The initial release of the ICARUS platform has been used to execute the first scenario defined in the PACE demonstrator: „Pollution Data Analysis“, where a trained model is used to predict fuel consumption and pollution emissions for customer defined city pairs like Berlin/Athens.
The focus of the early release of the PACE demonstrator was on the evaluation and validation of basic platform functionalities like
- Data upload and data related operations like mapping, cleaning and encryption
- Analytics definition and execution
- Result visualization and evaluation
Scenario Execution Flow
As a preparation, a portion of „knowledge“ has been extracted from the Pacelab Mission Suite (PLMS), the only commercial available analysis tool which integrates a comprehensive aircraft model covering payload and performance definition and allows investigation of route network and economic investigations in a single environment, delivering reliable projections of key metrics such as payload capacity, maximum range and direct operating costs.
The knowledge is represented by the calculation of fuel consumed and pollution emitted for a set of almost 65.000 city pairs with varying conditions in distance, weather and payload factor. This data has been stored in a table for upload into the ICARUS platform and integration into the ICARUS data model.
Further data preparation activities can be executed on the platform during the data check-in like:
- Metadata addition
- License configuration
In a next step, an analytics workflow has been designed, where the uploaded data were used to train a linear regression model. This model owns all the knowledge exported from the PLMS and is enabled to create predictions for fuel consumption and pollution emissions for customer defined city pairs and customer defined conditions.
Therefore, the customer has to define and upload own city pair tables or the customer collects the data from the ICARUS platform, e. g. from a free or commercial data provider like OAG.
Then, the prediction part of the fuel consumption and pollution emissions data analysis has to be executed, where the linear regression prediction algorithm provides values for fuel consumption and pollution emission for each city pair, based on the knowledge transferred into the trained model.
Finally the report and visualization features of the platform have been used to find the best fitting results of the data analyses.
The first experiences were very promising and we are looking forward to execute the first and the second PACE scenario: “Massive Route Network Analysis and Evaluation” on the intermediate and final version of the ICARUS platform with improved analytics and reporting features.
Blogpost prepared by PACE-TXT