Inconsistent Equipment Data Formats Are Blocking AI Adoption
The same temperature sensor gets tagged as `TEMP_01` on one vendor's equipment, `Zone1.PV` on another, and `t_actual` on a third. Units mix Celsius and Fahrenheit. Sampling intervals range from 100ms to one second. That is why every new quality prediction model on every new line starts over from data preprocessing.
The deeper problem is that none of this work carries forward. Mapping rules built for machine #2 do not transfer cleanly to machine #3, and moving to another plant means starting from scratch. A large share of AI project effort ends up spent on cleansing and mapping rather than on modeling.
Does OPC-UA solve this? OPC-UA is a communication standard that defines how data is exchanged. It standardizes connectivity, but what the transmitted values actually mean remains scattered across vendor documentation. That is precisely why a semantic layer standard is needed alongside the transport layer.
What Is the AAS? A Digital ID Card for Equipment
The Asset Administration Shell (AAS) is an international standard that gives assets — equipment, components, products — a standardized digital identity. It consists of three elements:
The most practical concept is the separation of Type AAS and Instance AAS. A Type AAS defines the equipment model — 'a press of this type carries these data items' — while an Instance AAS holds the actual values for a specific unit such as 'Line 3, Machine 2.' Define the model once and it applies to dozens of identical machines.
Korea's Manufacturing Data Standardization Program and Reference Models
Korea's Ministry of SMEs and Startups, together with the AI Innovation Task Force at TIPA, has been developing and distributing AAS-based manufacturing data reference models by industry. The program began with 50 models in 2024, added 115 more in 2025, and targets a cumulative 500. In 2026, five consortia comprising 17 organizations were selected, and a cooperation agreement was signed with the IDTA, which leads AAS international standardization.
For small and mid-sized manufacturers the implication is straightforward: standard schemas can be adopted rather than built. In sectors where reference models already exist — injection molding, pressing, surface treatment — much of the data definition phase can simply be skipped.
Four Steps to Adding AAS to Your MES, ERP, and AI Stack
Why Now
Requests from prime contractors for quality and traceability data have already become routine. For exporters to Europe, this intersects with regulations demanding product-level traceability and carbon data. AAS submodels are purpose-built containers for exactly these requirements.
Standardization gets more expensive the longer it is postponed. The more non-standard data and integration code accumulates, the larger the eventual rework scope becomes.
The POLYGLOTSOFT Approach
POLYGLOTSOFT proposes an integration design that preserves your existing MES and equipment connectivity while layering the standard on top. Because this means adding adapters rather than replacing systems, it can be deployed without halting production.
Our subscription development model lets you expand coverage machine by machine as reference models continue to roll out, spreading the initial investment across monthly payments. Start with a free prototype so we can validate the AAS mapping for one class of equipment in your plant together. Share your requirements and we will assess feasibility against your actual tag structure.