[Factory Control | Factory Analytics | Factory Design | Miscellaneous]
Factory Control

   - New Control Program Language for SW-Oriented & PLC-less Control
   - Execution Neural Network (ENN) : A Deep Learning Model to Train and Generate Control Logics for Autonomous Manufacturing Systems
   - Digital Twin Simulation for Lookahead Intelligence
   - Workcell Agent for Autonomous Control

Execution Neural Networks (ENN) & MEL-Generator (MEL-Gen): Filab is developing a new AI that has the capability to autonomously generate MEL programs.
  • ENN is a deep learning model for control logic learning using MEL Execution Feature Matrix
  • MEL-Gen is an ENN-learned AI for Control Logic Generation

Execution Neural Networks

Manufacturing Execution Language (MEL): MEL is a new user-friendly programming language that enables the design, execution, and validation of all manufacturing operations under a software-oriented architecture. FiLab is currently developing a new task & flow-based control diagram, called MEL-diagram, to mathematically represent manufacturing processes and their control logics. To do this, we first define a hierarchical structure of Manufacturing Execution Entities, and furthermore, devise Manufacturing Execution Feature Matrices to facilitate advanced algorithm-based higher-level logic programming. These matrix definitions play a crucial role in AI training, empowering autonomous generation of control programs.


Manufacturing Execution Language

Workcell Agent for Autonomous Control : Filab is developing agents for autonomous control in workcells and a structure to facilitate their easy configuration. These workcell agents contribute to addressing labor shortages and meeting market demands through autonomous optimal control. Workcell agents possess the following capabilities: perception and control of manufacturing equipment included in the workcell, decision-making and optimal control within the workcell based on internal information, and higher-level decision-making and optimal control through communication among agents.


Workcell Agent for Autonomous Control

Digital Twin - look ahead intelligence: FiLab' Digital Twin approach emphasizes look ahead intelligence armed with model-driven simulation and data-driven learning & prediction. To do this, standardized information & interoperable neutral data format must be developed. Furthermore, seamless information synchronization between real and digital systems must be provided by a super-efficient communication protocol.


Digital Twin

Factory installation wizard: Manufacturing paradigm has been changed from mass production, to mass customization, and to personalization. To meet this paradigm shift, we are developing 'Factory Installation Wizard' for rapid factory configuration and implementation. The Factory Installation Wizard consists of the following main six steps: layout design, controller configuration, 3D factory modeling and control logic design, factory-in-the-loop simulation, factory OS installation, test and calibration.

Factory Installation Wizard

Manufacturing process monitoring: Scheduling and planning are the central functions to increase the productivity in manufacturing. In a shop floor, these functions should be deployed in a real-time manner by considering the dynamics conditions of manufacturing processes. In this regard, the prerequisite is seamless manufacturing process monitoring to acquire live workplace data.

Design form failure

Self-resilient process control: This research aims to integrate (i) automatic root cause analysis of product and process faults, (ii) prediction of the product/process quality for a given system, and (iii) feedforward/feedback process adjustment & control to enhance the system's response to fault or quality/productivity degradation. This integration of in-process quality monitoring with closed-loop process control will make manufacturing processes more flexible, tolerable, and resilient to unforeseen changes and events.

Factory Installation Wizard

Smart machine and fixture: According to personalization, jig and fixture system should be usually redesigned in advance. Since it spends a lot of time and costs, it is neceesary to develop a reconfigurable jig and fixture system for any kinds of products. Therefore, this research aims to constrcut (i) transformable jig and fixture system, and (ii) its control software including optimal path generations, feedback-position control.

transformable jig and fixture

©2020 Factory Intelligece Lab POSTECH
Cheongam-Ro 77, Pohang, South Korea, 37673
Tel:+82 54 279 8261 Fax:+82 54 279 2870