KR20240074074A - AI-based risk prediction anomaly detection model platform service method - Google Patents

Abstract

The present invention relates to an AI-based risk prediction anomaly detection model platform service method. According to the present invention, heterogeneous process equipment data at the production site is collected and interfaced with an AI deep learning system, and risk prediction and abnormal signs are predicted in real time through big data analysis of the heterogeneous process equipment data at the production site based on AI. By monitoring, production site risk critical information can be managed efficiently.

Description

AI-based risk prediction anomaly detection model platform service method {AI-based risk prediction anomaly detection model platform service method}

The present invention relates to an AI-based risk prediction anomaly detection model platform service method.

Smart manufacturing refers to a technology that actively applies information and communication technologies to the current manufacturing process, and is revolutionizing the global manufacturing industry economy.

The Fourth Industrial Revolution, which is emerging today, refers to the era of building a smart manufacturing environment based on cutting-edge information and communication technology (ICT). Although there is no exact definition, SMLC is a national R&D consortium established under the leadership of the U.S. federal government. (Smart Manufacturing Leadership Coalition) refers to smart manufacturing as 'the intensified application of advanced intelligent systems that enable rapid manufacturing of new products, active response to product demand, and real-time optimization of production and supply chain networks.'

Through this, it is possible to combine more advanced digital technologies with production systems, and accompanying technologies include wireless communication technology, Internet of Things technology, cloud computing technology, reprogrammable robot technology, and machine intelligence technology.

The final result of this smart manufacturing will be a smart factory, and the core of this smart manufacturing is data, and how to acquire, manage, and use data is important.

Recently, as automation processes have developed, various sensors and actuators have been introduced, and higher-level automation performance and functions are being implemented through sequence programs using these. Accordingly, the complexity of sequence programs inevitably increases.

As complexity increases, the time and cost required to analyze the cause or recover when an error or malfunction occurs in an automated process is also increasing.

In particular, when a malfunction occurs in an automated process, if the cause of the malfunction is not traced and action is not taken quickly, it is very important to enable a rapid response because it is directly connected to a large loss for the company.

(1) Domestic registered patent No. 10-2410459 B1 (2) Domestic published patent 10-2021-0111502 A (3) Domestic published patent 10-2022-0059837 A

The present invention aims to solve the above-described needs and/or problems.

In addition, data from heterogeneous process equipment at the production site can be collected and interfaced with an AI deep learning system, and risk prediction and abnormal signs can be monitored in real time through big data analysis of heterogeneous process equipment data at the production site based on AI. The purpose is to provide an AI-based risk prediction anomaly detection model platform service method.

According to one aspect of the AI-based risk prediction anomaly detection model platform service method according to the present invention to achieve the above object, collecting heterogeneous process equipment data at the production site; Collecting the collected heterogeneous process equipment data from the production site and interfacing them to be linked to an AI deep learning system; Performing big data analysis on heterogeneous process equipment data at the production site based on AI; And it may include predicting risks and monitoring abnormal signs in real time based on heterogeneous process equipment data at the production site analyzed based on the AI.

The step of performing big data analysis on the heterogeneous process equipment data at the production site based on the AI includes receiving the collected heterogeneous process equipment data at the production site as input from the AI deep learning system; Learning the normal range through deep learning on heterogeneous process equipment data at the production site; Modifying an AI-based risk prediction anomaly detection model based on heterogeneous process equipment data at the production site; And it may include performing big data analysis on heterogeneous process equipment data at the production site based on the modified AI-based risk prediction abnormality symptom detection model.

In the step of real-time risk prediction and monitoring of abnormal signs based on heterogeneous process equipment data at the production site analyzed based on the AI, abnormal signs that deviate from the normal range learned through deep learning can be automatically detected.

In the step of modifying the AI-based risk prediction anomaly detection model based on the heterogeneous process equipment data at the production site, if it is impossible to extract characteristic variables applied to the AI-based risk prediction anomaly detection model, iterative analysis is performed through additional data collection. This may further include steps to be performed.

In the step of real-time monitoring of risk prediction and abnormal signs based on the heterogeneous process equipment data at the production site analyzed based on the AI, if a data trend outside the normal statistical range is detected in the heterogeneous process equipment data at the production site. It is possible to generate an AI information-based warning alarm and change the alarm value through AI logic by linking it with an intelligent algorithm using deep learning.

In the step of real-time monitoring of risk predictions and abnormal signs based on the heterogeneous process equipment data at the production site analyzed based on the AI, the information monitored by the AI deep learning system is searched by the production site system controller and the corresponding process equipment is checked. can be controlled.

The effects according to an embodiment of the present invention will be described as follows.

According to the present invention, heterogeneous process equipment data at the production site is collected and interfaced with an AI deep learning system, and risk prediction and abnormal signs are predicted in real time through big data analysis of the heterogeneous process equipment data at the production site based on AI. By monitoring, production site risk critical information can be efficiently managed.

The accompanying drawings, which are included as part of the detailed description to aid understanding of the present invention, provide embodiments of the present invention, and together with the detailed description, explain technical features of the present invention.
Figure 1 is a diagram showing an example of the configuration of an AI-based risk prediction anomaly detection model platform according to an embodiment of the present invention.
Figure 2 is a diagram showing an example of providing an AI-based data collection and analysis service by collecting and analyzing data related to sensors and relay devices in industrial sites.
Figure 3 is a diagram showing an AI-based risk prediction anomaly detection model platform service method.
Figure 4 is a diagram showing an example of AI-based alarm notification triggering.
Figure 5 is a diagram illustrating the overall system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.
Figure 6 is a block diagram showing the configuration of an information collection and analysis device applied to a system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.
Figure 7 is a procedural flowchart sequentially showing the operation of a system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.

Hereinafter, embodiments disclosed in the present invention will be described in detail with reference to the attached drawings. However, identical or similar components will be assigned the same reference numbers regardless of reference numerals, and duplicate descriptions thereof will be omitted. The suffixes “module” and “part” for components used in the following description are given or used interchangeably only for the ease of preparing the specification, and do not have distinct meanings or roles in themselves. Additionally, in describing the embodiments disclosed in the present invention, if it is determined that detailed descriptions of related known technologies may obscure the gist of the embodiments disclosed in the present invention, the detailed descriptions will be omitted. In addition, the attached drawings are only for easy understanding of the embodiments disclosed in the present invention, and the technical idea disclosed in the present invention is not limited by the attached drawings, and all changes included in the spirit and technical scope of the present invention are not limited. , should be understood to include equivalents or substitutes.

Terms containing ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

When a component is said to be "connected" or "connected" to another component, it is understood that it may be directly connected to or connected to the other component, but that other components may exist in between. It should be. On the other hand, when it is mentioned that a component is “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

Singular expressions include plural expressions unless the context clearly dictates otherwise.

In this application, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Figure 1 is a diagram showing an example of the configuration of an AI-based risk prediction anomaly detection model platform according to an embodiment of the present invention.

As shown, an intelligent data collection and analysis platform can be implemented by applying the AI-based risk prediction anomaly detection model platform of the present invention to a smart factory.

The AI-based risk prediction anomaly detection model platform of the present invention can be composed of a sensor interface, middleware, and applications.

The sensor interface includes a human-machine interface (HMI) monitoring system, and the human-machine interface (HMI) monitoring system may include at least one process facility. For example, as processing equipment, a raw material conveyor may be equipped with a movement measurement sensor, and a processing machine may be equipped with a motion detection sensor. Additionally, the good/defective inspection machine may be equipped with a pattern detection sensor, and the actuator may be equipped with a warning light, a siren, etc.

Middleware may include an integrated dashboard, AI big data platform, and data collection and analysis device. An integrated dashboard can provide process sensor monitoring, real-time status analysis, and alert/response action services. The AI big data platform can provide AI support linked to deep learning and machine learning big data, statistical analysis pattern AI algorithms, real-time facility operation information, and data DB storage services. Additionally, the data collection and analysis device can provide data collection, signal interface, AI-based embedded module, and IoT platform real-time linked API service.

The application can provide real-time information analysis, sensor reference information, risk information alarm, and quality/safety management services. Of course, additional services can be provided as needed.

Figure 2 is a diagram showing an example of providing an AI-based data collection and analysis service by collecting and analyzing data related to sensors and relay devices in industrial sites.

As shown, data from industrial site sensors, switches, relays, and production site process equipment 100 related to PLCs can be collected and analyzed by the data collection and analysis device 200. The data collection and analysis device can perform on-site big data analysis, heterogeneous signal processing, protocol management linkage, advance failure prevention function, and real-time control monitoring in conjunction with the AI-based data collection and analysis service of the AI deep learning system 300. The AI-based data collection and analysis service can predict AI-based risks including deep learning, statistics, and databases, and can provide services such as a proactive platform, monitoring and notification, and fault diagnosis.

Figure 3 is a diagram showing an AI-based risk prediction anomaly detection model platform service method, and Figure 4 is a diagram showing an example of AI-based warning notification triggering.

As shown, after collecting and aggregating field data, the data can be analyzed by reflecting the AI algorithm. In other words, when data is input through AI logic and deep learning, the input data can be learned, the model modified, and the data analyzed. For example, it can display real-time sensor values and trends. Additionally, detailed analysis of sensor values, maximum, minimum, average, change, etc. can be displayed. By this, an AI information-based warning alarm can be generated.

Normal and anomaly are defined differently for each field and problem, and defining a classification model using an AI network may be necessary to establish an AI solution necessary for business promotion. . Here, in classification to determine which category the data belongs to, binary classification is used to classify data with two labels as 0 or 1 when input, and uses the sigmoid function as the activation function. (Sigmoid) is used (output as 0 or 1) and the alarm is determined through the class. If it is an alarm, it can be classified as Alert, and if it is a normal temperature, it can be classified as Normal.

In addition, as AI-based real-time monitoring management, it can provide optimal on-site analysis information for each data based on big data analysis using AI and provide smart monitoring functions necessary for repetitive and standardized errors and automated alarm warnings.

It provides intelligent AI algorithms such as pattern analysis, object detection, visualization, big data, and machine learning, improves data errors that occur in advance, provides monitoring, control, and failure prevention alarms, and analyzes and utilizes field data to provide advance failure alarms. and an alarm system can be provided.

In addition, when data trends are detected that are beyond data collection and analysis and manufacturing site statistics, alarm and warning values can be linked to an intelligent algorithm using deep learning to change alarm warning values through AI logic.

Information can be interfaced with the AI system, and the information displayed by the AI system can be viewed and controlled by the production site system controller. Deep learning and machine learning (AI business pattern analysis, risk prediction) can be applied to provide information on field data. Implement an automated solution of big data analysis and intelligent logic, install a prior risk warning function, acquire and process data, input data through AI deep learning → learn data → modify model → create a model through repetition of preparing a model (draft). You can implement real services and find the optimal solution.

As the first step in developing a model through machine learning, the first step in developing a machine learning model through data collection is data collection, which is an important step in model development. Variable extraction, model selection, and application techniques applied to the development model according to the characteristics of the collected data. Most involved in selection.

In the case of a process that has been operated for a long time, it is relatively easy to collect data representing the process because it has been operated based on important data. However, if it is impossible to extract characteristic variables, repeated analysis must be performed through additional data collection.

Data such as temperature, speed, and flow at a manufacturing production site can be integrated through an AI controller, a heterogeneous collection and analysis device, and interfaced with this data by automatically linking it to the AI deep learning system. In addition, by interfacing the results of the AI engine system, such as implementing AI modeling learning data sets, by implementing an analysis web system for production site data, it is possible to manage production site risk critical information through analysis monitoring.

Provides a basis for deriving alarms from AI and monitoring functions utilizing AI analysis information, sets up AI multiple regression prediction model development API implementation and big data cloud environment implementation, and learns the normal range using deep learning technology. It is possible to provide an AI engine that automatically detects anomalies that fall outside the normal range.

Figure 5 is a diagram illustrating the overall system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.

As shown in Figure 5, the system of one embodiment is largely comprised of equipment installed for a plurality of different processes at a manufacturing production site, that is, sensor devices (110-1 to 130-2, etc.), and a self-network. It includes a data collection analyzer 200 connected to each sensor device (110-1 to 130-2, etc.). Here, in addition to sensor devices (110-1 to 130-2, etc.), relays, devices, and HMI monitoring devices are also used as equipment. Additionally, the private network includes, for example, a TCP/IP network, an LTE network, a wireless communication network, etc.

In this case, the system according to one embodiment further includes a management information processing device (including a PC) and an administrator terminal that collectively control the operation of the control device, for example, the operation of the air conditioner and the heater, through the data collection and analysis device 200. Includes.

In addition, a wider range of services is provided by connecting external contacts, and in this case, the external contacts include troubleshooting centers, information processing devices, etc.

The sensor devices (110-1 to 130-2, etc.) detect sensor information or manufacturing information for each of a plurality of different processes or facilities at a manufacturing production site and provide it to the data collection and analyzer 200. For example, process status information, including temperature, speed, and flow rate, is detected for each process at a manufacturing production site for a number of different models.

The data collection and analysis device 200 basically collects information detected by sensor devices (110-1 to 130-2, etc.) through its own network. In addition, the data collection and analysis device 200 determines the process status for each process through the sensed information, and controls the operation of the control device differently depending on the process status or facility status.

In addition, the management information processing device or data collection and analysis device 200 configures the function of configuring the program screen as a web service and uses a web development interface on the worker's PC. In this case, it includes a service that configures a web screen from the web user interface and executes it to create an operable application.

As above, the administrator terminal configures the function of configuring the program screen as an app service, and uses the app development interface as the worker's portable terminal. Additionally, it includes a service that configures the app screen from the app user interface and executes it to create an operable application.

Figure 4 is a block diagram showing the configuration of an information collection and analysis device applied to a system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.

As shown in FIG. 4, the data collection and analysis device 200 according to one embodiment is largely comprised of a key signal input unit 201, a signal input/output unit 202, a signal processing unit 203, a control unit 204, and a storage medium ( 205) and a display unit 206.

The key signal input unit 201 provides various setting information for advance failure notification and warning according to an embodiment, for example, equipment device information and control device information installed for each process at each manufacturing production site, regression analysis, and manager device. Information about the system, etc. is input according to the user's key operations.

The signal input/output unit 202 transmits various process status information and control information through input and output to the equipment 100 installed for each process at each external manufacturing production site, that is, sensor devices, relays, devices, and HMI monitoring devices.

For example, when heterogeneous process state information is received, the signal processing unit 203 converts it to link it to a preliminary failure notification and warning format according to an embodiment. In addition, various signals are converted so that they can be processed according to the pre-failure notification and warning format.

The control unit 204 checks the process status for each process according to the process status information of the signal input/output unit 202, generates different control signals for each process status or facility status, and transmits them to the control device. This control unit 204 performs the function of collecting and analyzing input signals in connection with the above-described prior failure notification and warning format, and transmitting the analyzed information to an external registration manager device.

The storage medium 204 stores device registration information of various equipment, administrator setting information, prior failure notification, and alarm format under the control of the control unit 204.

The display unit 205 displays analysis information, such as failure notification information or alarm information, under the control of the control unit 204.

Figure 7 is a procedural flowchart sequentially showing the operation of a system that provides a preliminary risk warning alarm through abnormality detection through AI analysis according to an embodiment.

As shown in FIG. 7, in one embodiment, first, a facility that collects and provides process status information through one or more of the field sensors, relays, devices, and HMI monitoring devices installed for each of the multiple different processes at the manufacturing production site. Includes devices.

In addition, it includes a data collection and analyzer 200 that is located remotely from the equipment and monitors and manages the process status of each process by collectively analyzing the process status information provided by the equipment.

Additionally, in this state, according to one embodiment, the information collection and analysis device 200 operates as follows.

That is, when the process status of each process is collectively analyzed, on-site process status information, including temperature, speed, and flow rate, for each process at the manufacturing production site is collected separately for a plurality of different models.

Then, based on this on-site process status information, if the on-site process status information is temperature information, the current region and season are confirmed.

And, based on this on-site process status information, the extent of climate change as of today compared to yesterday and the temperature difference occurrence value as the extent of change exceeds the set value are confirmed.

Additionally, based on the process status information at this site, a learning method corresponding to the process is extracted.

So, based on each information, it is determined whether or not there is an alarm according to the pre-set failure alarm and warning format.

So, this determination information is output on the screen.

In this case, the pre-failure alarm and warning format is as follows.

First, a learning method is defined by classifying process status information, including temperature, speed, and flow rate, for each process at the manufacturing production site for a plurality of different models to obtain learning information.

And, when obtaining such learning information, the learning information is obtained by classifying the process state information through set logistic regression analysis using a sigmoid function that calculates the probability of success when process state information is provided.

In particular, when the process state information is temperature, the linear relationship between temperature and meteorological environmental conditions is used to perform multiple regression analysis by region and season based on learning information of solar radiation, outdoor temperature, and wind speed variables measured by weather observation devices. Perform.

Then, the information obtained from the multiple regression analysis is verified by comparing it with the actual process temperature during the set period.

Based on this verification information, the change in today's climate temperature compared to yesterday is read, and a temperature prediction method is created that indicates a warning or alarm according to the temperature difference occurrence value where the change range exceeds the set value.

Next, characteristic variables are extracted according to the process characteristics for each of the multiple different processes, and divided into supervised (classification, regression) or unsupervised (cluster) learning methods according to the selected information characteristics, and supervised or unsupervised learning for each process. Select a method. For example, in the case of a temperature separation process, a regression analysis method is used.

Therefore, a preliminary failure alarm and warning format is created based on the various information described above.

So, in one embodiment, in the analysis of information collection and analysis devices, AI deep learning based on gateway implementation for heterogeneous big data compatibility in industrial sites is considered, taking into account that normal and abnormal signs are defined differently for each field and problem. Provides integrated information analysis using

Furthermore, through this analysis, process problems are prevented in advance.

As described above, in one embodiment, the information collection and analysis device collects process status information from sensors, relays, devices, HMI monitoring devices, etc. provided for each process or facility at the manufacturing production site and collectively analyzes the process status. In this case, on-site process status information, including temperature, speed, and flow rate, for each process at the manufacturing production site is collected separately for a plurality of different models.

In addition, the on-site process status information is analyzed according to the pre-failure alarm and warning format set according to one embodiment, the presence of an alarm is determined, and it is output on the screen and provided to managers, etc.

Therefore, in one embodiment, considering that normal and abnormal signs are defined differently in each field and problem in the analysis of information collection and analysis devices, AI deep learning based on gateway implementation is utilized for heterogeneous big data compatibility in industrial sites. Provides integrated information analysis.

Furthermore, through this analysis, process problems are prevented in advance.

The above-described present invention can be implemented as computer-readable code on a program-recorded medium. Computer-readable media includes all types of recording devices that store data that can be read by a computer system. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SDD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. There is. Accordingly, the above detailed description should not be construed as restrictive in all respects and should be considered illustrative. The scope of the present invention should be determined by reasonable interpretation of the appended claims, and all changes within the equivalent scope of the present invention are included in the scope of the present invention.

100: Production site process equipment
200: data collection analyzer
300: AI deep learning system

Claims (6)

As an AI-based risk prediction anomaly detection model platform service method,
Collecting heterogeneous process equipment data at the production site;
Collecting the collected heterogeneous process equipment data from the production site and interfacing them to be linked to an AI deep learning system;
Performing big data analysis on heterogeneous process equipment data at the production site based on AI; and
An AI-based risk prediction anomaly detection model platform service method comprising the step of predicting risk and monitoring abnormal signs in real time based on heterogeneous process equipment data at the production site analyzed based on the AI.

In claim 1,
The step of performing big data analysis on heterogeneous process equipment data at the production site based on the AI is:
Receiving the collected heterogeneous process equipment data from the production site as input from the AI deep learning system;
Learning the normal range through deep learning on heterogeneous process equipment data at the production site;
Modifying an AI-based risk prediction anomaly detection model based on heterogeneous process equipment data at the production site; and
An AI-based risk prediction anomaly detection model platform service method comprising the step of performing big data analysis on heterogeneous process equipment data at the production site based on the modified AI-based risk prediction anomaly detection model. In claim 2,
In the step of predicting risks and monitoring abnormal signs in real time based on the heterogeneous process equipment data at the production site analyzed based on the AI,
An AI-based risk prediction anomaly detection model platform service method that automatically detects abnormalities that deviate from the normal range learned through deep learning. In claim 2,
In the step of modifying the AI-based risk prediction anomaly detection model based on the heterogeneous process equipment data at the production site,
An AI-based risk prediction anomaly detection model platform service method further comprising performing repetitive analysis through additional data collection when it is impossible to extract characteristic variables applied to the AI-based risk prediction anomaly detection model. In claim 1,
In the step of predicting risks and monitoring abnormal signs in real time based on the heterogeneous process equipment data at the production site analyzed based on the AI,
When a data trend outside the normal statistical range is detected in the heterogeneous process equipment data at the production site, an AI information-based alarm is generated and the alarm value is changed through AI logic in conjunction with an intelligent algorithm using deep learning. AI-based risk prediction anomaly detection model platform service method. In claim 1,
In the step of predicting risks and monitoring abnormal signs in real time based on the heterogeneous process equipment data at the production site analyzed based on the AI,
An AI-based risk prediction abnormality detection model platform service method that queries the information monitored in the AI deep learning system from the production site system controller and controls the relevant process equipment.