🎓 Lesson 11
D5
Real-Time KPI Calculation for Fleet Orchestration
Real-time KPI calculation for fleet orchestration means instantly measuring how well your mining equipment (like haul trucks and shovels) is performing—so you can fix problems or adjust plans before delays pile up.
🎯 Learning Objectives
- ✓ Calculate real-time fleet utilization rate from GPS and engine-hour telemetry streams
- ✓ Design a dashboard logic flow that triggers an alert when shovel-truck matching efficiency drops below 85%
- ✓ Analyze cycle time variance across shifts using timestamped event data (load, haul, dump, return)
- ✓ Apply payload deviation thresholds to identify underloading/overloading trends in haul truck operations
- ✓ Explain how latency in sensor data ingestion impacts KPI validity and dispatch decision integrity
📖 Why This Matters
In modern remote operations centers, a 30-second delay in detecting a shovel’s idle time or a truck’s repeated underload can cascade into 12+ minutes of lost production per shift—costing $250K+/month at scale. Real-time KPIs aren’t just dashboards; they’re the nervous system of fleet orchestration, turning raw telemetry into actionable decisions that maintain throughput, reduce wear, and uphold safety compliance.
📘 Core Principles
Real-time KPI calculation rests on three interdependent layers: (1) Data acquisition fidelity—ensuring synchronized, calibrated telemetry (GPS, engine RPM, payload sensors, CAN bus) with <100ms end-to-end latency; (2) Stream processing semantics—applying windowed aggregations (tumbling windows of 60s), stateful event correlation (e.g., linking shovel ‘finish-load’ to truck ‘arrive-at-shovel’), and anomaly-aware validation (e.g., rejecting payloads >105% rated capacity); and (3) Operational semantics—mapping computed metrics to physical constraints (e.g., utilization = active-minutes / scheduled-minutes, where ‘active’ excludes planned maintenance but includes warm-up/idle during dispatch). Mastery requires understanding not just *what* is calculated—but *when*, *how often*, and *under what data quality conditions* the result remains trustworthy.
📐 Real-Time Fleet Utilization Rate
Utilization rate measures the percentage of scheduled time that a vehicle spends in productive motion or loading—excluding unscheduled downtime but including dispatch-authorized idling. It is computed continuously using sliding 5-minute windows to support rapid intervention.
Fleet Utilization Rate (UR)
UR = (Σ t_active / Σ t_valid) × 100Measures percentage of valid operational time spent in productive activity (motion, loading, dumping, authorized idle).
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| t_active | Active time | seconds | Cumulative time in productive operational states per window |
| t_valid | Valid window time | seconds | Window duration minus time with invalid or unsynchronized sensor data |
Typical Ranges:
High-efficiency iron ore fleet: 85 - 95%
Coal fleet with frequent road conditions: 72 - 84%
💡 Worked Example
Problem: Truck #T-421 is scheduled for 8 hours (480 min). Over a 5-minute sliding window, its telematics log shows: 210 sec moving, 90 sec loading, 60 sec dumping, 30 sec idling (dispatch-authorized), and 30 sec offline (GPS dropout).
1.
Step 1: Identify productive states — moving (210 s), loading (90 s), dumping (60 s), and authorized idle (30 s) = total 420 s.
2.
Step 2: Exclude non-productive states — GPS dropout (30 s) is invalid data; subtract from window duration: 300 s − 30 s = 270 s valid window time.
3.
Step 3: Compute UR = (productive seconds / valid window seconds) × 100 = (420 / 270) × 100 = 155.6% → capped at 100% (physically impossible >100%), indicating data misalignment; investigation reveals duplicate motion events — corrected UR = 360 / 270 = 133% → still invalid → confirms sensor sync error requiring CAN-GPS time alignment.
Answer:
The initial result (155.6%) violates physical bounds, revealing a data synchronization flaw. After correcting timestamp alignment, UR = 92%, which falls within the typical range of 80–95% for high-performing fleets.
🏗️ Real-World Application
At Rio Tinto’s Nammuldi Iron Ore ROC (Pilbara, WA), real-time KPI pipelines calculate shovel-truck matching efficiency every 30 seconds using synchronized PLC timestamps from shovels and GNSS-IMU fusion data from CAT 793 trucks. When matching efficiency dropped to 78% for two consecutive minutes (triggering Tier-2 alert), the system auto-paused dispatch for Truck Group B and rerouted 4 units to Shovel S-07—restoring efficiency to 91% within 92 seconds. Post-event root-cause analysis traced the dip to a 1.2-second GNSS clock drift in three trucks—corrected via OTA firmware update within 4 hours.
✏️ KPI Integrity Check Exercise
Given a 60-second telemetry window for Truck #T-77: Engine-on = 58 s, Payload ≥90% rated = 42 s, GPS speed >0 km/h = 36 s, ‘Dispatched’ flag = true for entire window, and 2 invalid accelerometer readings flagged. Calculate (a) real-time utilization rate, (b) payload compliance ratio, and (c) assess whether this window meets ISO 20487 ‘Valid Operational Window’ criteria (requires ≥95% sensor uptime & no conflicting state flags).
📋 Case Connection
📋 Limestone Mine ROC for Autonomous Haulage Fleet Coordination
Fleet coordination conflicts during simultaneous loading/unloading causing queue buildup and fuel waste