Our solution is a deep learning-based seismic detection system designed to process and filter seismic data collected on planetary missions. It works in the following five steps:

Windowing the Seismic Data: The seismic wave is split into a fixed number of timestamps (or windows), allowing us to process the data in smaller, manageable segments.
Proximity/Heat Feature Creation: We then generate a proximity/heat feature for each timestamp. This feature represents how close a given timestamp is to the start of a seismic event, helping the model focus on important data points.
Deep Learning Model: A Convolutional Neural Network (CNN-1D) forms the backbone of the model. This model includes Pooling, Dropout, and BatchNormalization layers to enhance the model's ability to detect seismic events within noisy data.
Post-Processing Step: After the model makes predictions, the seismic wave is divided into smaller windows. Predictions are summed for each window, and we select the window with the highest summed predictions, indicating a potential seismic event.
Timestamp Selection: Finally, the system picks the timestamp with the highest predicted value from the selected window. This ensures only the most scientifically relevant data is transmitted back to Earth.

Benefits: This system reduces the power consumption of planetary missions by minimizing the amount of data transmitted, ensuring only important seismic events are sent. It also increases the efficiency of data transmission, making missions more sustainable for long-term exploration.

Tools & Technologies Used: We used TensorFlow for building and training the deep learning model. The coding languages used include Python for model development and data preprocessing. The system was developed using open-source libraries for deep learning and scientific computing.

Our goal is to make planetary seismology more efficient by using AI to identify and transmit only relevant data, addressing the key challenge of power management during space missions.

2024 NASA Space Apps Challenge

Cosmic Tremor Detectives