================================================================================
DATASET README
================================================================================
Title: Synthetic dataset for "A Comprehensive Performance Evaluation Framework 
       for Machine Learning-Based Oil, Gas and Water Leak Detection Systems"
Authors: Sina Alizadeh Tabrizi, Bahador Fatehi-Nobarian, Farshid Nematzadeh
Corresponding author: bfatehinobarian@iau.ac.ir
================================================================================

This dataset supports the case study presented in Section 5 of the manuscript 
(Hybrid IoT Gas Leak Detection Node for Residential Kitchens). All data are 
synthetic but statistically consistent with the experimental results reported 
in the paper (detection accuracy: 96%, false alarm rate: 3%, mean latency: 
15.6 s, energy per detection: 50 mJ, packet loss: 2%, battery lifetime: 10 years).

================================================================================
FILES INCLUDED
================================================================================
The dataset is provided as a single Excel file with the following sheets:

1. Leak_Detection_Events       (100 rows)  – 100 leak injection experiments
2. NonLeak_Interference_Events (100 rows)  – 100 non-leak / interference scenarios
3. Timing_Breakdown            (100 rows)   – Latency breakdown for true detections
4. Energy_Consumption          (100 rows)   – Energy per detection (mJ)
5. LoRaWAN_Communication       (100 rows)  – Packet log with 2% loss rate
6. Reliability_LifeTest        (10 rows)   – Accelerated life test results
7. Battery_Lifetime            (1 table)   – Calculation of 10-year battery life

================================================================================
SHEET DESCRIPTIONS
================================================================================

Sheet 1: Leak_Detection_Events
Columns:
- Event ID         : Unique identifier (L001 to L100)
- Ground Truth     : "Leak" for all events
- Detected         : "Yes" for 96 events, "No" for 4 events (false negatives)
- Detection Time   : Time of leak onset (always 0)
- VOC Sensor (ppm) : Simulated VOC reading (range 20–65 ppm for leaks)
- Pellistor (ppm)  : Simulated pellistor reading (range 140–500 ppm for leaks)
- RF Probability   : Random Forest output probability (0.81–0.99 for TP, 0.62 for FN)
- Alarm Triggered  : "Yes" if probability >0.7, else "No"
- Latency (s)      : Total detection latency (15.2–16.3 s)
- Energy (mJ)      : Total energy per detection (47–54 mJ)

Sheet 2: NonLeak_Interference_Events
Columns:
- Event ID           : N001 to N100
- Ground Truth       : "No Leak"
- Detected as Leak?  : "Yes" for 3 events (false alarms), "No" for 97 events
- VOC Sensor (ppm)   : 5–31 ppm (typical kitchen interferences)
- Pellistor (ppm)    : 12–102 ppm (well below leak threshold)
- RF Probability     : 0.05–0.22 (except false alarms: 0.73, 0.75)
- Interference Type  : Cooking vapor, cleaning product, steam, etc.
- Alarm (False?)     : "Yes" for false alarms, "No" otherwise

Sheet 3: Timing_Breakdown
Columns (for each true detection, L001to L100):
- Event ID           : Matching leak event ID
- Leak Onset (s)     : Always 0
- Sensor Response (s): 9.5–10.7 s (physical sensor transient)
- Edge Inference (ms): 1.1–1.3 ms (Random Forest on ESP32)
- LoRa TX (ms)       : 4.2–4.5 ms (LoRaWAN transmission)
- Total Latency (s)  : Sum of above (mean 15.6 s)

Sheet 4: Energy_Consumption
Columns (for each true detection):
- Event ID            : Matching leak event ID
- Sensing Energy (mJ) : 37.5–43.0 mJ
- Inference Energy (mJ): 0.8–0.9 mJ
- TX Energy (mJ)      : 8.2–9.6 mJ
- Total Energy (mJ)   : Sum (mean 50 mJ)

Sheet 5: LoRaWAN_Communication
Columns (100 packets from the IoT node to gateway):
- Packet ID           : P001 to P100
- Node ID             : Node01
- Gateway ID          : GW01
- SF                  : Spreading factor (7)
- BW (kHz)            : Bandwidth (125)
- TX Power (dBm)      : 14
- Sent Timestamp (ms) : 0 to 99,000 in 1,000 ms steps
- Received Timestamp (ms): Timestamp + latency (or blank if lost)
- Status              : "Delivered" (98 packets) or "Lost" (2 packets)
- Latency (ms)        : 85–120 ms for delivered packets

Sheet 6: Reliability_LifeTest
Columns:
- Tested Unit         : Unit 01 to 10
- Test Hours          : 5,000 (ambient) or 2,000 (accelerated)
- Temperature (°C)    : 25, 85, -20, 60, 40
- Humidity (%RH)      : 50, 85, 95, 80
- Failures Observed   : 0 for all units
- Calculated Failure Rate (λ): 0 (no failures during test)
- MTTF (years)        : >20 (extrapolated from accelerated testing)

Sheet 7: Battery_Lifetime
Calculation table showing:
- Average current draw per component (μA)
- Duty cycle (%)
- Average contribution (μA)
- Total average current: 35.3 μA
- Battery capacity: 3000 mAh
- Estimated lifetime: 85,000 hours ≈ 10 years

================================================================================
REPRODUCIBILITY
================================================================================
For any questions, please contact the corresponding author.
================================================================================