KR102379259B1 - AIoT-based integrated management system of injection manufacturing facilities and its operating method - Google Patents

Abstract

The present invention relates to an AIoT-based integrated management system for an injection manufacturing facility, capable of deriving an optimal process condition for an injection manufacturing facility based on injection molding condition data, injection peripheral data, mold interior data, and the like through artificial intelligence (AI) and supporting remote control suitable for the process condition to enable intelligent integrated management breaking from existing management schemes relying only on the proficiency of workers, and an operation method thereof. In accordance with the present invention, an optimal injection molding condition (AI model) is derived through a cloud environment AI algorithm, and a molding condition control command generated based on the AI model and measurement data of each process parameter is configured to exchange information through a communication mode of a message queuing telemetry transport (MQTT) protocol between an edge cluster and a control server, thereby efficiently utilizing IT resources as well as contributing to quality and productivity improvements of injection-molded products. In accordance with the present invention, there can be effects of enabling automatic remote control and effectively responding to a realistic demand for material property improvement through a new type of molding material, and preventing the occurrence of defective items and a deterioration in productivity, caused by human errors or poor management.

Description

AIoT-based integrated management system of injection manufacturing facilities and its operating method

The present invention relates to an AIoT-based integrated management system for injection manufacturing equipment and an operating method thereof, and more specifically, injection molding condition data, injection peripheral device data, mold internal data, etc. through AI (artificial intelligence) technology. Integration of AIoT-based injection manufacturing facilities capable of intelligent integrated management by breaking away from the existing management method that relied only on the skill level of workers by deriving the optimal process conditions for manufacturing facilities and supporting remote control suitable for the process conditions It relates to a management system and its operation method.

The trend of digitalization is getting stronger in the industrial field centering on the keyword of the 4th industrial revolution. In particular, as the scale of production facilities grows and remote control technology becomes more common, the optimal process conditions that contribute to productivity improvement are derived by efficiently collecting and analyzing data pouring from numerous production facilities at the manufacturing site. A technology for an integrated management system that automatically and remotely controls is attracting attention.

Recently, with the spread of Internet of Things (IoT) technology, various sensors and devices are connected through the Internet (communication network), making it easy to exchange information between devices or to match related data, providing specialized services in various fields Platform services for this purpose are becoming the subject of interest. In particular, as the technological development of IIoT (Industrial Internet of Things) specialized in industrial fields is booming, the enthusiasm for the development of AIoT (Intelligent Internet of Things) platform technology in which artificial intelligence (AI) technology is grafted into IIoT is gradually increasing.

This movement is also actively appearing in the field of injection molding machines, which can be said to be the core of the manufacturing industry, and as the amount of data exchanged between manufacturing facilities for integrated control has rapidly increased, data processing efficiency and stable and The importance of intelligent communication network operation is being emphasized more and more.

In accordance with such a situation, Patent No. 10-2021-0056235 (Prior Patent 1) discloses an artificial intelligence (AI)-based injection molding system and a molding condition generation method. is attached, and through the molding condition generator, it is judged whether the product produced by the injection molding machine maintains the target molding quality. If the molding quality is not maintained due to disturbance, the molding conditions are changed to follow the target data Techniques are exemplified.

However, in Prior Patent 1 above, a certain amount of pure function can be expected in that the molding conditions are controlled through the molding condition generating device attached to the injection molding machine, but a separate molding condition generating device must be attached to each injection molding machine. and each molding condition generating device must have information processing resources (IT resources) such as databases and learning algorithms to support artificial intelligence (AI) learning. There was a structural problem that not only the operational efficiency of the resource (IT resource) was hindered, but also the increase in the installation cost was unavoidable.

In addition, Registration Patent No. 10-2264066 (Prior Patent 2) discloses a smart control system for injection molding machine by artificial intelligence (AI) and image analysis and an operation method thereof. When defects that exceed the standard occur continuously by inspecting the appearance (photographed image) with the learned artificial intelligence (AI), or the cumulative number of occurrences of defective products is confirmed to be greater than the standard value, external environmental variables around the injection molding machine are sequentially detected and artificial intelligence (AI) is used. A technical matter of sequentially adjusting preset molding condition variables by intelligent analysis is disclosed.

However, in the previous patent 2 above, only the technique of adjusting the molding conditions affecting the quality of the final product by applying the AI learning technique to compare the image image of the product produced through the injection molding machine with the reference image is exemplified. Since they were unable to escape the structure of attaching a separate molding condition changing device to each, they had no choice but to receive an evaluation that efficient integrated management was still insufficient in an environment where multiple injection molding machines were operated.

(Prior Patent 1) Patent Publication No. 10-2021-0056235 (Prior Patent 2) Registered Patent No. 10-2264066

The present invention has been made in view of the problems of the above background technology, and provides an integrated management system for an AIoT-based injection manufacturing facility that supports efficient integrated management in a large-scale factory equipped with a plurality of injection molding machines, and an operating method thereof. The purpose.

In addition, the present invention derives the optimal injection molding conditions (AI model) using the AI algorithm of the cloud environment, and the molding condition control command based on the measurement data of each process variable and the AI model is MQTT (Message Queuing Telemetry Transport) broker By configuring information exchange between the edge cluster and the control server through intended to provide.

In the integrated management system of the AIoT-based injection manufacturing facility of the present invention, the edge cluster, the control server and the cloud server installed in the injection molding machine are connected through a communication network, and the edge cluster has a sensor unit that measures process variables and changes the process conditions. It includes a control unit and an agent unit for controlling the data flow, the cloud server has a preprocessor, a database, and a machine learning algorithm, the edge cluster and the control server are configured to exchange information through an MQTT broker, and the agent unit is a standard It is configured to extract the measurement data in a steady state within the range and transmit it to the cloud server, the cloud server is configured to transmit an optimal model derived through a machine learning algorithm to the control server, and the control server is based on the optimal model. A control command for molding conditions is generated and transmitted to the edge cluster, and the edge cluster is configured to enable control of driving means related to molding conditions of the injection molding machine according to the control command.

In addition, in the integrated management system of the AIoT-based injection manufacturing facility of the present invention, the agent unit consists of at least one agent, each agent includes a communication unit, a processing unit, an execution engine, and a DB, and the processing unit includes measurement data received from the sensor unit It is characterized in that it is configured to determine whether it is in the reference range by comparing the reference data stored in the DB and the reference range, and to transmit only the measured data determined to be in a normal state within the reference range to the cloud server.

In addition, the integrated management system of the AIoT-based injection manufacturing facility of the present invention converts the normal state measurement data received by the pre-processing unit 504 of the cloud server into a structure of a learning pattern pair and supports accumulated storage in the database. do it with

In addition, the integrated management system of the AIoT-based injection manufacturing facility of the present invention classifies and groups the measured data of the sensor unit according to the physical characteristics of the process variables, and the optimal model ( 503), a new learning pattern pair is constructed based on the optimal model of the first stage to enable the derivation of the optimal model through the second stage of machine learning, and classified according to the physical characteristics. The grouped process variables include temperature, humidity, injection speed, and pressure.

On the other hand, the method of operating an integrated management system for an AIoT-based injection manufacturing facility of the present invention includes the steps of starting production after the injection molding machine is ready for operation, measuring process variables through the sensor unit of the edge cluster, and processing the measured data. The step of determining whether the measured data is in a normal state by comparing it with the set reference value collected in the Transmitting the optimal model derived from machine learning using data to the control server, the control server generating a control command based on the AI optimal model and delivering it to the edge cluster through the MQTT broker, driving the control means of the edge cluster It is characterized in that it is made including the step of executing the control command through the.

In addition, in the method of operating the integrated management system of the AIoT-based injection manufacturing facility of the present invention, the measurement period of the sensor unit is set in the operation preparation stage, and the measurement period is the shortest process time among the process times for each step included in the injection process time of one cycle. It is characterized in that it is set based on .

In addition, in the method of operating the integrated management system of the AIoT-based injection manufacturing facility of the present invention, the step of determining whether the normal state is in the reference range by comparing the measurement data transmitted from the sensor unit with the reference data stored in the DB is determined by the processing unit and, when it is determined that the measurement data transmitted from the sensor unit greatly deviates from the reference range, the step of sending an alert is further included.

In addition, the method of operating the integrated management system of the AIoT-based injection manufacturing facility of the present invention classifies and groups the process variables according to the physical characteristics of the measurement data of the sensor unit, and uses an agent specialized for each data group and a machine learning algorithm for each group in one step. After deriving the optimal model of , the step of deriving the integrated optimal model of the second stage through machine learning of a new learning pattern pair based on the optimal model of the first stage is further included.

According to the invention, it is possible to automatically remote control beyond the existing management method that relied only on the skill level of the operator, and effectively respond to the realistic demand for physical property improvement through a new molding material, and the occurrence of defective products due to human error or inexperienced management The effect of preventing a decrease in productivity can be expected.

In addition, it is possible to improve the efficiency and stability of data processing through appropriate function allocation of IT resources, and it is possible to manage sophisticated molding conditions, thereby omitting a part of the finishing process such as surface treatment of manufactured products, and It has the effect that it is easy to deal with factors that cause defects such as whitening.

1 (a), (b) is a configuration diagram for explaining a prior art artificial intelligence (AI) based injection molding system.
2 is an overall configuration diagram for explaining an integrated management system of an injection manufacturing facility according to a preferred embodiment of the present invention.
3 is a configuration diagram for explaining the function of the agent employed in the present invention.
4 is an exemplary diagram for explaining the function of the MQTT broker adopted as the information communication protocol in the present invention.
5 is an exemplary diagram for explaining the machine learning algorithm of the cloud server used in the present invention.
6 is a flowchart for explaining the operation method of the integrated management system for the injection manufacturing facility of the present invention.

Hereinafter, an integrated management system for an AIoT-based injection manufacturing facility and an operating method thereof according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

For reference, in describing the present invention, if it is determined that a detailed description of a known configuration or function may unnecessarily obscure the gist of the present invention, a detailed description thereof will be omitted. And the technical terms used in the present invention are only used to describe specific embodiments and are not intended to limit the present invention, and the technical terms used in the present invention are not specifically defined in the present invention unless otherwise specified. It should also be interpreted in a meaning generally understood by those of ordinary skill in the art to which the invention pertains, but should not be interpreted in an excessively reduced or comprehensive sense.

1 (a), (b) shows a conventional artificial intelligence (AI) based injection molding system (Prior Patent 1). 1(a) shows a structure in which a molding condition generating device 200 is attached to the injection molding machine 100, and FIG. 1(b) shows a detailed configuration of the molding condition generating device 200. As shown in FIG. The molding condition generating device 200 of FIG. 1 (b) is characterized in that it is configured to change the molding conditions by determining whether the product produced by the injection molding machine 100 maintains the target molding quality, and the injection data acquisition unit 710 , a database 750 , a model generator 740 , and a software engine 735 are included, and the model generator 740 has a structure in which a molding quality maintenance model is derived through AI learning.

However, the prior art of Figures 1 (a), (b), as mentioned above, a certain degree of pure function in that the molding conditions are controlled through the molding condition generating device 200 attached to the injection molding machine 100 . can be expected, but it is a structure in which a separate molding condition generating device 200 must be attached to each injection molding machine 100, and a database 750 for supporting artificial intelligence (AI) learning for each molding condition generating device, AI-based Since it is inevitable to have information processing resources (IT resources) such as the model generator 740 and software engine 735 of It has a structural problem that not only is it hindered, but also the increase in the installation cost cannot be avoided.

The present invention is focused on the problems of the prior art as described above, and FIG. 2 shows a system according to a preferred embodiment of the present invention. The integrated management system of FIG. 2 is characterized in that it utilizes the hardware resources of the existing injection molding machine related to sensors and control means as much as possible while efficiently distributing IT resources through specialized solutions.

More specifically, the integrated management system of FIG. 2 includes an injection molding machine 10, an edge cluster 20, an MQTT (Message Queuing Telemetry Transport) broker 30, a control server 40, and a cloud server 50. It is included as the main component. At this time, the injection molding machine 10 is provided with a sensor unit 101 for measuring various injection process variables and a controller 102 capable of controlling the process variables according to an external signal. And the injection molding machine 10 and the edge cluster 20 are connected through a wired or local area network, and the edge cluster 20, the MQTT broker 30, the control server 40 and the cloud server 50 use the Internet protocol. It is functionally connected to enable information exchange using In addition, although not specifically shown in FIG. 2 , as shown in FIG. 5 , the cloud server 50 has an artificial model that supports the derivation of the optimal model 503 through machine learning based on the accumulated data stored in the database 501 It can be said that it is a technical feature that an intelligent (AI) algorithm 502 is provided, and the control server 40 has an integrated dashboard that can monitor a number of injection molding machines 10 installed at the production site as a whole. there is.

In general, Internet of Things (IoT) technology is premised on supporting data exchange between devices by using sensors, wired/wireless communication and network infrastructure, servers, and human machine interface (HMI) technology. In particular, in the Industrial Internet of Things (IIoT) applied to industrial sites, data is collected from manufacturing equipment, and the collected data is transmitted to a server through a communication network, undergoes data processing, and after deriving meaningful output variables, the output variables The process conditions are controlled using Or, it is a driving force that promotes the release of various open source-based APIs (Application Programming Interfaces).

In response to such a technology development trend, the integrated management system of the present invention shown in FIG. 2 efficiently distributes the functions of IT resources, adopts a user-friendly Human Machine Interface (HMI) solution, and at the same time artificial intelligence (AI) ) to propose a more intelligent integrated management system through the grafting of learning technology.

Hereinafter, each function of the main configuration included in the integrated management system of the present invention will be described in detail with reference to FIGS. 2, 3, 4 and 5 .

First, the sensor unit 101 of the injection molding machine 10 shown in FIG. 2 is provided with a plurality of sensors, so that the temperature of the cavity, core, hot runner, etc., the internal pressure of the mold, the injection speed, the temperature gradient of the mold or the hopper is mounted It is responsible for measuring process variables such as humidity of the molded molding material, and the control unit 102 operates heating, cooling, increasing pressure, decompression, acceleration, deceleration, humidification, dehumidification, etc. according to the control command of the control server 40 . It is responsible for the function of controlling the related driving means to make this happen. The sensor unit 101 and the control unit 102 can utilize the existing hardware resources to the maximum, and the number and types of control means and sensors can also be appropriately selected as well-known general-purpose means according to need. there will be no

In addition, although only one agent may be provided in the agent unit 103, it is preferably designed in a structure including a plurality of agents 21 and 22 in consideration of process variables and the characteristics or number of control objects. For example, it would be desirable to design agent 1 as a molding condition agent, agent 2 as a temperature control agent, agent 3 as a pool-proof agent, etc. to implement a function suitable for the characteristic. It would be desirable to design such a plurality of agents in the form of modules to support disassembly and assembly as needed, such as a Raspberry Pi-based computer (PC), so that individual agent functions can be selected. .

Meanwhile, each agent is designed to include a communication unit 211 , a processing unit 212 , an execution engine 213 and a database 214 as illustrated in FIG. 3 . Here, the communication unit 211 supports data communication between the cloud 50 and the MQTT broker 30, and the information exchange between each agent is a message-based known Message-Oriented Middleware (MOM) solution (API). designed to support In addition, the database 214 of the agent 21 is designed to support the storage of reference data for molding conditions and cookie information (Cookie file) of the measurement data transmitted from the sensor unit 101, and the processing unit 212 is After determining whether it is in the reference range by comparing the measurement data transmitted from the sensor unit 101 with the reference data stored in the DB 214 , the measured data determined to be in a normal state within the reference range is extracted and sent to the cloud server 50 . It is designed to support the transmission function. In this case, the processing unit 212 may be designed to support a data purification function such as removing noise included in the measurement data or extracting a feature, and a periodic update function of log data stored in the DB 214 .

And the execution engine 213 provided in the agent supports the function of delivering (Publish) topic information and measurement data to the MQTT broker 30, and subscribes to the control commands retrieved through the MQTT broker 30 Thus, it is designed to have a function of controlling the process conditions of the injection molding machine 10 according to the control command.

Figure 4 is an exemplary diagram for explaining the function of the MQTT broker (30). The MQTT protocol is generally adopted as an ISO standard to support exchange of information in a conversational format between terminals, and the MQTT protocol applied in the industry is a local server that collects measured sensor data and control Its function as a broker that mediates a server (Client Server) is well known. The MQTT broker 30 of FIG. 4 collects the data measured by the sensor unit 101 in accordance with the measurement period from the agent 21 and groups them according to the physical characteristics of the measured data to create a topic, and the topic When the measurement data is transmitted (Publish) to the MQTT broker 30 together with (Publish) is in charge of the intermediary function.

When the function of the MQTT broker 30 is described with reference to FIG. 4 , the measurement data transmitted from the sensor unit 101 is combined with topic information through the function of the execution engine 213 provided in the agent to the MQTT broker 30 . After being transmitted (Publish), the control server 40 subscribes to topic information and measurement data through the MQTT broker 30, and then controls generated based on the subscription data and an artificial intelligence (AI) model to be described later. When the command is sent back to the MQTT broker 30 (Publish), the execution engine 213 of the agent 21 subscribes to the control command and controls the driving means related to the control command, thereby controlling the control server 4 It is possible to change the process conditions for the injection molding machine (10) remotely in the.

The reason for adopting the MQTT protocol in the present invention is that, despite partial disadvantages such as time delay and data loss that occur in the data processing process, data related to a topic can be easily checked visually through an integrated dashboard, It is intended to take full advantage of the advantage that fast and accurate message exchange (information exchange) is possible through the process of publishing and subscribing in the Q&A method. Accordingly, the present invention has an effect that can satisfy expectations for productivity increase and quality improvement, which are important goals in the field of injection molding manufacturing due to the advantage of information exchange through the MQTT broker 30 .

5 shows the configuration of the cloud server adopted in the present invention.

Cloud computing is a widely used technology to maximize the expected value or output value of services by efficiently distributing and operating internal and external IT resources and to solve the heat problem caused by excessive power consumption.

As illustrated in FIG. 5 , the cloud server 50 used in the present invention receives the measurement data determined to be in a normal state from the processing unit 212 of the agent 20 , and accumulates and stores the database 501 and the accumulated data. An AI algorithm 502 for deriving an optimal model 503 by using as learning data is provided.

Here, the AI algorithm 502 learns certain signals and patterns from data without human assistance based on mathematical optimization or statistical analysis techniques, etc., and a machine learning algorithm that supports to derive an optimal molding condition model based on it. However, since such a machine learning algorithm can utilize various open sources that are already widely used, there will be no special technical difficulties in implementing the AI algorithm of the present invention. And, the machine learning method of the present invention includes supervised learning that informs and learns both the problem (input) and the correct answer (result), unsupervised learning that learns without teaching the correct answer (result), and a reward system It is okay to use any method of reinforcement learning that learns using ) would be preferable.

On the other hand, the injection molding manufacturing process premised in the present invention is a one-cycle process by sequentially passing the processes such as melting of the molding material, injection of the molten molding material into the mold, cooling the mold, product extraction (release), and finishing. In order to produce a good product, it is necessary to precisely control process variables such as temperature, humidity, injection speed, internal mold pressure, and time during these processes.

Among such process variables, the AI machine learning process applied to the present invention is described based on the temperature variable. First, the main temperature variables affecting product quality in injection molding manufacturing are the ambient temperature of the injection molding machine 10, the melting temperature of the molding material. , mold temperature, etc. are representative. Such temperature data is measured through the sensor unit 101 in accordance with the measurement cycle, and the measured data is transmitted from the processing unit 212 of the agent 21 and compared with preset reference data. Only the measurement data determined to be in a normal state that does not deviate from the reference range is extracted and stored in the database 501 of the cloud server 50 .

At this time, the weekly measurement data stored in the database 501 is pre-processed to have a structure of a training pattern pair so that it can be utilized as AI learning data. For example, when it is assumed that three measurement data of the ambient temperature (x), melting temperature (y), and mold temperature (z) of the injection molding machine are needed to control the temperature condition of the injection molding machine 10, the database ( Data stored in 501) is converted into a pattern pair structure of (x, y, z), and data measured according to a set period is sequentially stored in the DB 501 in the structure of the learning pattern pair above. And, the target value of the machine learning algorithm 502 sets the highest quality index among the quality indexes for each grade of the molded product as a reference value.

In such a situation in which the accumulated learning data and the target value are secured, the machine learning algorithm 502 receives the learning pattern pairs stored in the database 501 and correlates which combination of the learning pattern pairs is closest to the target value (the highest quality). Through continuous learning, an optimal model (AI model) is derived. Of course, in the initial stage in which the injection molding machine 10 is newly installed, the reliability of the optimal model derived from the lack of accumulated data may not reach the expected level. The effect can be expected, and if the accumulated data of the past for the injection molding machine 10 of the same standard is applied to the newly installed injection molding machine and can be utilized, it leads to the effect of increasing the amount of learning data, thereby enhancing the reliability of machine learning. It will be helpful.

And, as in the above example described with a focus on temperature variables, after deriving each optimal model through specialized machine learning for process variables such as humidity, injection speed, and pressure inside the mold, a new model was developed based on the optimal model for each process variable. By generating and storing a training pattern pair and performing the second stage of machine learning, an optimal model (AI model) in which the entire process variables of the injection molding machine 10 are reflected will be finally derived. The AI model derived through such a process is transmitted to the control server 40 to support generation of control commands for process variables of the injection molding machine 10 .

Here, the description of the machine learning process of the present invention as a configuration performed by the cloud server 50 is that the cloud computing environment can easily accommodate advanced applications such as artificial intelligence (AI), and large-scale workloads such as big data processing It is advantageous for the expansion of IT resources to cover the If necessary, an edge computing environment or a hybrid computing environment may be used. For example, a person of ordinary skill in the art would not insist on using a cloud computing environment while operating only one injection molding machine 10 .

6 is a flowchart for explaining the operation method of the integrated management system for the injection manufacturing facility of the present invention. In order to operate the integrated management system of the present invention, first, in the standby stage, a measurement cycle for process variables such as temperature and pressure, a reference value stored in the DB of the agent 20, and initial process conditions are set. In this case, it would be desirable to set the measurement cycle of the process variable variably within the time range required to complete one cycle of the injection process. For example, if the measurement cycle is set based on the shortest step among the process time of one cycle, which consists of steps such as melting, injection, pressurization, cooling, and extraction of molding materials, improved technology related to rapid heating and rapid cooling is applied. This will shorten the process time of one cycle and have the advantage of being able to respond flexibly to the general trend of the injection manufacturing industry that seeks to improve productivity through such time reduction. Of course, it will be natural that the pre-preparation step includes the process of loading the molding material into the hopper or the process of preheating the injection molding machine 10 so that it can immediately enter the normal operation level.

Next, the injection molding machine 10 is operated in earnest to start product production. At this time, if the injection molding machine 10 is a new machine, it starts production according to the standard value and process conditions set at the beginning, and if it is a machine with an operation history in the past, it inherits the reference value and process conditions stored in the last operation process and starts production. At the same time, data of various process variables are measured through the sensor unit 101 . Such measurement data is transmitted to a specific agent, and the measurement data collected by the agent is grouped by classifying process variables according to physical characteristics such as temperature, humidity, injection speed, and mold pressure, and agents (21, 22) specialized for each data group. ) is allocated and operated so that it can be processed. In this case, it is desirable to use an open source-based MOM (Message-Oriented Middleware) solution for information exchange between each agent.

And the processing unit 212 of each agent compares the measurement data of each sensor with the reference value of the DB 214 stored in advance to determine whether it is within the set reference range, and if it is determined that the normal state is within the reference range, the measurement data of the normal state to be stored in the DB 501 of the cloud server 50, and the accumulated data stored in the DB 501 provides a basis that can be utilized as machine learning data. On the other hand, the processing unit 212 of each agent sets the measurement data out of the set reference range and the measurement data classified according to the physical characteristics so that the MQTT broker 30 can publish it together with the topic. Support.

In the next step, the cloud server 50 processes the steady-state data received from the agents 21 and 22 to have the structure of a training pattern pair, sequentially accumulating and storing it in the DB 501, and the accumulation It supports to derive the optimal model 503 by using the pattern pair of data as the training data of the AI algorithm 502 .

In the next step, the control server 40 receives and manages the optimal model 503 derived from the cloud server 50, and subscribes to the measurement data published by the MQTT broker 30. And after analyzing the control level based on how far the subscribed measurement data deviates from the optimal model 503, if the analysis result is at a controllable level, a control command for bringing the current molding conditions closer to the optimal model 503 Publishes to the MQTT broker 30, and the MQTT broker 30 provides a basis for subscribing to the control commands from the agents 21 and 22, and each agent 21, 22 It supports to drive the control means related to the subscribed control command through the execution engine 213 .

On the other hand, if the control level analysis result is an uncontrollable level, the alarm is activated and the operation of the injection molding machine 10 is immediately stopped. This is because, due to the nature of the molding material that cannot be reused, it is necessary to reduce the waste of the molding material as much as possible. And it is desirable that the alarm operated at this time has a different alarm method depending on the severity of emergency measures. For example, when immediate emergency action is required, alarm means such as a warning light, speaker, siren, etc. are used, and when preventive or cautionary emergency action is required, the display provided on the integrated dashboard of the control server 40 ( If the warning message is delivered to the manager using the monitor), it can be expected to have the effect of alleviating the fatigue of the manager caused by excessively frequent alarm sounds.

According to the integrated management system and operation method of the present invention as described above, the efficiency and productivity of the integrated management can be improved through appropriate function allocation of IT resources, and the operation state of the injection molding machine 10 through the control server 40 Since the process conditions can be controlled while monitoring the equipment remotely, it is effective in reducing the occurrence of defective products due to human errors or inexperienced management by breaking away from the existing management method that relied only on the skill level of the workers.

In addition, since subsequent processes such as surface treatment of products can be partially omitted through sophisticated molding condition management, it is possible not only to increase production per unit time, but also to easily cope with defects-inducing factors such as weld defects or whitening. This has the effect of improving the quality of the manufactured product.

Although the preferred embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and any person skilled in the art may make various modifications within the scope of the present invention without departing from the gist of the present invention. It will be possible. However, it should be noted that whether it departs from the gist of the present invention is determined solely by the description of the claims.

10: injection molding machine 101: sensor unit 102: control unit 103: agent unit
20: edge cluster 21, 22: agent 211: communication unit
212: processing unit 213: execution engine 214: agent DB
30: MQTT broker 40: control server 50: cloud server
501: DB 502: AI algorithm 503: optimal model 504: preprocessor

Claims (12)

In the integrated management system of the AIoT-based injection manufacturing facility in which the edge cluster 20 installed in the injection molding machine 10, the control server 40, and the cloud server 50 are connected through a communication network,
The edge cluster 20 includes a sensor unit 101 for measuring process variables, a control unit 102 for changing process conditions, and an agent unit 103 for controlling data flow,
The agent unit 103 includes at least one or more agents 21 and 22, and each of the agents 21 and 22 includes a communication unit 211, a processing unit 212, an execution engine 213, and a DB 214. is provided,
The processing unit 212 compares the measurement data transmitted from the sensor unit 101 with the reference data stored in the DB 214 to determine whether it is within a reference range, and returns the measurement data determined to be in a normal state within the reference range. It is configured to be extracted and transmitted to the cloud server 50,
The cloud server 50 includes a preprocessor 504, a database 501, and a machine learning algorithm 502,
The pre-processing unit 504 converts the received measurement data determined to be in the normal state into a structure of a learning pattern pair and is configured to be accumulated and stored in the database 501,
The cloud server 50 is configured to transmit the optimal model 503 derived through the machine learning algorithm 502 to the control server 40,
The control server 40 is configured to generate a control command for molding conditions based on the optimal model 503 and transmit it to the edge cluster 20,
The edge cluster 20 is configured to control the driving means related to the molding conditions of the injection molding machine 10 according to the control command,
The edge cluster 20 and the control server 40 are configured to exchange information through the MQTT broker 30, an integrated management system for AIoT-based injection manufacturing facilities delete delete delete According to claim 1,
The measured data of the sensor unit 101 is grouped by classifying process variables according to physical characteristics, and after deriving an optimal model 503 of the first stage through an agent specialized for each data group and a machine learning algorithm,
An integrated management system for AIoT-based injection manufacturing facilities, characterized in that a new learning pattern pair is created based on the optimal model of the first stage, and then an integrated optimal model can be derived through the second stage of machine learning 6. The method of claim 5,
An integrated management system for AIoT-based injection manufacturing facilities, characterized in that the process variables classified and grouped according to the physical characteristics include temperature, humidity, injection speed, and pressure According to claim 1,
The processing unit 212 further includes an alarm generating means capable of sending an alarm when the measured data transmitted from the sensor unit 101 greatly deviates from the reference range. In the operation method of the integrated management system of the AIoT-based injection manufacturing facility in which the edge cluster 20 installed in the injection molding machine 10, the control server 40, and the cloud server 50 are connected through a communication network,
After the operation preparation step of the injection molding machine 10, starting production,
measuring a process variable through the sensor unit 101 of the edge cluster 20;
Collecting the measured data in the processing unit 212 of the edge cluster 20 and comparing it with a set reference value to determine whether it is in a normal state;
In the step of determining whether the normal state is in the normal state, the measured data determined to be in a normal state is transmitted to the cloud server 50, and the preprocessor 504 of the cloud server 50 converts it into a structure of a learning pattern pair and converts it into a database ( 501) to accumulate and store,
transmitting the optimal model 503 derived by machine learning using the accumulated data of the database 501 to the control server 40;
The control server 40 generates a control command based on the optimal model 503 and transmits it to the edge cluster 20 through the MQTT broker 30;
A method of operating an integrated management system for an AIoT-based injection manufacturing facility, comprising the step of executing the received control command through the control means driving function of the edge cluster 20 9. The method of claim 8,
In the operation preparation step, the measurement period of the sensor unit 101 is set,
The measurement cycle is an operating method of an integrated management system for AIoT-based injection manufacturing equipment, characterized in that the shortest process time among the step-by-step process times included in the injection process time of one cycle is set as a standard 9. The method of claim 8,
In the step of determining whether the state is in the normal state, the processing unit 212 of the edge cluster 20 compares the measurement data transmitted from the sensor unit 101 with the reference data stored in the DB 214 to determine whether the edge cluster 20 is in the reference range. A method of operating an integrated management system for AIoT-based injection manufacturing facilities, characterized in that 9. The method of claim 8,
In the step of transmitting the optimal model 503 derived by machine learning using the accumulated data of the database 501 to the control server 40, process variables are selected according to the physical characteristics of the data measured by the sensor unit 101. After classifying and grouping, the first-stage optimal model 503 is derived through an agent specialized for each data group and a machine learning algorithm,
Operation of an integrated management system for AIoT-based injection manufacturing facilities, characterized in that the step of deriving an integrated optimal model of the second stage by performing machine learning with a new learning pattern pair based on the optimal model of the first stage is further included method 9. The method of claim 8,
The step of determining whether the normal state is in the AIoT-based injection manufacturing facility, characterized in that the step of sending an alarm when it is determined that the measurement data transmitted from the sensor unit 101 is significantly out of the reference range How to operate the integrated management system

Images