On an automotive fuel-tank assembly station, the hard part isn't "getting a clear shot of one tank" — it's making sure every unit follows the same order: locate the tank on the profiling block, bring it to the workbench, scan the barcode, connect the thick pipe, then operate and inspect with tools. A missed barcode, an unconnected pipe or a failed inspection must be caught right there.
This article is based on a real fuel-tank installation monitoring video and the VisionAgent UI, laid out as field scene → process timeline → workflow / events / state machine / parameter page → runtime dashboard. In the video, the AI already marks the tank, profiling blocks, barcode, thick pipe and tools with multi-color semantic-segmentation masks; the UI screenshots here are drawn to match the product's real controls, aligned with the video labels.
The real fuel-tank station in the video
The camera frames the fuel-tank assembly bench: an orange mask outlines "Tank on workbench", two purple masks mark "Tank on profiling block" (locating fixtures), and a cyan mask marks "Scanning barcode"; as the operator (in a hi-vis vest) connects the thick pipe, a blue mask marks "Thick pipe to box", and during tool work a red "Tool operating thick pipe" and a yellow "Tool inspecting thick pipe". These real labels are exactly what process monitoring must handle.
▲ Field: red mask "Tool operating thick pipe"
▲ Field: yellow mask "Tool inspecting thick pipe"
Six steps: what each one checks
The runtime checklist matches the state-graph event tensor — six ordered steps per the real labels in the video:
A missed fuel-tank step affects vehicle safety
The fuel tank involves sealing and fuel lines; a missed barcode, an unconnected pipe or a failed inspection often only surfaces at vehicle assembly or fuel-line testing. Process monitoring must lock "order + presence + inspection pass" on the station side.
Workflow: a five-operator tensor flow
Create a project "Automotive Fuel-Tank Installation" and drag operators from the left library, then connect them (matching the real canvas):
▲ Tensor-flow canvas: Camera → Preprocess → Segmentation → Events → State Machine (real UI style)
Orchestrating events: the GPU Event Orchestrator
Open the "GPU Event Orchestrator"; the header shows "Configuring event: Thick-pipe connection". Drag "Object Present" from the primitive library:
- The state-machine list on the right: E0 onto profiling block, E1 on workbench, E2 scan barcode, E3 pipe connection, E4 tool operation, E5 tool inspection… unfinished steps show "To configure";
- Scan steps can attach a "Barcode Recognition" primitive; miss-prone steps attach "Object Absent" to drive the interception branch.
▲ GPU Event Orchestrator: event "Thick-pipe connection" = primitive "Object Present" (real UI style)
State machine: the S0–S6 ring
- S0 await profiling —E0→ S1 await bench —E1→ S2 await scan —E2→ S3 await pipe —E3→ S4 await operate —E4→ S5 await inspect —E5→ S6 release;
- after E5 inspection pass, it returns to S0 — one loop per unit;
- the event tensor on the right lists EØ (void) and the installation events E0–E5; the state-action tensor is kept in sync.
▲ State-graph editor: fuel-tank installation ring S0–S6 (real UI style)
Camera & page parameter configuration
Open "Parameter Config" and switch to the Page Params tab. The "Defect Stats" on the left align with the installation steps, making it easy to count misses by NG type:
- Columns: Defect code (NG0…NG5), Defect name (profiling not located, tank missing, barcode not scanned, pipe not connected, operation incomplete, inspection failed), Count;
- You can "add / remove defect types"; edit name and description on the right, then "Save parameters" to apply;
- Result display: mask ON, defect name ON, font size 40, line width 2 — matching the on-site multi-class masks;
- On the camera side, configure channel, exposure and ROI at the "Industrial Camera" node so the whole station is framed.
| Defect code | Defect name | Count |
|---|---|---|
| NG0 | Profiling not located | 0 |
| NG1 | Tank missing | 0 |
| NG2 | Barcode not scanned | 2 |
| NG3 | Pipe not connected | 3 |
| NG4 | Operation incomplete | 1 |
| NG5 | Inspection failed | 2 |
▲ Parameter Config · Page Params: defect types & result-display settings (real UI style)
Go-live: video + six-step checklist
- Left: live view + multi-class masks, marking the tank, barcode and pipe work;
- Right: a six-step button checklist — onto profiling block, on workbench, scan barcode, pipe connection, tool operation, tool inspection; finished steps get a green tick;
- Verdict: inspection pass → release allowed; any step "object absent / inspection failed" → intercept and log the matching NG.
▲ Runtime monitor: live recognition on the left, six installation steps advancing on the right
▲ Live footage: the full automotive fuel-tank installation monitoring clip
Comparison table & takeaways
| Aspect | Traditional assembly line | VisionAgent fuel-tank installation |
|---|---|---|
| Recognize parts | Team-lead spot checks | Multi-class semantic segmentation identifies tank, profiling block, barcode and pipe in real time |
| Manage order | Habit and paper SOP | State graph S0–S6 enforces the order |
| Manage misses | Found at vehicle or fuel-line testing | Event "object present" step gating + NG stats |
| Manage inspection | Relies on tools and diligence | "Tool inspecting pipe" is an independent event; a fail is intercepted |
| Keep records | Paper ticks filled in later | Step trace and defect counts logged automatically |
- Business requirement: each unit finishes all six steps; barcode, pipe connection and inspection must not be missed;
- Decision rule: the current state only accepts its matching event; release is allowed only after inspection passes;
- Software config: five-operator tensor flow + GPU event orchestration + state graph + page params (NG dictionary).
Four questions run through it all: what the system looks at, what "done right" means, when to intercept, and how results reach the report.
In closing
The hard part of fuel-tank installation monitoring isn't recognizing one part — it's stringing recognize-in-place → report event → advance state → NG stats → inspection release into a chain the floor is willing to use and quality can stand up to review.