Battery Circularity: Recycling, Second-Life Integration, and Safety Standards

3-day, 90% hands-on program where you build a RoboDK robot cell, stand up an IoT/digital twin (OPC UA → MQTT → TSDB) with live KPIs, and deploy edge ML for quality, anomaly, and basic RUL. Leave with a working cell, twin stack, Dockerized models, and an ops playbook to lift OEE, cycle time, and first-pass yield.

Enable participants to take an energy-sector robot cell from offline simulation to online, data-driven operation by (1) designing and programming RoboDK cells, (2) building a standards-based IoT/digital-twin pipeline (OPC UA → MQTT → time-series DB) with a live dashboard, and (3) deploying edge ML models for quality, anomaly detection, and basic RUL—delivering measurable improvements in OEE, cycle time, and first-pass yield.

📅 Day 1 – RoboDK Cells (Offline → Online) — 90% Hands-On

• Concept sprint (10%): Robot cell basics for energy ops (pick-place, torqueing, inspection), safety envelopes, tool frames.
• Hands-on 1: Build a RoboDK cell (robot + turntable + fixtures) for an energy asset task (e.g., valve assembly or inverter heat-sink placement).
• Hands-on 2: Calibrate TCP & reference frames; import CAD (STEP) and create targets/paths with collision checks.
• Hands-on 3: Add sensors/IO stubs; map gripper open/close and torque gun signals.
• Hands-on 4: Generate offline programs (UR/KUKA/FANUC post); simulate cycle time & reach; export cycle report.
• Hands-on 5: Create a changeover: parametrize path for two SKUs; save recipe files and a quick SOP.

📅 Day 2 – IoT Twins & Edge Connectivity — 90% Hands-On

• Concept sprint (10%): What is an IoT/Digital Twin for a cell/line; OPC UA vs MQTT; time-series stores.
• Hands-on 1: Spin up an OPC UA server for the robot cell (simulated tags: pose, IO, cycle_ok); browse with a client.
• Hands-on 2: Bridge OPC UA → MQTT; publish twin topics (cell/status, robot/pose, quality/ok) with JSON schema.
• Hands-on 3: Ingest to a time-series DB (InfluxDB/Timescale) and persist recipe/cycle/energy metrics.
• Hands-on 4: Build a lightweight digital twin dashboard (Node-RED/Streamlit): live KPIs (OEE, cycle time, energy per unit), alarm banner, and recipe selector.
• Hands-on 5: Implement edge rules: debounce a sensor, create an interlock (door_open ⇒ robot_hold), and test a “fault → maintenance ticket” webhook.

📅 Day 3 – ML Analytics for Quality, Anomaly & Maintenance — 90% Hands-On

• Concept sprint (10%): Framing: anomaly vs quality classification vs RUL; labeling and drift.
• Hands-on 1: Build a feature table from Day-2 streams (cycle-level stats, torque traces, robot current, temperature).
• Hands-on 2: Train a quality classifier (gradient boosting) for pass/fail based on torque/current signatures; SHAP for explainability.
• Hands-on 3: Train an anomaly detector (isolation forest/autoencoder) on robot/IO telemetry; set calibrated alert thresholds.
• Hands-on 4: Simple RUL proxy: survival/Weibull fit for a consumable (gripper pad) using cycle counts and peak force.
• Hands-on 5: Deploy models at the edge (Docker container): subscribe to MQTT, score in real-time, publish alerts/quality & alerts/anomaly; update twin dashboard with badges & trendlines.
• Hands-on 6: Create a runbook: alarm triage, false-positive guardrails, and a rollback switch; export a one-page “ops playbook.

• Open to UG/PG students, researchers, and professionals in Biology, Biotechnology, Bioinformatics, or related fields.

• Basic computer literacy required; prior coding experience is not mandatory (we start from fundamentals).

• Recommended for participants comfortable with self-paced practice and evening live sessions