hmnshudhmn24/igbt-accelerated-aging-nasa

⚡ IGBT Accelerated Aging Dataset — NASA

A NASA prognostics dataset tracking accelerated thermal aging of Insulated Gate Bipolar Transistors (IGBTs) — for fault detection, remaining useful life prediction, and power electronics research.

---

🧭 Overview

This dataset contains experimental data from accelerated thermal aging of IGBTs (Insulated Gate Bipolar Transistors) under various stress conditions. Collected by NASA's Prognostics Center of Excellence, it is widely used in:

• Predictive maintenance research
• Remaining Useful Life (RUL) estimation
• Power electronics fault detection
• Prognostics & Health Management (PHM)

---

📁 Dataset Folders

Folder	Description
Thermal Overstress — DC Gate	Thermal cycling with DC waveforms. Device failed due to latch-up. Tail current changes observed as failure precursor. Single IGBT.
Thermal Overstress — Square Signal	Thermal cycling with square waveform at gate. Mimics real switching scenarios beyond safe operating area. Single IGBT.
Thermal Overstress — Square Signal + SMU	Square waveform aging of 4 devices + SMU characterization of threshold voltage, breakdown voltage, and leakage current.
SMU Data — New Devices	Electrical characterization of 40 pristine components (MOSFETs IRF520Npbf and IRG4BC30K). Baseline reference data.

---

📐 Parameters Measured

Parameter	Description
Collector Current	Transient current during turn-OFF
Gate Voltage	Applied gate waveform
Collector-Emitter Voltage	Switching voltage across device
Threshold Voltage	SMU characterization parameter
Breakdown Voltage	SMU characterization parameter
Leakage Current	SMU characterization parameter
Temperature	Detected via infrared sensor

---

🔬 Devices Covered

Device	Type
IRF520Npbf	N-Channel MOSFET
IRG4BC30K	IGBT

---

⚡ Quick Start

import pandas as pd
import scipy.io as sio
import os

# Load SMU characterization data (if .csv)
df = pd.read_csv("smu_data.csv")
print(df.head())

# Load MATLAB .mat files (if applicable)
data = sio.loadmat("igbt_aging_data.mat")
print(data.keys())

---

💡 Suggested Use Cases

1. Remaining Useful Life (RUL) Prediction — Predict how many cycles before device failure
2. Fault Detection & Classification — Detect latch-up and other failure modes early
3. Anomaly Detection — Identify abnormal tail current behavior as failure precursor
4. Degradation Modeling — Track electrical parameter drift over aging cycles
5. Transfer Learning — Use pristine SMU data as baseline for domain adaptation
6. Predictive Maintenance — Build real-world PHM pipelines for power electronics