Skip to main content

Data-driven research addressing aviation safety intelligence

Deliverables

Information representation

Reporting on Task 3.2 activities which aims at answering the following research questions: • Can the amount of data be reduced, without compromising subsequent analyses? For instance, reducing records (data sampling), variables (dimensionality reduction), values (data discretization)? • Should new variables and/or better data types be created? The mathematical structure for modelling the data will be chosen.

Visualisation and user experience

Reporting on Task 2.3 activities aimed at: • Develop visual metaphors able to create synergies between the human cognition and the artificial computation. • Develop of Semantic Interaction solutions, i.e. approaches in which data are visualised in an interactive form, and the user's analytical reasoning is inferred and in turn used to steer the underlying models (Endert et al., 2012a, b; Endert et al., 2014). This can be initially achieved by coupling different controls to the parameters of the visual representation, as the parameters of the underlying analytic models are represented by the constructs of the visualisation, tacit knowledge of the user's reasoning can be inferred through inverting these analytic models. Semantic interaction also allows a more direct interaction with the information; and the creation of sets of optimal inputs, which once inserted in the system, would produce the output desired by the user in his/her explorations. • Visual data exploration aims at the integration of the human in the data exploration process. Leveraging on human perceptual abilities, visual data exploration is especially useful when little is known about the data and the exploration goals are vague (Andrienko and Andrienko, 1999; Keim, 2001). While the visual exploration can in principle be complemented by other automatic techniques, a framework for such integration is still missing and needs to be addressed. Visual exploration can easily deal with highly heterogeneous and noisy data, and has the advantage of requiring no understanding of complex mathematical and statistical algorithms. • Provide both intelligently summarised fast-access reporting as well as more detailed in-depth reporting. Another aspect to be addressed is the prioritisation and ontological structuring of data, information and knowledge and the advanced ability to include certain trade-off comparisons in the visualisation layout concept for more holistic decision-making purposes.

Large Scale Infrastructure v0 development

Reporting on Task 5.1, including: - Defining interfaces for the automated data collection system, thus linking the users with the data storage infrastructure. These interfaces can be internal (i.e. between open data) and external, the latter needs to address privacy concerns of projects partners. The privacy concerns could be addressed by providing part of the data or anonymised or aggregated data. - Implementing interfaces: in the case of external interfaces these need to be implemented within the LSI and by the partners providing external data. - Testing interface to the automated data collection system - Starting to integrate workflows provided by WP4 (see Tasks T4.2 and T4.3) into the GPI-Space environment (see Section 3.1.2), and validate high performance, real-time analytics implementations and automatic safety data monitoring.

Historical Analysis through descriptive analytics

Reporting on task 4.2 activities which refers to procedures and algorithms that help describe, show, or summarise data in a meaningful way. Its focus is on the macro-scale, that is, the system as a whole, and not on individual events and elements. This deliverable will include: • Definition of descriptive analyses following the requirements developed in Task 4.1. • Implementation of basic historical analysis algorithms with an analysis of their computational cost and parallelisation possibilities. • Definition and execution of test, validation, and usability plans. • Creation of the documentation required for an optimal use of the algorithms by the end users.

Final Results Assessment

Final results assessment. Further research in the field, progress beyond the state of art in safety and in the data mining tools.

SafeOps Laboratory Validation

Reporting on the results of task 6.1 regarding the following activities: • Validating the laboratory environment following the requirements developed in WP1 • Integrating all of the technological elements with reasonably realistic supporting elements • Assessing the performance of the system through as set of correctly represented set of indicators and the sensibility of the indicators shall allows performance monitoring • Ensuring the performance of the algorithms must meet the application goals in terms of speed, precision and resolution to support the application concept. • Delivering information in time according to the user's requirements for each case study.

Large Scale Infrastructure v1 development

Deliverable reporting on task 5.2 which includes: • Functional tests of infrastructure to check that all required features are available. • Final high performance integration of finalised workflows provided by WP4 into GPI-Space, supporting the historical analysis, the predictive analytics and the identification of known hazards. • Automatic safety data monitoring for the selected Case Studies and known hazards. • Robustness checks, such as against missing data, inaccurate data or communication, and tests with a series of safety incidents to stress test the infrastructure.

Supporting techniques for knowledge discovery

Reporting on Task 4.1 findings, including: • Selection of a set of training/testing data sets for each problem to be solved. • Definition and first implementation of algorithms for predictive analysis, taking into account the low frequency of the events of interest • Performance evaluation of the algorithms, in terms of accuracy and speed, for the training/testing datasets • Definition and implementation of algorithms for causality analysis between sets of multivariate data, both static and time evolving (i.e. time series) and definition and implementation of advanced ways of representing causality relationships, e.g. by means of complex networks of relations. • Development of ways for including causality analysis in standard data mining algorithms. • Identification of suitable SMC primitives and protocols for performing secure computations. • Assessment of the computational cost of each devised solution (both for predictive analytics, causality and secure computation), and development of parallelisation strategies. • Definition of security test plans.

Regulation and Certification

Reporting on the results of task 6.3 activities, concretely: • Identifying implementation issues to avoid implementation constraints • Clarifying the need for coordination and identification of the necessary steps foreseen for demonstration • Elaborating a deployment plan and selection of the corresponding monitoring mechanism. • Assessing the regulations and regulatory material (CS, AMC or GM) to be applied across the different SafeClouds technological elements to ensure uniform implementation of proportional technical requirements and to facilitate their deployment • Assessing the regulations and regulatory material (CS, AMC or GM) needed to appropriately protect safety data/ information and related sources of SafeClouds solutions • Recommending regarding future certification of SafeClouds technology

Predictive analytics and identification of unknown hazards

Reporting on Task 4.3 activities: • Definition of predictive and hypothesis testing analyses following the requirements developed in WP2. • Definition of the methodologies for the identification of unknown hazards through feature selection and combination algorithms. • Implementation of algorithms with an analysis of their computational cost and parallelisation possibilities. • Implementation of the algorithms and protocols for secure computation and causality analysis, as developed in Task 4.1. • Definition and execution of test, validation, and usability plans. • Creation of the documentation required for optimal use of the algorithms by the end users.

SafeRunway Laboratory Validation

Reporting on the results of task 6.2 activities, specifically: • Validating the laboratory environment following the requirements developed in WP1 • Integrating all of the technological elements with reasonably realistic supporting elements • Providing necessary fast time and real time simulators facilities for the conduction of SafeRunway Validation • Utilising big data studies results as an input for the validation platform • Assessing the performance of the system through as set of correctly represented set of indicators and the sensibility of the indicators shall allows performance monitoring • Confirming the performance of the algorithms is meeting the application goals in terms of speed, precision and resolution to support the application concept. • Delivering the information in time according to the user's requirements

Data preparation

Reporting on Task 3.1. The communication interfaces developed to link with the SafeClouds Large Scale Infrastructure (WP 5) will be reported, and guidance will be given to the Consortium partners in the process of inputting available data into the system. Additionally, a first data preparation process will be defined.

Project website

An initial version of the project website, including its private domain (www.SafeClouds.eu), will be developed and launched within the first months of the project. The website content (including its blog, see Task 7.3) will be developed and updated according to the communication strategy defined in the Dissemination Policy Plan and in line with the visual identity defined. It will be regularly updated as part of Task 7.3.

Searching for OpenAIRE data...

Publications

Smart Data Fusion: Probabilistic Record Linkage adapted to Merge Two Trajectories from Different Sources

Author(s): Dario Martinez (Innaxis), Samuel Cristobal (Innaxis) and Seddik Belkoura (Innaxis)
Published in: SESAR INNOVATION DAYS, 2018

Energy Management for Unstable Approach Detection

Author(s): Javensius Sembiring, Changwu Liu, Phillip Koppitz, Florian Holzapfel
Published in: 2018 IEEE International Conference on Aerospace Electronics and Remote Sensing Technology (ICARES), 2018, Page(s) 1-6
DOI: 10.1109/icares.2018.8547140

Parameter Estimation on Low Observability Data

Author(s): Javensius Sembiring, Joachim Siegel, Florian Holzapfel
Published in: 2018 IEEE International Conference on Aerospace Electronics and Remote Sensing Technology (ICARES), 2018, Page(s) 1-7
DOI: 10.1109/icares.2018.8547044

A Boosted Tree Framework for Runway Occupancy and Exit Prediction

Author(s): Dario Martinez (Innaxis), Seddik Belkoura (Innaxis) and Samuel Cristobal (Innaxis), Floris Herrema (Delft University of Technology) Philipp Wachter (Austro Control)
Published in: SESAR INNOVATION DAYS, 2018

Forecasting Unstable Approaches with boosting frameworks and LSTM networks

Author(s): Antonio Fernández Llamas, Darío Martínez Romero, Pablo Hernández Fish, Samuel Cristobal, Florian Schwaiger, José María Nuñez, José Manuel Ruiz
Published in: SESAR INNOVATION DAYS, 2019

Flight Data Monitoring (FDM) unknown hazards detection using AutoEncoders

Author(s): Antonio Fernández Llamas, Darío Martínez Romero, Pablo Hernández Fish, Samuel Cristobal, Florian Schwaiger, José María Nuñez, José Manuel Ruiz
Published in: SESAR INNOVATION DAYS, 2019

Validation of the Runway Utilisation concept at Vienna airport

Author(s): Floris Herrema & Mohamed Ellejmi
Published in: SESAR INNOVATION DAYS, 2019

A machine learning model to predict runway exit at Vienna airport

Author(s): Floris Herrema, Ricky Curran, Sander Hartjes, Mohamed Ellejmi, Steven Bancroft, Michael Schultz
Published in: Transportation Research Part E: Logistics and Transportation Review, Issue 131, 2019, Page(s) 329-342, ISSN 1366-5545
DOI: 10.1016/j.tre.2019.10.002