prompt-grb-analysis

Bangs, Blips and Background: Decoding GRB Signals

This project implements a pipeline to detect and classify Gamma-Ray Bursts (GRBs) from AstroSat-CZTI data using statistical methods and signal processing techniques. It distinguishes astrophysical transients from background noise and charged particle hits, establishing a framework for reliable GRB detection and analysis.

---

Project Structure

• 1_GRB_detection.ipynb
  Constructs light curves from AstroSat CZTI data and applies multiple detrending filters (Savitzky-Golay, Median, Wavelet) to highlight potential transient features.

• 2_pipeline.ipynb
  Performs peak detection using SNR analysis and n-sigma thresholding, followed by spectrogram generation and classification of events into:

  • Bang (GRB)
  • Blip (charged particle)
  • Background (noise)

• report.pdf
  A detailed report explaining the physics motivation, methodology, detector details, results, and conclusions of the study.

---

Methods Overview

Data Processing

• Raw CZTI quadrant and veto data parsed from .evt FITS files using astropy.
• Binned into 1s and 10s light curves across multiple energy bands (20–100 keV for CZT and 100-500keV for Veto).

Detrending

Applied and compared:

• Savitzky-Golay filtering (unsuitable due to peak smoothing)
• Median filtering (best performance for impulsive events)
• Wavelet denoising (balanced suppression but slightly smoothed peaks)
• Matched filtering (template mismatch issues for GRBs)

Peak Detection

• Used SNR calculation and 5σ thresholding to identify statistically significant events.
• Spectrograms and energy-profile plots validate astrophysical signatures.

Classification Logic

1. Bang (GRB): Multi-quadrant, multi-band, >5σ, power-law energy profile.
2. Blip (Particle): Localized spike in one band/quadrant, 3σ–5σ range.
3. Background: Instrumental artifacts, noise, or post-SAA glitches.

---

Instrument: AstroSat-CZTI

The Cadmium Zinc Telluride Imager (CZTI) onboard AstroSat (India’s first space observatory) operates in the 20–200 keV range with a wide field-of-view. Its veto detectors (100–500 keV) help eliminate cosmic ray contamination. The data are quadrant-resolved and sensitive to GRBs both on- and off-axis.

---

Key Results

• Successfully recovered faint GRB candidates not captured by onboard triggers.
• Validated peaks through spectrograms and energy evolution patterns.
• Built a foundation for a future machine learning classifier on transient detection.

---

Future Work

• Extend the classifier with a machine learning pipeline (GRB vs blip vs noise).
• Integrate cross-matching with Fermi GBM/Swift BAT catalogs.
• Automate SAA masking and real-time event flagging.
• Tweak the pipeline more to make it highly robust and make the accuracy of flagging events to maximum

Requirements

• Python 3.8+
• numpy
• pandas
• matplotlib
• scipy
• astropy
• pywt (for wavelets)
• scikit-learn (for DBSCAN clustering)
• seaborn
• JupyterLab or Notebook

Install all dependencies:

pip install -r requirements.txt