Learning from past accidents to improve system design

Moura, Raphael and Beer, Michael and Patelli, Edoardo and Lewis, John and Knoll, Franz (2016) Learning from past accidents to improve system design. Safety Science, 84. pp. 37-45. ISSN 0925-7535 (https://doi.org/10.1016/j.ssci.2015.11.022)

[thumbnail of Moura-etal-SS2016-Learning-past-accidents-improve-system-design]
Text. Filename: Moura_etal_SS2016_Learning_past_accidents_improve_system_design.pdf
Accepted Author Manuscript
License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 logo

Download (825kB)| Preview


Despite the massive developments in new technologies, materials and industrial systems, notably supported by advanced structural and risk control assessments, recent major accidents are challenging the practicality and effectiveness of risk control measures designed to improve reliability and reduce the likelihood of losses. Contemporary investigations of accidents occurred in high-technology systems highlighted the connection between human-related issues and major events, which led to catastrophic consequences. Consequently, the understanding of human behavioural characteristics interlaced with current technology aspects and organisational context seems to be of paramount importance for the safety & reliability field. First, significant drawbacks related to the human performance data collection will be minimised by the development of a novel industrial accidents dataset, the Multi-attribute Technological Accidents Dataset (MATA-D), which groups 238 major accidents from different industrial backgrounds and classifies them under a common framework (the Contextual Control Model used as basis for the Cognitive Reliability and Error Analysis Method). The accidents collection and the detailed interpretation will provide a rich data source, enabling the usage of integrated information to generate input to design improvement schemes. Then, implications to improve robustness of system design and tackle the surrounding factors and tendencies that could lead to the manifestation of human errors will be effectively addressed.