Page 115 - Kỷ yếu hội thảo quốc tế: Ứng dụng công nghệ mới trong công trình xanh - lần thứ 9 (ATiGB 2024)
P. 115

106                              TRƯỜNG ĐẠI HỌC SƯ PHẠM KỸ THUẬT - ĐẠI HỌC ĐÀ NẴNG


                                AI-BASED PROCESS OPTIMAL
                 CONTROL SYSTEM FOR QUALITY TRACKING

                                           AND PREDICTION



                                                                                 1*
                                                                 2
                                          Song-Hee Park , Young-Se Lee , Young-Hyung Kim
                                                     1
                   1 Department of Electronic Engineering, Kumoh National Instityte of Technolog, Gumi,Korea; psh0203@kumoh.ac.kr
                                            2 AJUSTEEL, Gumi, Korea; young@ajusteel.co.kr
                                            *Young-Hyung Kim (email: kic126@kumoh.ac.kr)
                  Abstract: The AI-based process optimal control system   methods for a smart factory applying an AI-based process
               is  a  control  system  technology  that  combines  next-  optimization control system.
               generation new technologies such as autonomous and semi-  Keywords:  Quality  Management  System;  Predictive
               autonomous  control  algorithms,  intelligent  HMI,  sensors,   management;  Artificial  intelligence;  Control  system;
               and networks. In this paper, we discuss efficient operation   Quality tracking.
               plans  and  quality  tracking  and  predictive  management
                  1.  INTRODUCTION                            system can be built in the facility as shown below, and
                  There are many ways to run a factory efficiently.   the following effects can be achieved.
               If  you  apply  an  AI  appearance  inspection  algorithm   2.1. Build an intelligent manufacturing environment
               that  applies  deep  learning  to  the  quality  inspection   for self-diagnosis and improvement
               process, the equipment can detect product defects on   It detects the causal process that causes defects by
               its own without a human inspection, and if you build
               a process monitoring system in the equipment, it can   reflecting  the  sequential  manufacturing  process
               notify  the  user  in  advance  when  it  detects  an   situation,  visualizes  it,  and  presents  explainable
               abnormality in the equipment. Accidents and defects   process   management   guidelines.   Through
               can  be  prevented.  Efficient  management  is  possible   visualization,  equipment  can  express  the  relationship
               because the process flow can be viewed in real time.   between  parameters  collected  in  real  time  through  a
               In this paper, we discuss ways to build a smart factory   network, and through this, key equipment parameters
               that can increase productivity by applying the optimal   for diagnosing equipment status can be determined. In
               system to the quality process to remotely manage the   the  inspection  process,  a  deep  learning-based  defect
               process,  reduce  defect  rates,  and  increase  operating   activation  map  is  developed  to  identify  the  cause  of
               efficiency.  If  a  smart  factory  is  introduced,  it  can   product  defects,  and  the  defect  judgment  text  is
               overcome the limitations of manual work, increase the   generalized  to  provide  insight  to  field  workers.  By
               efficiency  of  the  production  system,  and  can  be   checking this, field workers recognize the presence or
               applied  to  most  factory  processes,  so  it  can  be   absence of abnormalities.
               expected to not only secure product quality but also   2.2.  Build  a  responsive  prediction  system
               operate efficiently overall.                   applicable to various environments

                  2.   AI-BASED     PROCESS      OPTIMAL         We  develop  a  monitoring  method  based  on  a
               CONTROL SYSTEM                                 process  progress  prediction  model  that  can  detect
                  Most  manufacturing  processes  that  mass  produce   stagnation  and  abnormalities  in  the  sequential
               small varieties mainly use sequential processes such as   manufacturing  process  early.  It  is  equipped  with  a
               a conveyor system. This process has the advantage of   deep  learning model to  enable  early  detection of  the
               being  able to operate  efficiently because  the labor  is   time and cause of failure of equipment that performs
               simple and production management is easy. However,   the  process.  In  the  event  of  a  breakdown  or
               due  to  repeated  simple  labor,  worker  fatigue  is  high   abnormality, it notifies the user or stops the operation
               and if a problem occurs in the front process, the back   of  the  equipment  to  prevent  defects,  improving  the
               process  cannot  proceed.  As  a  result,  production   yield  of  the  process  and  improving  the  equipment's
               stagnation and abnormal phenomena occur, which has   performance. Improves reliability and productivity.
               the  disadvantage  of  causing  serial  defects.  If   2.3. Optimization-based manufacturing process
               production is delayed and defects occur, the yield of   control to improve manufacturing efficiency
               the process drops, leading to a large loss. To solve this   Infer  and  verify  control  parameters  based  on
               problem,  a  self-diagnosis  and  responsive  prediction
                                                              historical  data  for  process  optimization.  In  order  to
                                                                                   ISBN: 978-604-80-9779-0
   110   111   112   113   114   115   116   117   118   119   120