This repository contains all notebooks that perform various data analyses and computational tasks for the analysis of the Martian Core and Seismic data. Each notebook is focused on a specific topic or methodology. Below is a summary of the actions performed and the analysis conducted in each notebook. Detailed Report Available at : Report.
This notebook focuses on signal processing techniques for analyzing the seismic data which is time-series wave-like data. The key actions include:
- Signal Loading: Time-series data was imported and preprocessed from the .mseed files using the Obspy Library.
- Filtering: Various filters (e.g., bandpass) were applied to denoise the signal and remove instrumental errors (like spikes).
- Fourier Analysis: The frequency components of the signal were extracted using Fast Fourier transforms.
- Time-Frequency Analysis: Techniques like wavelet transform were used to analyze signal behavior over time.
- Visualization: Spectrograms and time-domain plots were generated to interpret the results.
- Synthetic Data Simulation: Synthetic waveforms were generated from the available seismic data for shadow and non-shadow zones.
- Supervised ML to classify the Seismic Data: Logistic Regression was applied to various features obtained from the seismic data to predict whether it is a shadow zone or non-shadow zone.
This notebook focuses on analyzing and determining the core radius of Mars in a dataset. The following steps were performed:
- Martian Interior Simulation: Parameters were researched and assumed from various sources and extensive calculations were performed to model the P wave and S wave velocity and the density as a function of depth.
- Datasets: The model was compared with the other pre-existing datasets which simulate the P and S wave velocities and Denisty in the Martian Interior.
- Feature Engineering: Features like P and S waves velocities and Density were separated.
- Predictive Analysis: Algorithms like Random Forest were applied to calculate the optimal core radius.
- Visualization: Scatter plots to visualize the predicted and actual radius.
This notebook applies the DBSCAN algorithm for anomaly detection in seismic waveform data. Steps include:
- Introduction to DBSCAN: A brief overview of how DBSCAN works for clustering and anomaly detection.
- Dataset Preprocessing: Amplitude, Frequency, and Phase of the waveforms were separated for individual analysis of anomalies that might be caused by subsurface geological features. The dataset was normalized to improve algorithm efficiency.
- Parameter Tuning: Parameters such as
epsandmin_sampleswere adjusted to optimize clustering performance. - Anomaly Detection: Points flagged as anomalies in the amplitude, frequency, and phase plots were visualized and analyzed.
This notebook implements a Physics-Informed Neural Network (PINN) using a Gaussian Process Neural Network (GPNN) for solving partial differential equations (PDEs) of the P and S waves and mapping them as a function of depth. Highlights include:
- Physics-Informed Modeling: Derivation of PDEs and incorporation of physical laws into the neural network structure.
- Network Design: Implementing a Gaussian process neural network with PDE loss function.
- Training and Optimization: Loss functions were defined to minimize errors in satisfying the physical laws (PDE loss) and boundary conditions (MSE loss).
- Validation: Results were validated against analytical or numerical solutions.
- Visualization: The solution was visualized using 2D and 3D plots.
The 3D Simulation and the Scientific Functions were performed using Axisem and SPECFEM3D software. The output files are available at the drive link: 3D these files can be opened on PARAVIEW.
- Clone the repository:
git clone https://github.com/who-else-but-arjun/Martian_Core_Analysis.git - Install dependencies: Ensure Python and Jupyter Notebook are installed along with necessary libraries
pip install -r requirements.txt. - Follow the steps in the notebook to reproduce the results or modify them for your use case.