A common assumption in conversations about AI quality inspection is that the whole station — camera, lighting, industrial PC — needs to be replaced. In many cases that assumption is wrong, and it unnecessarily raises both cost and deployment time.
Why we start with qualification, not replacement
A "retrofit-first" approach checks whether the existing set — camera, optics, lighting, computer — meets the requirements of a specific process before we recommend replacing anything. Hardware replacement makes sense where it genuinely limits image quality or decision performance, not as the default starting point of every project.
What gets assessed
- Camera and optics — resolution, sharpness and field of view relative to the defect size that needs to be detected
- Lighting — stability and repeatability of imaging conditions between cycles
- Industrial computer — model decision time on the target hardware under near-production conditions
- Integration — available I/O interfaces and line communication protocols, e.g. Modbus TCP
When replacement is actually necessary
We recommend replacing a component when measurement shows a hard limit — for example, camera resolution physically cannot distinguish the defect from the background, or decision time on the current computer exceeds the line takt. In such cases, replacing a single component, not the whole station, is usually enough.
The practical effect
A retrofit-first approach shortens deployment time and avoids the cost of locking into one vendor's closed hardware ecosystem. It is also one of the things we already assess during the pilot — the qualification report states explicitly whether and where hardware replacement is needed.