Skip to content

AAI Intern Project - Machine learning model for predicting Navaid power levels using signal and operational characteristics. The dataset includes attributes like frequency, elevation, and magnetic variation to classify power levels as LOW, MEDIUM, or HIGH.

Notifications You must be signed in to change notification settings

shubhijoshi/Navaids-project

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Predicting Navaid Power Levels Using Signal and Operational Characteristics

Source of Data

https://ourairports.com/data/

Overview

Navigational Aids (Navaids) are essential for aviation safety, providing critical location and directional data for aircraft. This project develops a machine learning model to predict Navaid power levels based on multiple operational features.

Objective

Predict Navaid power levels (LOW, MEDIUM, HIGH) using features like frequency, elevation, and magnetic variation. Accurate predictions enhance aviation infrastructure planning, resource allocation, and navigation system reliability.

Why is Power Prediction Important?

  • Preventing Equipment Failure – Aviation navigation relies on continuous signal availability. Low or excessive power levels can indicate potential equipment degradation or impending failure, allowing for proactive maintenance before critical failures occur.
  • Smart Predictive Maintenance – By analyzing historical power data, the model can identify patterns leading to power failures, helping technicians schedule maintenance before a system goes down. This reduces unscheduled downtimes and keeps operations smooth.
  • Optimizing Power Consumption – Many Navaid stations operate continuously, consuming significant power. Predicting optimal power levels can help reduce unnecessary energy usage, improving efficiency and lowering operational costs.

Dataset Overview

Feature Description
frequency_khz Navaid operating frequency
dme_frequency_khz Frequency used for DME
magnetic_variation_deg Magnetic deviation at the location
slaved_variation_deg Additional magnetic variation
elevation_ft Elevation above sea level
usageType Classification of Navaid usage
power (Target) Categorized as "LOW", "MEDIUM", or "HIGH"

Feature Relationships

  1. Power vs. Frequency:
    • Higher frequencies may require more precise tuning and power adjustments due to increased atmospheric attenuation.
    • Lower frequencies can travel longer distances but may still need consistent power for stable transmission.
  2. Power vs. Elevation:
    • Higher elevations often face less interference, reducing power needs.
    • Lower elevations may have more obstructions (terrain, buildings), requiring higher power levels for effective signal reach.
  3. Power vs. Magnetic Variation:
    • Variations in Earth's magnetic field may impact signal calibration, requiring power adjustments for accurate transmission.
    • Sudden changes in magnetic variation might necessitate real-time power tuning to maintain system accuracy.

Tech Stack

Libraries Used:

  1. Data Manipulation & Cleaning

    • pandas – Used for loading the dataset, handling missing values, filtering data, and transforming categorical variables into numerical representations.

    • numpy – Utilized for numerical computations, array manipulations, and mathematical operations to process data efficiently.

  2. Data Visualization

    • matplotlib – Used for creating plots and visualizing trends in the dataset, such as distributions and class imbalances.

    • seaborn – Employed for advanced statistical data visualization, including correlation heatmaps and feature relationships.

  3. Machine Learning

    • scikit-learn
      • RandomForestClassifier – Model training.
      • StandardScaler – Feature scaling.
      • GridSearchCV – Hyperparameter tuning.
      • accuracy_score – Model evaluation.

Machine Learning Model

  • Random Forest with Hyperparameter Tuning
    The primary model used in this project is the Random Forest Classifier, which is an ensemble learning method that operates by constructing multiple decision trees and averaging their predictions to improve accuracy and reduce overfitting. By leveraging hyperparameter tuning, the best-performing model was selected to maximize classification accuracy for predicting Navaid power levels.

Challenges

  • Handling missing values in key features like dme_frequency_khz and slaved_variation_deg.
  • Scaling numerical data for better model performance.
  • Selecting the optimal machine learning model to achieve the highest classification accuracy.

Success Metric

The model will be evaluated based on classification accuracy, aiming to optimize prediction performance through feature engineering, data preprocessing, and hyperparameter tuning. By accurately predicting Navaid power levels, this project enhances aviation reliability, reduces energy consumption, and supports proactive maintenance, ensuring safer and more efficient air navigation systems.

Internship Completion Certificate

https://drive.google.com/file/d/1Qe4rv19icfesJ8UbJtxH9jGjHpF0acfi/view?usp=drivesdk

About Airports Authority of India, Raipur

AAI Official Website
Industry: Aviation & Air Navigation Services
Employees: 5000+ (Across India)
Founded: 1995

The CNS (Communication, Navigation, and Surveillance) department ensures seamless air traffic communication, advanced navigation aids, and robust surveillance systems for safe flight operations. This project was undertaken under AAI Raipur’s Navaids Department, managing Navaid installation and maintenance.


Contributions

Contributed by Shubhi Joshi (NITRR) and Sneha Kumar (NITRR) under the guidance of Sandeep Rathore Sir (AAI Raipur)

About

AAI Intern Project - Machine learning model for predicting Navaid power levels using signal and operational characteristics. The dataset includes attributes like frequency, elevation, and magnetic variation to classify power levels as LOW, MEDIUM, or HIGH.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published