Model-Based Analysis of Human Errors During Aircraft Cockpit System Design
State of the Art - Background
The safety of aircraft has been significantly enhanced during the last decades by technical improvements and new training concepts. However, the accident rate has remained almost the same, varying between three and four accidents per million departures, because the safety improvements could not outweigh the dramatic increase of the overall air transport. For the future, an even stronger increase in traffic density is anticipated, which leads experts to expect one serious accident a week if the rate is not drastically reduced.
Worldwide commercial jet fleet statistical information reports that 55% of accidents involve flight crew errors. The examination of human errors has been developed in the aircraft industry and is now considered to be an important analysis to accomplish during the design and certification of the cockpit. However, the current approach of analysing systems is prone to errors as well as being costly and time-consuming (based on engineering judgement, operational feedback from similar aircraft and simulator-based experiments). Therefore, in order to enhance the safety of the aircraft itself and its systems, innovative solutions for improved human-centred design are needed that allow for the more accurate detection of potential pilot errors at an earlier stage (in the design) and with reduced effort.
The objective of the HUMAN project is to develop a methodology with techniques and prototypical tools supporting the prediction of human errors in ways that are usable and practical for human-centred design of systems operating in complex cockpit environments.
The prediction of human errors will be achieved by developing and validating a cognitive model of crew behaviour. Cognitive models are a means to make knowledge about characteristic human capabilities and limitations readily available to designers in an executable form. They have the potential to automate parts of the analysis of human errors because they offer the opportunity to simulate the interaction with cockpit systems under various conditions, and to predict cognitive processes like the assessment of situations and the resulting choice of actions, including erroneous actions. In this way they can be used as a partial 'substitute' for human pilots in early developmental stages when design changes are still feasible and affordable. Model and simulation-based approaches are already well established for many aspects of the study, design and manufacture of a modern airliner, for the very same objective of detecting potential problems earlier and reducing the amount of testing required at a later stage. HUMAN will extend the modelling approach to the interaction of flight crews with cockpit systems.
Description of Work
The main research and development work in HUMAN will produce key innovations on three complementary research dimensions:
- Cognitive modelling: to develop an integrated cognitive crew model able to predict human error categories with regard to deviations from normative activities (standard operating procedure and rules of good airmanship).
- Virtual simulation platform: to develop a high-fidelity virtual simulation platform to execute the cognitive crew model in realistic flight scenarios in order to analyse the dependencies between the pilots, a target system in the cockpit, the aircraft and its environment.
- Physical simulation platform: to thoroughly investigate pilot behaviour on a physical simulation platform to produce behavioural and cognitive data as a basis for building a detailed knowledge base about cognitive processes leading to deviations from normative activities, and for validation and improving the predictions of the cognitive model generated on the virtual simulation platform.
The general idea of the virtual and physical platform is to use the same core system for both in order to ensure the functional equivalence between the two platforms. This equivalence is a fundamental precondition for validating the cognitive model by producing and comparing predicted crew activities (on the virtual platform) and actual crew activities (on the physical platform).
The output of the HUMAN project will be:
- An innovative means enabling the considerable improvement of the human-centred design of cockpit systems, including a cognitive crew model able to predict design-relevant pilot errors;
- A high-fidelity virtual simulation platform enabling execution of the cognitive crew model;
- A prototypical tool based on the virtual simulation platform supporting usability of the platform and cognitive model;
- Formal techniques and prototypical tools for analysis of simulator data;
- A detailed knowledge base about cognitive processes leading to pilot errors and derived guidelines for cockpit system design;
- A methodology that integrates all the techniques and tools for their application during aircraft cockpit system design.
HUMAN will have an impact on aircraft safety. The project will contribute to the European Commission's objective of reducing the accident rate by enhancing the accuracy of pilot error prediction. Furthermore, it will contribute to the objective of achieving a substantial improvement in the elimination of and recovery from human error by reducing the design effort of active and passive safety measures, and by reducing the effort of flight simulator tests for active and passive safety measures.
- Related Info
- Acronym: HUMAN
- Name of proposal: Model-Based Analysis of Human Errors During Aircraft Cockpit System Design
- Grant Agreement: 211988
- Instrument: CP - FP
- Total cost: 3 909 789 €
- EU contribution: 2 777 379 €
- Call: FP7-AAT-2007-RTD-1
- Starting date: 01/03/2008
- Ending date: 28/02/2011
- Duration: 36 months
- Technical domain:: Avionics, Human Factors and Airports
Mr. Andreas Luedtke
Offi s e.V.
DE 26121 Oldenburg
- E-mail: Luedtke@offi s.de
- Tel: +49 (0)441 9722101
- Fax: +49 (0)441 9722102
- EC Officer: Ms. Stéphanie Stoltz-Douchet
- Airbus France FR
- Alenia Aeronautica S.p.A. IT
- Université catholique de Louvain - Belgian Laboratory of Human-Computer Interaction BE
- Deutsches Zentrum für Luft- und Raumfahrt e.V. DE
- Netherlands Organization for Applied Scientific Research - TNO Human Factors NL