Implementing AR Visualization of Industrial Equipment
AR in industry — not marketing tool, but working tool. Technician aims tablet at machine and sees connection diagram, parts list under cover not yet opened, or maintenance instruction right over real equipment. Here accuracy matters more than beauty: if annotation "oil filter" hangs 15 centimeters from real filter — not AR, it's clutter.
Technical Binding to Equipment
Main question: how does AR system "recognize" specific unit and know where everything is. Three approaches:
Image Tracking (ARKit / ARCore). Printed QR code or unique plate on equipment — reference image. ARKit ARImageTrackingConfiguration detects and tracks. Reliable, cheap, but requires physical marker.
Object Scanning (ARKit). Scan 3D object via ARObjectScanningConfiguration → get .arobject file. In production session ARWorldTrackingConfiguration with detectionObjects — ARKit recognizes physical object without markers. Works on non-LiDAR devices, but requires sufficient object texture. Smooth metal pump case — poor candidate.
LiDAR Scene Reconstruction. On iPad Pro with LiDAR — build environment mesh, compare with reference equipment mesh via ICP (Iterative Closest Point). Accurate localization without markers, but computationally expensive.
In practice for production, most common: QR marker on equipment + ARKit Image Tracking. Simple, reliable, positioning accuracy ~1-2 mm.
Maintenance Instructions
Step-by-step instructions over real equipment — killer feature for AR MRO (Maintenance, Repair, Operations):
- Annotations to specific nodes: arrow + text "loosen M8 × 4 fasteners" tied to specific point on CAD model
- Animated disassembly: CAD model part "slides out" from body, showing extraction path
- Checklist with auto-advance: completed step → system moves to next, repositions annotations
Content prepared from CAD data (STEP, IGES, SolidWorks) → conversion via Datasmith (Unreal), PTC Creo Illustrate, or custom STEP → GLTF pipeline (via Open CASCADE Technology, open-source C++ SDK).
Stacks for Enterprise AR
Custom native development on ARKit/ARCore — for specific requirements. But in enterprise AR there are ready platforms worth considering:
PTC Vuforia Engine — de facto standard in industrial AR. Model Target (CAD recognition without markers), Area Target (space scanning via mobile app). SDK for iOS/Android/Unity. License from $840/year.
Scope AR WorkLink — no-code platform for creating AR instructions. Authoring via browser, player-app on iOS/Android. For typical MRO tasks — faster than custom development.
Upskill Skylight — WorkLink analog, focus on checklist-driven workflows.
Custom development justified when: deep ERP/CMMS integration needed, specific UI, offline work on protected devices.
Offline Work
Production facilities often limited or zero internet. All AR resources (3D models, instructions, reference images) must be stored locally. Scheme:
- Initial download when Wi-Fi connected (delta sync)
- All AR sessions work without network
- Log completed steps locally → sync when connection restored
SQLite + Core Data (iOS) / Room (Android) for local storage. Sync via REST API or GraphQL when connection restored.
ERP and CMMS Integration
Typical integrations: SAP PM (Plant Maintenance), IBM Maximo, Infor EAM. AR app gets from ERP: equipment serial number → loads needed instructions and 3D models. After maintenance completion — records work order back to ERP via API.
Timeline
MVP: image tracking + static annotations from JSON file, no ERP integration — 4-6 weeks. Complete solution with animated instructions, offline sync, SAP PM integration — 3-6 months. Cost calculated individually after technical requirements audit.