Description for Emission Prediction Program

Overview

The Emission Prediction Program is designed to predict vehicle emissions (CO2, NOX, PM, VOC, SO2) based on various input parameters using a neural network. The program features a graphical user interface (GUI) that allows users to input vehicle data, visualize predictions, and manage historical data.

Program Structure

• Main Functionality: The neural network preparation is divided into two main modes:

  • Training Mode: Trains the neural networks using a dataset and saves the trained weights.
  • Testing Mode: Loads pre-trained weights and allows users to input vehicle parameters for predictions.

• GUI Components: The GUI provides an interactive interface for users to input data and view results. It includes:

  • Input fields for vehicle parameters.
  • Buttons for actions such as predicting emissions, saving inputs, loading historical data, and clearing inputs.
  • A graph area to visualize emissions data.

Getting Started

1. Compile the Program: Ensure that all dependencies are installed and compile the program using CMake.
2. Run the Program: Execute the compiled binary to launch the GUI.

Using the GUI

1. Input Parameters:

   • Enter the following vehicle parameters in the provided input fields:
     • Vehicle Type (Car, Truck, Motorcycle, Bus)
     • Fuel Type (Petrol, Diesel, Electric, Hybrid)
     • Engine Size (liters)
     • Age of Vehicle (years)
     • Mileage (km)
     • Acceleration (m/s²)
     • Road Type (City, Rural, Highway)
     • Traffic Conditions (Heavy, Moderate, Free flow)
     • Temperature (°C)
     • Humidity (relative %)
     • Wind Speed (m/s)
     • Air Pressure (hPa)

   Then you need to input the max speed value. It is the limit up to which the graph will be drawn.

2. Buttons:

   • Predict: Click this button to calculate emissions based on the input parameters. The results will be displayed in the graph area.
   • Save: Saves the current input parameters in file.
   • Load: Loads previously saved input parameters from file.
   • Clear: Clears all input fields and resets the graph.
   • Help: Displays a help window with instructions on how to use the program.
   • Next/Previous Gas: Navigate through different gas emissions (CO₂, NOx, PM2.5, VOCs, SO₂) to view their respective graphs.
   • Journal: Expend the window and show last input configurations.

3. Graph Visualization: The program will display a graph representing the predicted emissions based on the input parameters. The graph will update when predictions are made.

Error Handling

• If invalid data is entered, the program will display an error message and prompt the user to correct the input.
• Ensure that the input values are within reasonable ranges to avoid errors during prediction.

Conclusion

This program provides a user-friendly interface for predicting vehicle emissions using advanced neural network techniques. By following the instructions above, users can effectively utilize the program to analyze and visualize emissions data.

Team:

• Team lead & DevOps Engineer - Alexander Semichastnov
• ML Engineer - Ruslan Sharafutdinov
• Frontend developer - Timur Safiullin

The project was made as part of an engineering workshop, MIPT RSE.
Link to Dataset