Frequently Asked Questions

There is no fixed threshold. The test is whether the equipment has been through a full cycle of the variation it experiences. If conditions shift with the seasons, both summer and winter need to be present; if conditions differ by product, the main products should appear. Typically a first model is built on three to six months of data and improved in operation. Where defect data is scarce, anomaly detection trained on normal data alone is a workable start.

Yes. In that case building the collection infrastructure is the first stage: extracting signals from equipment, standardizing formats and creating a storage structure. Attaching AI before that step produces results you cannot trust.

It depends on scope. A project addressing one specific problem in one process runs to some months; building data infrastructure from scratch takes longer. Two thirds of the deployments IMPIX has delivered ran two years or more. Validating on one line and then extending tends to produce better outcomes than targeting the whole plant at once.

Yes. We apply industrial communication standards (OPC-UA) and asset description standards (AAS) to connect equipment and existing systems. Where older machines do not support OPC-UA directly, a gateway sits in front of them. In some projects analysis results were surfaced inside the existing MES screens rather than in a new system.

Yes. IMPIX has delivered government programs as a supplier since beginning smart factory support projects in 2018, and can work with you from the planning stage. See the Government Support Programs page for detail.

Yes. In pharmaceutical and biotech sites we first establish whether the system falls under computer system validation, then design the collection path to meet data integrity (ALCOA+) and audit trail requirements. Adding AI does not change the regulatory requirements — if anything, recording where data came from and how it was processed becomes more important.

Models lose accuracy over time, because equipment condition and process conditions keep changing. IMPIX continues to handle model retraining and operational maintenance after delivery; after-sales support is a distinct business area for us.

Yes. A²LAB applies explainable AI (XAI), so the reasoning behind any action the system proposes can be followed. We take the view that a system offering only results rarely earns trust on the shop floor.

It depends on the state of your data. If data is not being collected, start with B²LAB or data standardization; if data exists but is not being used, start with a module that solves a specific problem, such as A² for analysis or V² for vision inspection. There is no need to adopt everything at once.

OWP was IMPIX's original flagship smart factory solution, made up of unit modules — MES, SCM, GMP management, Smart HACCP, WMS — on top of an equipment data collection layer (OWP-F). It still runs as the foundation of core operational systems. The company's current flagship platform, however, is A²LAB. Where OWP gathers data and supports the operational systems, A²LAB is the layer where AI reasons and controls on top of it.

Get in touch and we will review the conditions. Where data export is restricted, we approach it with on-premises or edge-based configurations. We can discuss in English and Japanese as well as Korean.