured data can be an unstructured type of data that cannot determine pass or fail.

[0108] Structured data can be a type of data that a person can visually judge as good or bad, and unstructured data can be a type of data, such as sound data, that a person cannot visually see.

[0109] Structured data can be either image data or sensor sensing value data.

[0110] Unstructured data can be audio data.

[0111] The edge conductor server (200-1) can generate a first modeling code to be used for modeling the model based on the received first data set (S403).

[0112] The first modeling code may be a code required for modeling a model to be mounted on the first edge device (100-1). The first modeling code may be a code indicating a base model to be applied to the first edge device (100-1) and parameters used in the base model.

[0113] The first modeling code may include multiple codes corresponding to each of the multiple algorithms. Each of the multiple algorithms may be used to train multiple models in parallel.

[0114] The processor (260) of the edge conductor server (200-1) can obtain a model suitable for the type of the first data set and obtain a first modeling code for modeling the obtained model.

[0115] The memory (230) of the edge conductor server (200-1) may store modeling codes corresponding to each of a plurality of data types.

[0116] The processor (260) of the edge conductor server (200-1) can obtain a first modeling code matching the first data set type from the memory (230).

[0117] The edge conductor server (200-1) can transmit the first data set and the generated first modeling code to the AI ​​conductor server (200-2) (S405).

[0118] The processor (260) of the edge conductor server (200-1) can transmit the first data set and the first modeling code to the AI ​​conductor server (200-2) via the communication unit (210).

[0119] The AI ​​conductor server (200-2) can select an available server by considering resources upon receiving the first data set and the first modeling code (S407).

[0120] The AI ​​Conductor server (200-2) may include multiple modeling servers. Each of the multiple modeling servers may generate multiple models in parallel. The multiple modeling servers may be provided separately from the AI ​​Conductor server (200-2).

[0121] The processor (260) of the AI ​​conductor server (200-2) can determine an available server capable of performing modeling by considering the resources of each of the multiple modeling servers.

[0122] The processor (260) of the AI ​​conductor server (200-2) can determine an available server by considering one or more resources of the memory capacity and CPU performance of each modeling server.

[0123] The AI ​​conductor server (200-2) can create multiple models using the first data set and the first modeling code through the selected available server (S409).

[0124] The processor (260) of the AI ​​conductor server (200-2) can transmit a control command to a determined available server to train multiple models using the first data set and the first modeling code. The available server can train multiple models in parallel using the first data set and the first modeling code according to the received control command.

[0125] Each of the multiple models may be a model that aims to infer as accurately as possible whether a product is good or bad by an edge device operating in the field.

[0126] Each of the multiple models may be an artificial intelligence-based model that receives one or more of the transformed data corresponding to the structured data or unstructured data for the product as input and infers whether the product is good or bad.

[0127] An available server can train multiple models in parallel using the same primary data set. Each of the multiple models can be trained using different algorithms or the same algorithm.

[0128] The AI ​​conductor server (200-2) can select the first model with the optimal performance among the multiple generated models (S411).

[0129] The processor (260) of the AI ​​conductor server (200-2) may select the model with the highest prediction accuracy among the multiple learned models as the first model with optimal performance. If the model is a judgment model, the prediction accuracy may indicate the rate at which a good or defective product is accurately classified.

[0130] The AI ​​conductor server (200-2) can transmit the selected first model to the edge conductor server (200-1) (S415).

[0131] The processor (260) of the AI ​​conductor server (200-2) can transmit the selected first model to the edge conductor server (200-1) through the communication unit (210).

[0132] The processor (260) of the AI ​​conductor server (200-2) can transmit the selected first model to the edge conductor server (200-1) to distribute the first model to the first edge device (100-1).

[0133] The processor (260) of the AI ​​conductor server (200-2) can transmit one or more of the parameters constituting the selected first model and the code for driving the first model to the edge conductor server (200-1) through the communication unit (210).

[0134] The edge conductor server (200-1) can generate a first docker image based on the received first model and first device information (S415).

[0135] The processor (260) of the edge conductor server (200-1) can generate a first docker image based on the first device information so that the first model can operate normally on the first edge device (100-1).

[0136] The first Docker image may include a software package for independent execution of edge applications mounted on edge devices.

[0137] The first Docker image can contain the files required to run the edge application, including the edge application's code, runtime, system tools, and libraries.

[0138] The edge conductor server (200-1) can transmit the first model and the first docker image to the first edge device (100-1) (S417).

[0139] The processor (260) of the edge conductor server (200-1) can transmit the first model and the first docker image to the first edge device (100-1) via the communication unit (210) to load the first model onto the first edge device (100-1).

[0140] The first edge device (100-1) can load the first model onto the first edge device (100-1) based on the received first docker image (S419).

[0141] The processor (180) of the first edge device (100-1) may be equipped with a first model optimized for the first edge device (100-1) based on the first docker image. The processor (180) of the first edge device (100-1) may control the execution of the first model through an edge application.

[0142] The edge application installed on the first edge device (100-1) can infer the judgment result of a product or component through the first model. The edge application installed on the first edge device (100-1) can transmit the inferred judgment result to the edge conductor server (200-1). The edge conductor server (200-1) can transmit the judgment result to the AI ​​conductor server (200-2).

[0143] The second edge device (100-2) can transmit second device information and a second data set to the edge conductor server (200-1) (S421).

[0144] The second device information may include one or more of the operating system, memory capacity, CPU, and hardware information of the second edge device (100-2).

[0145] The second data set may be data collected by the second edge device (100-2).

[0146] In one embodiment, the processor (180) of the second edge device (100-2) can collect sensing data through the sensing unit (140).

[0147] In another embodiment, the processor (180) of the second edge device (100-2) can collect sensing data through a sensor that communicates with the second edge device (100-2).

[0148] The processor (180) of the second edge device (100-2) can transmit the second data set, which is sensing data collected through the communication unit (210), to the edge conductor server (200-2).

[0149] The second data set may include one or more of structured data or unstructured data.

[0150] The edge conductor server (200-1) can generate a second modeling code to be used for modeling the model based on the received second data set (S423).

[0151] The second modeling code may be a code required for modeling a model to be mounted on the second edge device (100-2). The second modeling code may be a code indicating a base model to be applied to the second edge device (100-2) and parameters used in the base model.

[0152] The second modeling code may include multiple codes corresponding to each of the multiple algorithms. Each of the multiple algorithms may be used to train multiple models in parallel.

[0153] The processor (260) of the edge conductor server (200-1) can obtain a model suitable for the type of the second data set and obtain a second modeling code for modeling the obtained model.

[0154] The memory (230) of the edge conductor server (200-1) may store modeling codes corresponding to each of a plurality of data types.

[0155] The processor (260) of the edge conductor server (200-1) can obtain a second modeling code matching the second data set type from the memory (230).

[0156] The first modeling code and the second modeling code may be the same or different.

[0157] The edge conductor server (200-1) can transmit the second data set and the generated second modeling code to the AI ​​conductor server (200-2) (S425).

[0158] The processor (260) of the edge conductor server (200-1) can transmit the second data set and the second modeling code to the AI ​​conductor server (200-2) through the communication unit (210).

[0159] The AI ​​conductor server (200-2) can select an available server by considering resources upon receiving the second data set and the second modeling code (S427).

[0160] The AI ​​Conductor server (200-2) may include multiple modeling servers. Each of the multiple modeling servers may generate multiple models in parallel. The multiple modeling servers may be provided separately from the AI ​​Conductor server (200-2).

[0161] The processor (260) of the AI ​​conductor server (200-2) can determine an available server capable of performing modeling by considering the resources of each of the multiple modeling servers.

[0162] The processor (260) of the AI ​​conductor server (200-2) can determine an available server by considering one or more resources of the memory capacity and CPU performance of each modeling server.

[0163] The AI ​​conductor server (200-2) can create multiple models using the second data set and the second modeling code through the selected available server (S429).

[0164] The processor (260) of the AI ​​conductor server (200-2) can transmit a control command to a determined available server to train multiple models using a second data set and a second modeling code. The available server can train multiple models in parallel using the second data set and the second modeling code according to the received control command.

[0165] Each of the multiple models may be a model that aims to infer as accurately as possible whether a product is good or bad by an edge device operating in the field.

[0166] Each of the multiple models may be an artificial intelligence-based model that receives one or more of the transformed data corresponding to the structured data or unstructured data for the product as input and infers whether the product is good or bad.

[0167] An available server can train multiple models in parallel using the same second data set. Each of the multiple models can be trained using different algorithms or the same algorithm.

[0168] The AI ​​conductor server (200-2) can select a second model with optimal performance among the multiple generated models (S431).

[0169] The processor (260) of the AI ​​conductor server (200-2) may select the model with the highest prediction accuracy among the multiple trained models as the second model with optimal performance. If the model is a judgment model, the prediction accuracy may indicate the rate at which a good or defective product is accurately classified.

[0170] The AI ​​conductor server (200-2) can transmit the selected second model to the edge conductor server (200-1) (S435).

[0171] The processor (260) of the AI ​​conductor server (200-2) can transmit the selected second model to the edge conductor server (200-1) through the communication unit (210).

[0172] The processor (260) of the AI ​​conductor server (200-2) can transmit the selected second model to the edge conductor server (200-1) to distribute the second model to the second edge device (100-2).

[0173] The processor (260) of the AI ​​conductor server (200-2) can transmit one or more of the parameters constituting the selected second model and the code for driving the second model to the edge conductor server (200-1) through the communication unit (210).

[0174] The edge conductor server (200-1) can generate a second docker image based on the received second model and second device information (S437).

[0175] The processor (260) of the edge conductor server (200-1) can generate a second docker image based on the second device information so that the second model can operate normally on the second edge device (100-2).

[0176] The second Docker image may include a software package for independent execution of edge applications mounted on edge devices.

[0177] The second Docker image can contain the files required to run the edge application, including the edge application's code, runtime, system tools, and libraries.

[0178] The edge conductor server (200-1) can transmit the second model and the second docker image to the second edge device (100-2) (S437).

[0179] The processor (260) of the edge conductor server (200-1) can transmit the second model and the second docker image to the second edge device (100-1) via the communication unit (210) to load the second model onto the second edge device (100-2).

[0180] The second edge device (100-2) can load the second model onto the second edge device (100-2) based on the received second docker image (S439).

[0181] The processor (180) of the second edge device (100-2) may be equipped with a second model optimized for the second edge device (100-1) based on the second docker image. The processor (180) of the second edge device (100-1) may control the execution of the second model through an edge application.

[0182] The edge application installed on the second edge device (100-2) can infer the judgment result of a product or component through the second model. The edge application installed on the second edge device (100-2) can transmit the inferred judgment result to the edge conductor server (200-1). The edge conductor server (200-1) can transmit the judgment result to the AI ​​conductor server (200-2).

[0183] In this way, according to embodiments of the present invention, an optimized model tailored to each of a plurality of edge devices can be distributed. This significantly improves the accuracy of determining whether each edge device is a good or bad product.

[0184] FIG. 5a is a drawing illustrating a specific configuration of an AI advisor system having a three-stage structure according to an embodiment of the present invention, and FIG. 5b is a drawing illustrating a configuration of an available server that learns multiple models in parallel according to an embodiment of the present invention.

[0185] Referring to FIG. 5, the AI ​​conductor server (200-2) may include a wireless communication unit (310), a plurality of available servers (320-1 to 320-n), an optimal model verification unit (330), a model selection unit (350), a display (370), and a processor (390).

[0186] The AI ​​conductor server (200-2) may be an example of the AI ​​server (200) of FIG. 2. In this case, the wireless communication unit (310) of FIG. 5 may be an example of the communication unit (210) of FIG. 2, and the processor (390) of FIG. 5 may have the same configuration as the processor (260) of FIG. 2.

[0187] The wireless communication unit (310) can perform wireless communication with the edge conductor server (200-1). The wireless communication unit (310) can receive structured data and unstructured data from the edge conductor server (200-1).

[0188] The wireless communication unit (310) may also receive converted data, in which unstructured data is converted into structured data, from the edge conductor server (200-1).

[0189] A data conversion unit (not shown) can create converted data by assigning labels to unstructured data.

[0190] The data transformation unit can convert structured data into a graph, assign a label, and combine the label with the structured data to create transformed data.

[0191] The data transformation unit can generate transformation data for input data required for model learning.

[0192] Each of the multiple available servers (320-1 to 320-n) can train multiple models in parallel using structured data and transformed data.

[0193] Each of the multiple models may be a judgment model that aims to infer as accurately as possible whether a product is good or bad by an edge device operating in the field.

[0194] Each of the multiple models may be a model that receives one or more of the transformed data corresponding to the structured data or unstructured data for the product as input, and infers whether the product is good or bad.

[0195] Each available server can train four models in parallel using the same data. Each of the four models can be trained using different learning methods.

[0196] Referring to FIG. 5b, each available server (320) may include a first model learning unit (321), a second model learning unit (322), a third model learning unit (323), and a fourth model learning unit (324).

[0197] In Fig. 5b, it is assumed that each available server (320) includes four model learning units, but this is merely an example and may include a greater number of model learning units.

[0198] The first model learning unit (321) may include a first preprocessing unit (321-1), a first data set processing unit (321-2), a first training unit (321-3), and a first model generation unit (321-4).

[0199] The first preprocessing unit (321-1) can preprocess the first data set received from the first edge device (100-1). The first preprocessing unit (321-1) can remove unusable data from the first data set.

[0200] The first data set processing unit (321-2) can calculate a data processing value including at least one of the average value, median value, and maximum value of the data from which useless data has been removed.

[0201] The first training unit (321-3) can learn the first judgment model using the first learning algorithm based on the data processing value.

[0202] The first learning algorithm may be a random forest classification algorithm. Random forest classification may be a technique that generates multiple decision trees and combines the prediction values ​​of each of the generated decision trees to generate a final prediction value.

[0203] The first model generation unit (321-4) can generate the first judgment model by determining the parameters of the first judgment model learned through the first training unit (321-3).

[0204] The second model learning unit (322) may include a second preprocessing unit (322-1), a second data set processing unit (322-2), a second training unit (322-3), and a second model generation unit (322-4).

[0205] The second preprocessing unit (322-1) can preprocess a first data set including structured data and transformed data. The second preprocessing unit (322-1) can remove unusable data from the first data set.

[0206] The second data set processing unit (322-2) can calculate a data processing value including at least one of the average value, median value, and maximum value of the data from which useless data has been removed.

[0207] The second training unit (322-3) can learn the second judgment model using the second learning algorithm based on the data processing value.

[0208] The second learning algorithm may be the Categorical Boosting (CAT Boost) classification algorithm. The CAT Boost classification algorithm may be an algorithm that classifies data by processing categorical features based on gradient descent.

[0209] The second model generation unit (322-4) can generate a second judgment model by determining the parameters of the second judgment model learned through the second training unit (322-3).

[0210] The third model learning unit (323) may include a third preprocessing unit (323-1), a third data processing unit (323-2), a third training unit (323-3), and a third model generation unit (323-4).

[0211] The third preprocessing unit (323-1) can preprocess a first data set including structured data and transformed data. The third preprocessing unit (323-1) can remove unusable data from the first data set.

[0212] The third data processing unit (323-2) can calculate a data processing value including at least one of the average value, median value, and maximum value of the data from a data set from which useless data has been removed.

[0213] The third training unit (323-3) can learn the third model using the third learning algorithm based on the data processing value.

[0214] The third learning algorithm may be the Gradient Boosting Classification algorithm. Gradient Boosting Classification uses gradient descent to adjust the weights of multiple learners to minimize prediction errors, and allows the next learner to learn from the error between the previous learner's prediction and the actual value.

[0215] The third model generation unit (323-4) can generate a third judgment model by determining the parameters of the third judgment model learned through the third training unit (323-3).

[0216] The fourth model learning unit (324) may include a fourth preprocessing unit (324-1), a fourth data processing unit (324-2), a fourth training unit (324-3), and a fourth model generation unit (324-4).

[0217] The fourth preprocessing unit (324-1) can preprocess the first data set. The fourth preprocessing unit (324-1) can remove unusable data from the structured data and converted data.

[0218] The fourth data processing unit (324-2) can calculate a data processing value including at least one of the average value, median value, and maximum value of the data from which useless data has been removed.

[0219] The fourth training unit (324-3) can learn the fourth judgment model using the fourth learning algorithm based on the data processing value.

[0220] The fourth learning algorithm may be the NGB Classification algorithm. The NGB Classification algorithm uses the assumption of independence between features, a fundamental assumption of Naive Bayes, to estimate the probability distribution of data and classify the data based on this.

[0221] The fourth model generation unit (324-4) can generate the fourth judgment model by determining the parameters of the fourth judgment model learned through the fourth training unit (324-3).

[0222] Each training unit can measure the weight values ​​of each intermediate layer among the multiple layers that make up the model and verify the measured weight values.

[0223] Each training unit can stop model learning if the validation score of each weight value of the intermediate layers is less than the reference score.

[0224] That is, each training unit can stop learning or adjust the weight values ​​if there are weight values ​​in the intermediate layers that have an effect of lowering the model's prediction accuracy.

[0225] Each training unit can adjust the weight values ​​of intermediate layers, which have the effect of lowering the prediction accuracy as each layer is stacked.

[0226] The optimal model verification unit (330) can measure the accuracy of each of the multiple judgment models.

[0227] The optimal model verification unit (330) can measure the prediction accuracy of each of multiple models using the N-fold cross validation method.

[0228] The N-fold cross validation method can be a method of dividing a given data set into N parts, using one part as validation data, and the remaining N-1 parts as training data, repeating N times so that each part is used as validation data once.

[0229] The model selection unit (350) can select the judgment model with the highest accuracy among multiple judgment models as the first model to be transmitted to the first edge device (100-1).

[0230] Through the above process, the AI ​​conductor server (200-2) can generate multiple judgment models based on the second data set received from the second edge device (100-2), and select the judgment model with the highest prediction accuracy among the generated multiple judgment models as the second model.

[0231] The AI ​​conductor server (200-2) can generate multiple judgment models based on the second data set using other available servers.

[0232] A plurality of judgment models based on a first data set received from a first edge device (100-1) and a plurality of judgment models based on a second data set received from a second edge device (100-2) can be simultaneously learned and generated through different available servers.

[0233] The processor (390) of the AI ​​conductor server (200-2) can distribute different models to each of the multiple edge devices. That is, the processor (390) of the AI ​​conductor server (200-2) can generate a model with different parameters based on data received from each edge device.

[0234] The edge conductor server (200-1) can transmit different judgment models to each of the multiple edge devices (100-1, 100-2) it manages.

[0235] The edge conductor server (200-1) can transmit a first model to a first edge device (100-1) and a second model with different parameters from the first model to a second edge device (100-2).

[0236] In another embodiment, multiple models can be trained using a single algorithm. In this case, each of the multiple models can be input with identical or different data.

[0237] Each model learning unit can learn the model using any one of the four algorithms described above.

[0238] The optimal model verification unit (330) and model selection unit (350) of FIG. 5b may be included in the running processor (240) of FIG. 2.

[0239] As another example, the optimal model verification unit (330) and the model selection unit (350) may be included in the processor (260, 390) of FIG. 2.

[0240] The display (370) can display the learning process of multiple models.

[0241] The processor (390) can control the overall operation of the AI ​​conductor server (200-2).

[0242] FIG. 6 is a diagram illustrating a process of measuring the prediction accuracy of each of a plurality of models learned by each of a plurality of algorithms according to one embodiment of the present disclosure.

[0243] Referring to Fig. 6, the training data set can be divided into N parts. The training data set can include structured data, unstructured data, and transformed data.

[0244] In Figure 6, N may be 4, but this is only an example.

[0245] The optimal model verification unit (330) can classify the training data set into four folds.

[0246] The optimal model verification unit (330) has one of the four folds as test data and the remaining three folds as training data.

[0247] Training data can be data used to learn a model, and test data can be data used to test whether the model has been learned well while preventing overfitting of the model.

[0248] The optimal model verification unit (330) can measure the first prediction result of the model when the first fold is used as test data.

[0249] The optimal model verification unit (330) can use the second fold as test data and the remaining folds as training data. When the second fold is used as test data, the optimal model verification unit (330) can measure the second prediction result of the model.

[0250] In this way, the optimal model verification unit (330) can measure the first to fourth prediction results.

[0251] The optimal model verification unit (330) can obtain the average of the first to fourth prediction results as the prediction accuracy of the model.

[0252] The optimal model verification unit (330) can perform N-fold cross-validation for each of the multiple models and obtain multiple prediction accuracies corresponding to each of the multiple models.

[0253] Meanwhile, the optimal model verification unit (330) can obtain the prediction accuracy of each of multiple models simultaneously and in parallel.

[0254] Below, we describe in detail the edge applications installed on each edge device.

[0255] Edge application features

[0256] Edge applications can be applications that enable edge devices to perform computing functions. The actions performed by edge applications can be actions performed by edge devices.

[0257] Edge applications can be installed without restrictions on personal computers, local servers, cloud servers, embedded devices, etc., and can perform computing according to the hardware performance of the installed environment.

[0258] Edge applications can be upgraded to the latest version via wireless communication with the Edge Conductor server (200-1). Edge devices can receive updated Edge applications from the Edge Conductor server (200-1).

[0259] The edge application can periodically transmit the operating status or failure of the model installed on the edge device, and the operating status and failure of the edge device to the edge conductor server (200-1) to maintain a stable computing state.

[0260] An edge application can transmit the inference results of the installed model to the edge conductor server (200-1). The inference results may include probability values.

[0261] The edge conductor server (200-1) can simultaneously communicate with multiple edge applications and control their functions. In other words, the edge conductor server (200-1) can centrally control multiple edge applications.

[0262] Edge applications can download a selected AI solution from among multiple AI solutions from the Edge Conductor server (200-1). Furthermore, edge applications can reinstall and re-execute new AI solutions based on user demand.

[0263] When an AI solution is installed on an edge device, an artificial intelligence model related to the AI ​​solution is trained in an AI conductor server (200-2), and the trained artificial intelligence model can be automatically distributed to the edge device.

[0264] An AI solution can be configured as a pipeline that receives data collected from edge devices, preprocesses the received data, creates an AI model using an AI algorithm based on the preprocessed data, and transmits the created AI model to the edge device.

[0265] As another example, an AI solution may consist of a pipeline that receives data collected from edge devices, preprocesses the received data, generates inference results using AI algorithms based on the preprocessed data, and delivers the generated inference results to edge applications.

[0266] Information, warnings, and errors that occur as the pipeline runs can be collected by the edge conductor server (200-1) to detect whether the AI ​​solution is operating normally.

[0267] When data requiring inference is generated on an edge device, the edge application can detect this and generate inference results.

[0268] Edge devices can display AI inference results on their displays through the edge viewer of edge applications.

[0269] Edge applications can transmit the inference results generated each time AI inference is performed, as well as the input data used for inference, to the edge conductor server (200-1). This allows users to review the inference results and determine whether a new AI model is needed.

[0270] When a user requests the creation of a new AI model, the AI ​​Conductor server (200-2) can create the new model and automatically deploy it to the edge application. This ensures that the edge device's inference performance remains fast.

[0271] The original data (input data) and inference results used in AI inference can be transmitted to the edge conductor server (200-1) at each inference. The edge conductor server (200-1) can manage the history of the received original data and inference results.

[0272] The original data (input data) used for AI inference may be sensing data collected by edge devices.

[0273] The original data can be used to retrain AI models. This allows for better management of data that impacts model performance, potentially enabling the development of high-performance AI solutions.

[0274] FIG. 7 is a drawing showing an execution screen of an edge UI application according to one embodiment of the present invention.

[0275] Edge UI applications can communicate with edge applications and execute various functions of edge applications.

[0276] Describes the execution screen (700) of the Edge UI application.

[0277] Number 1 can indicate the connection status with the edge application.

[0278] Number 2 can indicate the name of the model currently applied to the edge device.

[0279] Number 3 can indicate the operation mode of the edge application.

[0280] Button 4 may be a button indicating a request for inference results or data collection from a model installed on an edge device.

[0281] Number 5 can display the inferred image in real time.

[0282] Number 6 can represent real-time inference results. The inference result can be either OK or NG.

[0283] Number 7 can display statistics on the inferred results for the day. The statistics can include the number of counts judged OK and the number of counts judged NG.

[0284] FIG. 8 is a ladder diagram for explaining an operation method of an artificial intelligence system according to one embodiment of the present invention.

[0285] The first edge device (100-1) can collect a first data set (S801) and transmit the collected first data set to the edge conductor server (200-1) (S803).

[0286] The first data set may include one or more of sensing data or image data.

[0287] The first edge device (100-1) may be a device capable of inspecting the quality of a product at a manufacturing site. The first edge device (100-1) may include all of the components of FIG. 1.

[0288] The first edge device (100-1) may have the above-described edge application installed.

[0289] The first edge device (100-1) can transmit first device information along with the first data set to the edge conductor server (200-1). The first device information can include hardware specification information and operating system information of the first edge device (100-1).

[0290] The edge conductor server (200-1) can select a first AI solution based on the received first data set (S805) and transmit a request to create the first data set and the first AI solution to the AI ​​conductor server (200-2) (S807).

[0291] The edge conductor server (200-1) can select a first AI solution from among multiple AI solutions that matches the type of data included in the first data set. The edge conductor server (200-1) can also receive the selection of the first AI solution through user input.

[0292] Each of the multiple AI solutions can correspond to multiple data types. The data types can be sensor data, image data, or a first-stage business situation data set, but these are merely examples.

[0293] The AI ​​solution may be a solution that creates a model suitable for the type of data and distributes the created model to the first edge device (100-1).

[0294] The first AI solution creation request may be a request to create a data anomaly detection model that performs quality inspection and to deploy the created data anomaly detection model to the first edge device (100-1).

[0295] The first AI solution creation request may be for the creation of a data anomaly detection model as well as for the creation of a Docker image. In this case, the edge conductor server (200-1) may also transmit the first device information for the creation of the Docker image to the AI ​​conductor server (200-2).

[0296] The edge conductor server (200-1) may also create a docker image based on the first device information received from the first edge device (100-1).

[0297] The first AI solution creation request may include modeling code for creating a data anomaly detection model.

[0298] The AI ​​conductor server (200-2) can generate a first data anomaly detection model and a first quality criterion value according to a request for generation of a first data set and an AI solution (S809).

[0299] The first data anomaly detection model may be an artificial neural network-based model trained using deep learning or machine learning algorithms.

[0300] The first data anomaly detection model is a model for inspecting the quality of a product based on input data, and can be a model that infers whether the product is good or defective.

[0301] The first data anomaly detection model can be trained through supervised learning. The training data set of the first data anomaly detection model may include training data and labeled data labeled with the training data.

[0302] The training data may be sensing data or image data included in the first data set.

[0303] Labeling data can be correct data indicating whether a product is good or bad.

[0304] The AI ​​conductor server (200-2) can learn a first data anomaly detection model by performing labeling tasks on each piece of data included in the first data set. The AI ​​conductor server (200-2) can then generate a first data anomaly detection model for which learning has been completed.

[0305] When the first data anomaly detection model determines whether a product is good or defective, a first quality criterion value may be required as a reference. For example, if the first quality criterion value for classifying a product as good is set to 0.9, if the first data anomaly detection model outputs a value greater than or equal to 0.9, the product is considered good. If the model outputs a value less than 0.9, the product is considered defective.

[0306] The first quality criterion value can be a default value or a value set by the user.

[0307] Any one of the plurality of parameters constituting the first data anomaly detection model may represent a first quality criterion value.

[0308] The AI ​​conductor server (200-2) can transmit the generated first data anomaly detection model and the first quality criterion value to the edge conductor server (200-1) (S811).

[0309] The AI ​​conductor server (200-2) can transmit a first data anomaly detection model and a first quality value to which a first quality criterion value is applied to the edge conductor server (200-1).

[0310] The AI ​​conductor server (200-2) can transmit a first data anomaly detection model to which a first quality criterion value is applied, the first quality criterion value, and a docker image to the edge conductor server (200-1).

[0311] The edge conductor server (200-1) can transmit the first data anomaly detection model and the first quality criterion value to the first edge device (100-1) (S811).

[0312] The edge conductor server (200-1) can transmit a first data anomaly detection model, a first quality criterion value, and a docker image to the first edge device (100-1).

[0313] The edge conductor server (200-1) can transmit a first data anomaly detection model, a first quality criterion value, and a docker image to target edge devices including the first edge device (100-1).

[0314] The first edge device (100-1) can be equipped with a received first data anomaly detection model (S815).

[0315] The first edge device (100-1) can execute the first data anomaly detection model through an edge application and obtain the inference result of the first data anomaly detection model.

[0316] FIG. 9a is a diagram illustrating a process of adjusting a quality criterion value of a data anomaly detection model installed in a first edge device based on a first management situation data set.

[0317] The second edge device (100-2) can collect a first management situation data set including one or more of sales trend information, weekly product sales, weekly inventory, and weekly returns (S901).

[0318] Sales trend information can include product sales over a long period of time. This long period could be a specific season, a year, or a product event period, but these are examples only.

[0319] Weekly product sales may be the sales of products collected on a weekly basis.

[0320] Weekly inventory may be the inventory of products collected on a weekly basis.

[0321] Weekly returns may be the number of products returned collected on a weekly basis.

[0322] The second edge device (100-2) may be a server for collecting the first management situation data set. The second edge device (100-2) may collect the first management situation data set through an edge application.

[0323] The second edge device (100-2) can transmit the collected first management situation data set to the edge conductor server (200-1) (S903).

[0324] The edge conductor server (200-1) can select a second AI solution that matches the received first management situation data set (S905), and transmit a request for creating the second AI solution and the first management situation data set to the AI ​​conductor server (200-2) (S907).

[0325] The edge conductor server (200-1) can select a second AI solution from among multiple AI solutions, based on the type of data included in the first management situation data set. The edge conductor server (200-1) can also receive the selection of the second AI solution through user input.

[0326] The request for creating a second AI solution may be a solution that creates a quality criterion prediction model that infers the criteria for quality inspection based on the first management situation data set, and distributes the quality criterion values ​​predicted through the created quality criterion prediction model to the first edge device (100-1).

[0327] A request to create a second AI solution may be a request to create a Docker image as well as to create a quality criteria prediction model.

[0328] The request for creation of a second AI solution may include modeling code for creation of a quality criteria prediction model.

[0329] The AI ​​conductor server (200-2) can generate a quality criterion prediction model having a second quality criterion value based on the received AI solution generation request (S909).

[0330] The quality criteria prediction model can be an artificial neural network-based model trained using deep learning or machine learning algorithms.

[0331] A quality criterion prediction model may be a model that infers a quality criterion value used to determine whether a product is good or bad using an input first management situation data set.

[0332] A quality-based prediction model can be trained through supervised learning. The training data set for the quality-based prediction model may include a learning-use business situation data set and labeled data labeled in the learning-use business situation data set.

[0333] The learning management situation data set may include long-term sales trend information, short-term forecast inventory information, and short-term forecast return information, which are generated based on the first management data set.

[0334] Labeling data can be correct data that represents quality criteria values.

[0335] The AI ​​Conductor server (200-2) can learn a quality criterion prediction model by labeling learning management environment data. The AI ​​Conductor server (200-2) can then generate a quality criterion prediction model upon completion of learning.

[0336] The quality criterion prediction model may be a model for adjusting the quality criterion value of the data anomaly detection model installed in the first edge device (100-1).

[0337] The AI ​​conductor server (200-2) can obtain a second quality criterion value inferred by the quality criterion prediction model using the first management situation data set received from the edge conductor server (200-1) as input (S911).

[0338] The second quality criterion value may be a value adjusted from the first quality criterion value set in the first data anomaly detection model of the first edge device (100-1).

[0339] That is, the quality criterion value of the first data anomaly detection model can be adjusted based on the first management situation data set.

[0340] For the process of obtaining the second quality criterion value, see Fig. 9b.

[0341] FIG. 9b is a flowchart illustrating a process of inferring a second quality criterion value from a first management situation data set using the quality criterion prediction model of the present disclosure.

[0342] The AI ​​conductor server (200-2) can obtain seasonal predicted sales trend information from the sales trend information included in the first management situation data set (S931).

[0343] The AI ​​conductor server (200-2) can obtain seasonal predicted sales trend information from sales trend information using a seasonal autoregressive cumulative moving average model.

[0344] The seasonal autoregressive cumulative moving average model can be a statistical model for analyzing and forecasting time series data. This model considers seasonality, a characteristic of time series data, and can be used to predict data trends by combining autoregressive and cumulative moving averages.

[0345] Autoregressive (AR) can be a model in which the current value is expressed as a linear combination of previous values, and cumulative moving average (MA) can be a model that uses the average of observations.

[0346] The seasonal autoregressive cumulative moving average model is used to identify and predict seasonal patterns in time series data, and is effective in data sets with seasonal components.

[0347] Seasonal forecast sales trend information may include a seasonal sales trend function that predicts sales volume over a long-term period (or season).

[0348] Seasonal forecast sales trend information may include forecast sales volumes for a long period of time (or on a seasonal basis).

[0349] The AI ​​conductor server (200-2) can obtain short-term predicted sales trend information from the weekly product sales volume included in the first management situation data set (S933).

[0350] The AI ​​conductor server (200-2) can obtain short-term predicted sales trend information from weekly product sales volume using an artificial neural network-based time series data prediction model.

[0351] Short-term forecast sales trend information may include a short-term sales trend function that predicts weekly sales volume.

[0352] Short-term forecast sales trend information may include weekly forecast sales volumes.

[0353] Short-term forecast sales trend information may be information to specify the amount of change in long-term sales trend information.

[0354] The AI ​​conductor server (200-2) can obtain long-term sales trend information based on seasonal predicted sales trend information and short-term predicted sales trend information (S935).

[0355] The AI ​​conductor server (200-2) can generate a long-term sales trend function by combining a seasonal sales trend function and a short-term sales trend function.

[0356] The AI ​​conductor server (200-2) can generate a long-term sales trend function by merging a short-term sales trend function with a seasonal sales trend function.

[0357] The AI ​​conductor server (200-2) can obtain long-term sales trend function as long-term sales trend information.

[0358] Long-term sales trend information can include long-term unit sales forecasts. The long-term period can be months, several months, or years.

[0359] The AI ​​conductor server (200-2) can obtain short-term forecast inventory information based on the weekly inventory amount included in the first management situation data set (S937).

[0360] The AI ​​conductor server (200-2) can obtain short-term forecast inventory information from weekly inventory levels using an artificial neural network-based time series data prediction model.

[0361] Short-term forecast inventory information may include an inventory trend function that represents the trend in inventory levels over a week.

[0362] Short-term forecast inventory information may include weekly forecast inventory levels.

[0363] The AI ​​conductor server (200-2) can obtain short-term predicted return information based on the weekly return amount included in the first management situation data set (S939).

[0364] The AI ​​conductor server (200-2) can obtain short-term predicted return information from weekly return volume using an artificial neural network-based time series data prediction model.

[0365] Short-term forecast return information may include a return trend function that represents the trend in return volume on a weekly basis.

[0366] Short-term forecast return information may include weekly forecast return amounts.

[0367] The AI ​​conductor server (200-2) can obtain a second quality criterion value from long-term sales trend information, short-term predicted inventory information, and short-term predicted return information using a quality criterion prediction model (S941).

[0368] The AI ​​conductor server (200-2) can obtain final sales trend information by differentiating short-term forecast inventory information and short-term forecast return information from long-term sales trend information.

[0369] A quality criterion prediction model can be a model that infers quality criterion values ​​from final sales trend information. Final sales trend information can include long-term unit final forecast sales volumes that take into account predicted inventory and predicted returns.

[0370] The quality criterion prediction model can generate a second quality criterion value that is different from the first quality criterion value set in the first data anomaly detection model.

[0371] The second quality criterion value may be a value used to control the production volume of a product at the manufacturing site.

[0372] Again, Figure 9a is described.

[0373] The AI ​​conductor server (200-2) can transmit the second quality criterion value to the edge conductor server (200-1) (S913), and the edge conductor server (200-1) can transmit the received second quality criterion value to the first edge device (100-1) (S915).

[0374] The edge application of the first edge device (100-1) can apply the received second quality criterion value to the first data anomaly detection model. Accordingly, the first data anomaly detection model can be updated to the second data anomaly detection model.

[0375] Figures 10 and 11 are drawings explaining a process for updating a data anomaly detection model mounted on a first edge device.

[0376] FIG. 10 is a diagram illustrating a process of performing re-learning of a second data anomaly detection model when the judgment accuracy of the second data anomaly detection model is low according to an embodiment of the present disclosure.

[0377] The operations performed by the edge device (100-1) below may be operations performed by the edge application.

[0378] Referring to FIG. 10, the edge device (100-1) can update the first data anomaly detection model to the second data anomaly detection model based on the second quality criterion value received from the edge conductor server (200-1) (S1001).

[0379] The edge device (100-1) can obtain an inference result from input data using the second data anomaly detection model (S1003).

[0380] The input data can be sensing data or image data.

[0381] The edge device (100-1) can obtain judgment accuracy based on the inference result and determine whether the obtained judgment accuracy is less than a preset accuracy (S1005).

[0382] The edge device (100-1) can determine the accuracy of each of the multiple inference results.

[0383] The inference result may be an outcome indicating whether the product represents a good product or a bad product.

[0384] The preset accuracy can be either default or set based on user input.

[0385] The edge device (100-1) can determine that there is no abnormality in the second data abnormality detection model if the judgment accuracy of the second data abnormality detection model is higher than the preset accuracy.

[0386] The edge device (100-1) can determine that there is an abnormality in the second data abnormality detection model when the judgment accuracy of the second data abnormality detection model is less than the preset accuracy.

[0387] If the edge device (100-1) determines that the judgment accuracy is less than the preset accuracy, it can perform re-learning of the second data anomaly detection model (S1007).

[0388] That is, the second data anomaly detection model can be retrained to improve judgment accuracy.

[0389] FIG. 11 is a diagram illustrating a re-learning process of a data anomaly detection model according to an embodiment of the present disclosure.

[0390] The first edge device (100-1) can collect a second data set (S1101) and transmit a request for re-learning the collected second data set and the data anomaly detection model to the edge conductor server (200-1) (S1103).

[0391] The second data set may include one or more of sensing data or image data re-collected by the first edge device (100-1).

[0392] The edge conductor server (200-1) can transmit a second data set and a re-learning request to the AI ​​conductor server (200-2) (S1105).

[0393] The AI ​​conductor server (200-2) can generate a third quality criterion value and a third data anomaly detection model according to the second data set and re-learning request (S1107).

[0394] The AI ​​Conductor server (200-2) can retrain a data anomaly detection model by performing labeling tasks on each data included in the second data set. The AI ​​Conductor server (200-2) can then generate a third data anomaly detection model after completing the retraining.

[0395] When the third data anomaly detection model makes a final decision on whether a product is good or bad, a third quality criterion value may be required as a reference.

[0396] The AI ​​conductor server (200-2) can transmit the generated third data anomaly detection model and third quality criterion value to the edge conductor server (200-1) (S1109).

[0397] The AI ​​conductor server (200-2) can transmit a third data anomaly detection model and the third quality criterion value to which the third quality criterion value is applied to the edge conductor server (200-1).

[0398] The edge conductor server (200-1) can transmit the third data anomaly detection model and the third quality criterion value to the third edge device (100-1) (S1111).

[0399] The first edge device (100-1) can be equipped with a received third data anomaly detection model (S1113).

[0400] The first edge device (100-1) can update the second data anomaly detection model to the third data anomaly detection model through the edge application.

[0401] The first edge device (100-1) can obtain inference results through the updated third data anomaly detection model.

[0402] Figure 12 is a diagram illustrating a process of updating the quality criteria values ​​of a data anomaly detection model installed on an edge device according to re-collection of a business situation data set.

[0403] Fig. 12 may be an embodiment performed after the embodiment of Fig. 11. That is, Fig. 12 may be an embodiment of updating the quality criterion value of the third data anomaly detection model mounted on the edge device (100-1) from the third quality criterion value to the fourth quality criterion value.

[0404] As another example, FIG. 12 may be an embodiment performed after the embodiment of FIG. 9a. That is, FIG. 12 may be an embodiment of updating the quality criterion value of the second data anomaly detection model mounted on the edge device (100-1) from the second quality criterion value to the fourth quality criterion value.

[0405] Referring to FIG. 12, the second edge device (100-2) can re-collect a second business situation data set including sales trend information, weekly product sales, weekly inventory, and weekly return amount (S1201).

[0406] Sales trend information can include product sales over a long period of time. This long period could be a specific season, a year, or a product event period, but these are examples only.

[0407] Weekly product sales may be the sales of products collected on a weekly basis.

[0408] Weekly inventory may be the inventory of products collected on a weekly basis.

[0409] Weekly returns may be the number of products returned collected on a weekly basis.

[0410] The second edge device (100-2) may be a server for collecting a second set of business situation data. The second edge device (100-2) may re-collect the second set of business situation data through an edge application.

[0411] The second edge device (100-2) can transmit a re-generation request for re-generation of the re-collected second management situation data set and quality criterion values ​​to the edge conductor server (200-1) (S1203).

[0412] The edge conductor server (200-1) can transmit the received second management situation data set and re-generation request to the AI ​​conductor server (200-2) (S1205).

[0413] A regeneration request may be a request to regenerate a quality criterion value that is an inference result of a quality criterion prediction model.

[0414] The AI ​​conductor server (200-2) can obtain a fourth quality criterion value from the second management situation data set using a quality criterion prediction model (S1207).

[0415] That is, the AI ​​conductor server (200-2) can infer the fourth quality criterion value from the second management situation data set using the previously learned quality criterion prediction model.

[0416] The AI ​​conductor server (200-2) can obtain long-term sales trend information, short-term forecast inventory information, and short-term forecast return information from the second management situation data set through the example of FIG. 9b.

[0417] The AI ​​conductor server (200-2) can obtain a fourth quality criterion value from long-term sales trend information, short-term predicted inventory information, and short-term predicted return information using a quality criterion prediction model.

[0418] The AI ​​conductor server (200-2) can transmit the fourth quality criterion value to the edge conductor server (200-1) (S1209), and the edge conductor server (200-1) can transmit the received fourth quality criterion value to the first edge device (100-1) (S1211).

[0419] The edge device (100-1) can update the third data anomaly detection model to the fourth data anomaly detection model using the fourth quality criterion value (S1213).

[0420] The edge device (100-1) can generate a fourth data anomaly detection model by adjusting the value of a parameter corresponding to a quality criterion value among the parameters of the third data anomaly detection model to a fourth quality criterion value.

[0421] According to the embodiment of Fig. 12, even if the prediction of the quality standard fails and the inventory amount increases, there is an effect that allows for flexible response as the quality standard is readjusted.

[0422] The aforementioned disclosure can be implemented as computer-readable code on a program-recorded medium. The computer-readable medium includes any type of recording device that stores data readable by a computer system.

[0423] Examples of computer-readable media include hard disk drives (HDDs), solid state disks (SSDs), silicon disk drives (SDDs), ROMs, RAMs, CD-ROMs, magnetic tapes, floppy disks, optical data storage devices, etc. In addition, the computer may include a processor (180) of an artificial intelligence device.

Claims

1. In an artificial intelligence system that adjusts the quality standards of models installed on edge devices by considering the business situation, A first edge device equipped with a first data anomaly detection model having a first quality criterion, wherein the first data anomaly detection model is an artificial intelligence model that determines whether a product is good or defective; A second edge device that collects a first set of business situation data and transmits the collected first set of business situation data to an edge conductor server; An edge conductor server that transmits a request for generating the first management situation data set and quality criteria prediction model received from the second edge device; and Including an artificial intelligence conductor server that generates a quality criterion prediction model that infers a quality criterion value of a data anomaly detection model based on the first management situation data set, obtains a second quality criterion value inferred by the quality criterion prediction model, and transmits the obtained second quality criterion value to the edge conductor server; The above edge conductor server Transmitting the above second quality criterion value to the first edge device, The above first edge device Based on the second quality criterion value, the first data anomaly detection model is updated to the second data anomaly detection model. Artificial intelligence system.

2. In paragraph 1, The above first management situation data set is Contains one or more of the following information: sales trend information, weekly product sales, weekly inventory, and weekly returns. Artificial intelligence system.

3. In paragraph 1, The above first edge device Obtain the judgment accuracy of the above second data anomaly detection model, and if the acquired judgment accuracy is less than the preset accuracy, transmit a re-learning request of the above second data anomaly detection model and a re-collected data set to the edge conductor server. Artificial intelligence system.

4. In paragraph 3, The above edge conductor server Requesting re-learning of the above second data anomaly detection model and transmitting the re-collected data set to the artificial intelligence conductor server, The above artificial intelligence conductor server Create a third data anomaly detection model with a third quality criterion value based on the recollected data set, The generated third data anomaly detection model is transmitted to the edge conductor server, The above edge conductor server Transmitting a third data anomaly detection model having the third quality criterion value to the first edge device, The above first edge device The above second data anomaly detection model is updated to the third data anomaly detection model having the third quality criterion value. Artificial intelligence system.

5. In paragraph 4, The above second edge device Collect a second business situation data set, and transmit the collected second business situation data set to the edge conductor server, The above edge conductor server Sending a request to regenerate the above second management situation data set and quality criteria values ​​to the artificial intelligence conductor server. Artificial intelligence system.

6. In paragraph 5, The above artificial intelligence conductor server Obtain a fourth quality criterion value inferred by the quality criterion prediction model based on the second management situation data set, and transmit the obtained fourth quality criterion value to the edge conductor server. The above edge conductor server Transmitting the above fourth quality criterion value to the above first edge device, The above first edge device Update the above third data anomaly detection model to a fourth data anomaly detection model having the fourth quality criterion value. Artificial intelligence system.

7. In paragraph 1, The above quality criteria values ​​are The above first data anomaly detection model is the cut-off value used when determining whether a product is good or bad. Artificial intelligence system.

8. In the operation method of an artificial intelligence system that adjusts the quality standards of a model installed in an edge device by considering the management situation, A step in which a first edge device is equipped with a first data anomaly detection model having a first quality criterion, wherein the first data anomaly detection model is an artificial intelligence model that determines whether a product is good or bad; A step in which a second edge device collects a first management situation data set and transmits the collected first management situation data set to an edge conductor server; A step in which the edge conductor server transmits a request for generation of the first management situation data set and a quality criteria prediction model received from the second edge device; A step in which an artificial intelligence conductor server generates a quality criterion prediction model that infers a quality criterion value of a data anomaly detection model based on the first management situation data set; A step in which the artificial intelligence conductor server obtains a second quality criterion value inferred by the quality criterion prediction model; A step in which the artificial intelligence conductor server transmits the acquired second quality criterion value to the edge conductor server; The step of the edge conductor server transmitting the second quality criterion value to the first edge device; and The first edge device includes a step of updating the first data anomaly detection model to a second data anomaly detection model based on the second quality criterion value. How an artificial intelligence system works.

9. In paragraph 8, The above first management situation data set is Contains one or more of the following information: sales trend information, weekly product sales, weekly inventory, and weekly returns. How an artificial intelligence system works.

10. In paragraph 8, The step of the first edge device obtaining the judgment accuracy of the second data anomaly detection model; and The first edge device further includes a step of transmitting a re-learning request for the second data anomaly detection model and a re-collected data set to the edge conductor server when the acquired judgment accuracy is less than the preset accuracy. How an artificial intelligence system works.

11. In paragraph 10, The step of the edge conductor server transmitting a re-learning request for the second data anomaly detection model and the re-collected data set to the artificial intelligence conductor server; A step in which the artificial intelligence conductor server generates a third data anomaly detection model having a third quality criterion value based on the re-collected data set; A step in which the artificial intelligence conductor server transmits the generated third data anomaly detection model to the edge conductor server; The step of the edge conductor server transmitting a third data anomaly detection model having the third quality criterion value to the first edge device; and The first edge device further includes a step of updating the second data anomaly detection model to the third data anomaly detection model having the third quality criterion value. How an artificial intelligence system works.

12. In paragraph 11, The second edge device collects a second management situation data set and transmits the collected second management situation data set to the edge conductor server; and The edge conductor server further comprises a step of transmitting a request for re-creating the second management situation data set and quality criterion values ​​to the artificial intelligence conductor server. How an artificial intelligence system works.

13. In paragraph 12, A step in which the artificial intelligence conductor server obtains a fourth quality criterion value inferred by the quality criterion prediction model based on the second management situation data set; A step of transmitting the acquired fourth quality criterion value to the edge conductor server; The step of the edge conductor server transmitting the fourth quality criterion value to the first edge device; and The first edge device further includes a step of updating the third data anomaly detection model to a fourth data anomaly detection model having the fourth quality criterion value. How an artificial intelligence system works.

14. In paragraph 8, The above quality criteria values ​​are The above first data anomaly detection model is the cut-off value used when determining whether a product is good or bad. How an artificial intelligence system works.

Images