AAS Asset Definition
Facilities, sensors, normal ranges, relationships and manuals are organized into a standard asset model.
Digital Twin
Connect actual facility and process data to an AAS-based asset model and pre-verify facility and process abnormalities and quality changes in a virtual environment.
OVERVIEW
Not a static 3D replication, An operating model that moves with field data.
Manufacturing digital twin is a structure that connects equipment properties, sensors, normal range, and operation history to virtual assets, and verifies abnormal conditions and response results through simulation. Because the health of real assets and virtual models is linked, you can expand beyond monitoring to prediction and pre-verification.
Define assets, synchronize field data, reproduce anomalies in a virtual environment, and connect to field response.
On-site data synchronization
Connect PLC · Sensor · MES · LIMS · Quality data with real assets.
Normal/abnormal state verification
Inject conditions to reproduce expected changes in equipment and processes in a virtual environment.
Judgment/Response Connection
It fuses sensors and visual opinions and connects them to action, approval, and task execution.
Actual facility → virtual verification → on-site responseVerification/response operation loopCreate .
CHALLENGES
Abnormal situations are visible only after they occur, It is difficult to repeat verification in actual facilities.
Manufacturing sites have different data formats and normal ranges for each facility, and it is costly and risky to reproduce failures and quality deviations on actual lines.
Facility information and operating standards are scattered.
Sensor items, normal ranges, process relationships and manuals are distributed across different systems and personnel. Digital models are difficult to create consistently.
It is difficult to reproduce abnormal conditions in actual equipment.
Repeated testing at high temperatures, vibrations, speed changes and quality deviations can lead to production interruptions, safety hazards and loss of raw materials.
Sensor anomalies and visual anomalies are analyzed separately.
If sensor values and vision inspection results are separated, it is difficult to comprehensively determine internal defects that are not visible on the outside.
There is insufficient data to learn about failures and quality deviations.
Compared to normal data, there are fewer actual abnormal images and failure cases, which limits the verification scope of predictive maintenance and quality prediction models.
Analysis results do not directly lead to on-site response.
Even after detecting an abnormality, the person in charge must find the manual, review the cause, and perform separate action and approval procedures.
SOLUTION
Connect asset model, simulation and AI Agent Results are verified before field application.
01
AAS-based asset standardization
Define facility attributes, sensors, normal ranges, relationships and operational documentation with an AAS-based asset specification to create a common baseline for physical assets and virtual models.
02
Real-time connection to field data
Connect PLC · SCADA · MES · LIMS · quality data with facility assets and accumulate time flow and status changes of batch units.
03
Scenario-based facility simulator
By injecting temperature, humidity, vibration, spray amount and speed conditions, normal and abnormal operation and process changes are repeatedly verified in virtual facilities.
04
Creating future states based on NVIDIA Cosmos
Create future states based on sensors and images, Compare changes to steady state.
05
Sensor/Visual Fusion Judgment
By analyzing time series abnormalities and vision inspection results together, we support comprehensive judgment including internal defects that are not visible from the outside.
06
Agent response and verification history
SAMANDA analyzes the cause, suggests or implements actions, and leaves human approval and audit records at necessary steps.
ARCHITECTURE
Synchronize real and virtual assets, Connect simulation results to operational decisions.
KosmosAI manages AAS assets and simulators, and AkashiQ · ORKESTRIX · SAMANDA · IKIN is responsible for data, execution environment, judgment, and workflow.
PRODUCT ROLES · SERVICE FLOW
Digital twin service flow linking product roles
Centered on KosmosAI, we connect actual facility data and virtual verification results to on-site response.
DIGITAL TWIN
KosmosAI
Integrated management of AAS assets, digital twins, and scenarios.
DATA FOUNDATION
AkashiQ
Stores and manages sensor, MES, quality data and lineage.
AI EXECUTION
ORKESTRIX
Provides an on-premises GPU Kubernetes and model execution environment.
AI OPERATIONS
SAMANDA
Responsible for status observation, cause analysis, action decision and implementation.
WORKFLOW
IKIN
Connects manual RAG, approval, and work workflow.
AAS Asset Model
Asset Specification · Relationship · Normal Range · Manual
Data Synchronization
MES · LIMS · Quality · Time series data
Virtual Twin
Injection of abnormal conditions · Creation of expected state · Comparison verification
AI Agent
Cause analysis, action proposal, approval, implementation
BUSINESS VALUE
Verify first without stopping on site, Accelerates response to abnormalities and quality improvement.
01
Risk-free pre-verification
Conditions and response results are repeatedly tested in a virtual environment without causing problems in actual equipment.
02
Accelerated analysis of cause of abnormalities
Reduce the time spent checking multiple systems by connecting assets, sensors, video, and manuals.
03
Secure synthetic abnormal data
It compensates for the lack of training and verification data by generating rare failure and quality deviation scenarios.
04
Predictive maintenance and quality prediction
We analyze the correlation between equipment abnormalities and product quality to predict failures and quality deviations in advance.
05
Standardize asset knowledge
Based on AAS, facility properties, normal ranges, relationships and operational knowledge are accumulated as standard assets.
06
On-premise operations
It is operated in the GPU environment inside the factory without exporting sensitive equipment and production data to the outside.
APPLICATIONS
설비 · 공정 · 운영 변화를 실제 적용 전에 가상 환경에서 검증합니다.
현장 데이터와 자산 모델을 연결해 조건 변화와 이상 시나리오가 운영과 품질에 미치는 영향을 미리 확인합니다.
제약 · 바이오 디지털 트윈
설비 · 공정 · 품질 데이터를 바탕으로 배치 조건 변화와 이상 시나리오를 가상 환경에서 검증합니다.
- 설비 · 센서 · 품질 데이터 연계
- 배치 · 공정 조건 변화 시뮬레이션
- 설비 상태 · 품질 영향 분석
- 매뉴얼 · 품질 기준 기반 대응 연계
The content below is a representative application example, and the target equipment, process, data and verification scope are defined depending on the customer environment.