PRODUCTION SPECIFICATION v1.0.0 • NON-INTRUSIVE OPTICAL SENTRY

Sightline Edge Vision AI: Non-Intrusive Optical Watchdog & Telemetry

⚙️ System Prerequisites & Standalone Architecture

Field-tested for clean bare-metal factory IPCs and HMIs. Zero Docker, Zero Python, and Zero external CDN dependencies required.

APPLIANCE RUNTIME INTEGRITY: Core modules and in-memory UI assets are compiled via Cython into native C-machine shared objects (.so). Pre-baked with official embedded runtime engines. Internal Audit Verified: Score 97/100 (Grade A+ Exemplary).

Windows 10 / 11 / IoT IPC ZERO-DOCKER .BAT
  • Self-Contained Engine: Bundled with embedded 64-bit Python runtime. No system PATH edits.
  • 1-Click Launch: Runs via START_SIGHTLINE.bat; auto-binds to http://localhost:8000.
  • RAM Footprint: Strict < 304 MB Resident RAM limit.
Linux Bare-Metal IPC STANDALONE .SH
  • Pre-Baked Libraries: Portable glibc engine with headless OpenCV and ONNX Runtime.
  • 1-Click Launch: Runs via ./start_sightline.sh; no root daemon or Docker needed.
  • Crash Durability: SQLite WAL mode survives factory electrical blackouts safely.

1. Optical Sentry Physics & Optics Positioning

Unlike broad security surveillance, Sightline operates as a calibrated optical sentry. Proper mounting and illumination are critical for accurate sub-millisecond edge telemetry:

  • Mounting Geometry: Mount the camera housing on a rigid overhead arm or pillar at 30° to 45° elevation, angled toward the tool retract position and output feedbed.
  • Illumination: High-frequency LED ring lighting or polarized cross-illumination is recommended to eliminate spindle metallic glare and shadows.
  • Coolant Protection: In wet CNC machining, use an IP67 sealed camera enclosure with an active air-knife (curtain) nozzle to keep oil mist off the front lens element.

2. Camera & Sensor Compatibility Matrix

Sightline is a software appliance compatible with standard industrial optical hardware (Cameras supplied separately):

InterfaceProtocol / StreamSupported HardwareConfiguration
RTSP IP Stream RTSP / H.264 over TCP Hikrobot, Dahua, Basler IP, Uniview CAMERA_SOURCE=rtsp://user:pass@ip:554/stream
Industrial USB3 USB3 UVC / V4L2 High-Bandwidth Basler dart/pulse, FLIR, Imaging Source CAMERA_SOURCE=0 (or device index)
Embedded CSI MIPI-CSI Sony IMX 2K Raspberry Pi HQ Camera, Arducam IMX477 Direct V4L2 pipeline binding

3. Multi-OS Standalone Execution Runbook (Zero Docker)

Follow the exact step-by-step instructions below. No Docker installation or developer tools are required:

[PATH A] Windows 10 / 11 / Server IPC Runbook
  1. Extract cnc-sightline-v1.0-windows.zip into C:\cnc-sightline\.
  2. Double-click START_SIGHTLINE.bat. The embedded engine boots in seconds.
  3. Your browser automatically opens: http://localhost:8000.
  4. Copy the 16-character Machine Fingerprint shown in the red banner.
  5. Email your ID to support@systems.creativesnext.com to receive your signed 72-Hour Key.
  6. Drop license.key into the data/ folder. The system instantly unlocks.
# To terminate safely at any time:
Double-click STOP_SIGHTLINE.bat
[PATH B] Linux Bare-Metal & IPC Standalone Runbook
# Step 1: Extract standalone archive
tar -xzf cnc-sightline-v1.0-linux.tar.gz
cd cnc-sightline-v1.0-linux

# Step 2: Launch standalone watchdog engine
./start_sightline.sh

# Step 3: Access telemetry dashboard at:
http://localhost:8000

# Step 4: Drop your received license.key into data/ to unlock full 72-hour quota:
cp /path/to/license.key ./data/license.key

4. Dual-Zone ROI & Threshold Calibration

Sightline divides each 2K frame into two dedicated Regions of Interest (ROI) directly adjustable via the Dashboard UI:

  • Zone A (Spindle Tooling - 800x800): Evaluates cutting insert edge sharpness and chipping. Calibrated threshold: 0.40 edge density.
  • Zone B (Workpiece Deliverable - 1200x800): Evaluates finished part surface fractures and burrs. Calibrated threshold: 0.45 edge density.
  • Custom ONNX Model Provision: Drop any pre-trained model.onnx (YOLO / PatchCore) into data/models/ to autonomously switch from mathematical telemetry to deep neural defect classification.

5. 72-Hour Cumulative Shift Quota & Node-Locking

Sightline operates on an active operating quota rather than calendar days:

  • Cumulative Machine Runtime: System measures active seconds recorded in SQLite. An 8-hour shift consumes exactly 8 hours; if the machine stops, the quota pauses.
  • Anti-Clock Rollback Guard: Monotonic SQLite database verification terminates trial extension attempts if system RTC time is manually rolled back.
  • Seamless Perpetual Upgrade: Replacing license.key with a commercial key permanently unlocks the appliance with zero re-installation.

6. Production Packages & Download Portal

Acquire the certified zero-Docker standalone bundle for Windows (.zip) or Linux (.tar.gz) from our official portal:

Go to Sightline Edge AI Portal →
PRODUCTION SPECIFICATION v1.0.0 • KUBERNETES-NATIVE

Kubernetes-Native (K3s) CNC Machine Tool Wear Monitoring & Predictive Maintenance

1. Architecture & ISO 3685 Flank Wear Physics

During high-speed CNC cutting, carbide insert degradation follows a non-linear wear curve governed by ISO 3685. Operating past the critical flank wear boundary ($VB \ge 0.30\text{ mm}$) leads to exponential friction surges and spindle bearing damage.

2. 5-Protocol Unified Gateway & ZeroMQ IPC Bus

Ingests spindle electrical and vibration telemetry across Siemens OPC UA, Fanuc Modbus TCP, Haas MTConnect, China NC-Link, and retrofit MQTT.

3. Multi-OS Shopfloor Execution Runbook

Pre-compiled ELF/PE standalone executables deployable via Helm or batch scripts.

4. 0.08ms Inference & Air-Gapped Licensing

Dual-mode ONNX runtime executing deterministic tool wear classification in 0.08 ms.

5. Production Packages & Download Portal

Go to CNC Tool Wear (K3s) Portal →
PRODUCTION SPECIFICATION v3.0.0

CNC Spindle Thermal Drift Predictive Maintenance

1. Architecture & Physics Engine

Thermal expansion modeling across Z-axis high-speed spindle operations.

5. Packages & Download Distribution

Go to Spindle Prototype Portal →