From a site under the Ministry of SMEs and Startups' autonomous factory build support programme to MES and PLM builds, we show what we built and with which technology, as it is. Planned figures are stated as plans. Six main cases have separate detailed records, from site conditions to results.
When the work runs under a standalone contract without a government support programme, it takes 2 weeks of consultation and 2 weeks to working software screens. On-site testing then takes usually 1 month. Up to this point is the core build period of about 2 months. The whole period including handover and stabilisation is, depending on scope, 6~10 months based on commercial contract cases; government support programmes take longer because agreement, interim review and settlement are added. It also takes longer as the systems and equipment to be connected increase — scope sets the duration.
For customer requests, AI carries out intake, development and deployment to the development server automatically, and deployment to production is checked by the developer in charge after customer approval. The values are a comparison against the period when people handled requests, from a full count of the same records.
| Item | When people handled it | Now |
|---|---|---|
| Time until half receive an answer | 9 days | 18.4 minutes |
| Answered within 24 hours | 19% | 69% |
| Same issue submitted again | 2.4% | 1.7% |
Each case is written in the same frame of six items: site conditions · problem · build scope · screens · results · next. Results carry an evidence tag — [Measured] means values measured in the completion report submitted to the government, [Verified] means demonstration completed, and [Case] means measured by the customer. Cases without result figures are written up to the build scope.
Existing equipment connected, DES accuracy 85% (year-2 target), demonstration of direct AI → PLC control. The record of an autonomous factory build at a volume press plant.
Hourly output 306→383 EA/HR (target 368 exceeded), process defect rate 2.3→1.7%. Measured values from the completion report submitted to the government.
Phase 1 built up the standard tables with MES, and phase 2 added predictive maintenance, quality prediction and process optimisation on the same data.
Hourly output 47→56 EA/h (+18.3%), finished-product defect rate 2.3→1.18% (-48.7%). These are measured values from the completion report submitted to the government, and LOT tracking is 100% automated.
Douzone iCube ERP replaced with VEXPLOR. Only HR and accounting stay; the company-wide systems move onto a single data model.
Younglimwon K-System ERP connected through an ontology without replacement. Five agents and an integrated dashboard for two plants.
In the diagnosis, we look at your data structure and propose a starting point.