BIW Quality Gate: AR Checklist, 900 s per Body
A body-in-white sits on the shop floor and must be confirmed for component completeness, mounting orientation, position and overall structure before it enters the next process. The body is large and carries many components, visual inspection depends on experience, and skipped items and inconsistent standards occur, so problems reaching the next process trigger rework. The purpose of a quality gate is to keep the problem inside the previous process.
Deploy the items and steps, or the standard drifts with the person
The 3D CAD of the body is imported into the tablet, inspection items and steps are built in the system and deployed to the device. CAD becomes the digital inspection baseline and fixes what each item should reveal into the workflow, replacing paper checklists and memory. Steps follow process order and are confirmed item by item, which prevents skipping. The real gain here is standard consistency, not speed.
Position, orientation and scale all have to be right
The tablet camera identifies and tracks the physical workpiece, extracts body feature points and matches them to CAD features to establish correspondence, then overlays CAD at the correct position, orientation and scale. Tracking keeps updating the correspondence as the body is large and the viewing angle changes. If any of the three is off, the difference on screen is no longer the workpiece's problem.
Item-by-item marking is about adoption, not technology
The tablet supports item marking and handwritten annotation, close to familiar manual habits. For overall structure items, the overlay shows the theoretical state directly, and deviation from the physical body is obvious. Whether a tool is comfortable to use decides whether the shop will replace paper with the new process.
Reflection, exposure and tracking stability
Camera tracking depends on feature visibility, and light that is too dim or too bright affects it; reflections on body panels interfere with feature recognition. The treatment is not complicated: keep even lighting, adjust the viewing angle, control tablet exposure and add light where needed. What undermines tracking is usually these basic conditions, not the algorithm.
The report loop: classify, annotate, recheck, archive
When an anomaly appears, it is photographed by category and annotated on the overlay, and the system generates an inspection report. After rectification, a recheck confirms the fix before the body moves on. Reports are generated with photos and annotations, archived as quality-gate records for traceability, synchronized to the quality system, and linked to rectification status. The inspection cycle is 900 seconds, subject to site acceptance results.
Why this is not an AOI replacement
AR comparison is human-machine inspection: the inspector judges from the overlay, with high flexibility and fast deployment, which suits low-frequency gate checks. Automated optical inspection is machine judgment, suited to high-frequency inline work. Selection should assess CAD completeness and site lighting; when model and physical state disagree, synchronize the model revision first, and redeploy to the tablet after any model update. Once the checklist is digital, it can be replicated to other quality gates quickly.
What the quality gate team asks first
How does CAD align to the body?
Body feature points are tracked and matched to CAD features, and position, orientation and scale are solved in real time, with correspondence updated as the viewing angle changes.
How are inspection items built into the tablet?
Import the 3D CAD, build the item and step list for the inspection need, deploy the set to the tablet, and execute step by step on site.
What if the model is revised?
Synchronize the model revision before deployment. When the baseline and the physical part disagree, overlay comparison is meaningless.
How does it divide work with AOI?
AOI suits high-frequency inline automatic judgment; AR suits low-frequency human-machine confirmation. Cycle and inspection frequency decide the split.