Measuring and improving adaptive capacity in resilient systems by means of an integrated DEA-Machine learning approach

Abstract

Resilient systems strive to enhance the safety of complex systems through building and developing adaptive technological and organizational capacities. This study aims at analyzing and improving the level of adaptive capacity in a petrochemical plant by means of an integrated quantitative approach. The data were gathered by a questionnaire whose reliability is examined by statistical methods. To compute and analyze the influence of resilience engineering (RE) indicators, teamwork, and redundancy on adaptive capacity, data envelopment analysis (DEA) method was used. The results indicate that teamwork and redundancy have a positive effect on enhancing the level of adaptive capacity. Multilayer perceptron (MLP), a machine learning approach, was used to estimate the level of adaptive capacity on the basis of a dataset. The results of DEA and MLP approaches are confirmed by statistical methods. To the best of our knowledge, this is the first study that measures quantitatively and improves adaptive capacity by an integrated DEA-MLP approach based on the stress-strain model. The outcomes of this study could assist managers and other decision-makers of complex systems to compute and improve the level of adaptive capacity for coping with upcoming events in abnormal conditions.

Keywords: Adaptive capacity; Data envelopment analysis (DEA); Machine learning; Redundancy; Resilience engineering (RE); Teamwork.