Related Experiment Videos
Neural-network-based classification of Space Acceleration Measurement Systems (SAMS) data via supervised learning.
A Smith1, A Sinha, C M Grodsinsky
1Department of Mechanical Engineering, The Pennsylvania State University, University Park 16802, USA.
Microgravity Science and Technology
|January 1, 1996
Summary
Neural networks effectively classify events using Space Acceleration Measurement System (SAMS) data. This approach surpasses traditional methods, offering a powerful tool for analyzing space mission data.
Area of Science:
- Space Science
- Machine Learning
- Data Analysis
Background:
- Space missions generate vast amounts of data, requiring advanced analytical techniques.
- Classifying events from sensor data is crucial for mission monitoring and safety.
Purpose of the Study:
- To demonstrate the effectiveness of neural networks for event classification using SAMS data.
- To compare the performance of neural networks against other classification methods.
Main Methods:
- Developed MATLAB programs for SAMS data retrieval and power spectral density computation.
- Trained a multi-layer neural network (MNN) using the backpropagation algorithm.
- Utilized SAMS data from the STS-50 Space Shuttle mission.
Main Results:
- The trained MNN achieved high accuracy in classifying crew exercise events.
- Neural network performance exceeded that of the nearest neighbor classifier.
- Demonstrated the practical applicability of neural networks in space data analysis.
Conclusions:
- Neural networks provide a robust and accurate method for event classification in space missions.
- The developed methodology offers a significant improvement over existing classification techniques.