A proposal of a mixed fuzzy logic/rough sets FRAM model for complex systemic analyses of sociotechnical systems

Abstract

Modern sociotechnical systems have been continuously developing at a fast pace leading to more complexity and interconnectivity among their components. The traditional view in wellestablished risk and safety assessment methods adopted a linear cause-effect philosophy focusing on what could go wrong (SAFETY-I). The focus was mainly and simply directed to the identification of errors, root causes and singular events, which could directly cause accidents and lead to undesired outcomes. However, considering the complexity of sociotechnical systems nowadays, classical approaches might not be sufficient to cover all influential factors affecting their performance. Different and innovative approaches capable of capturing such complexity are therefore required. Such approaches must adopt a systemic perspective and consider the various factors in a complex socio-technical system and their interactions (resonance). This shall provide an understanding of how undesired events might develop and emerge out of simple performance adjustments and the combination of systemic functional variability. Indeed, the field of safety management witnessed a significant amount of research efforts to address these issues leading to the proposition of several novel methods and approaches adopting fresh perspectives. One of these recently emerging disciplines is the field of Resilience Engineering adopting a new SAFETY-II perspective on the topic of safety management. The main method proposed in Resilience Engineering is the Functional Resonance Analysis Method (FRAM), which shall serve as the main tool in this project. Despite the many advantages offered by these new tools, they are still limited in many ways and require further research and development to mature and become more standardized and established. The use of qualitative scales in tools such as FRAM allows for the capture of complex and dynamic relationships using natural language. However, this induces other limitations for the interpretations of the significance and for determining the precise magnitude of the produced outcomes. The addition of quantification into such approaches without sacrificing the distinguishing characteristics of these tools would allow the analysts to benefit from the advantages offered by both quantitative and qualitative methods. A possible solution to this limitation can be offered by fuzzy logic, which facilitates computing with natural language. The advantages of fuzzy logic are represented in the capacity to account for uncertainties in the defined parameters and the addition of quantification to FRAM, which rely on linguistic variables. Through the combination of fuzzy logic and FRAM, the benefits of the two approaches can be utilized to present a new powerful tool for risk and safety assessments. The integration of fuzzy logic can facilitate as well the proposition of standardized framework to account for variability in FRAM and perform predictive assessments using the rule base of the Fuzzy Inference System (FIS). In contrast to retrospective accidents analyses, predictive safety assessments must deal with uncertainties and vagueness arising from the lack of sufficient data or knowledge on the true nature and magnitude of the evaluated concepts. Additionally, the rule base generation process in the FIS is exhaustive and requires significant efforts in case of a high number of input variables. This leads in turn to the generation of large rule bases heavy on computing resources and unfeasible in terms of expert elicitation. The Rough Set Theory (RST) framework can provide solutions to these issues as a data mining and classification tool. Through the application of several search algorithms to analyze input data provided by experts or field observations, the RST approach allows for an automatic classification of data (whether quantitative or qualitative) and consequently the generation of efficient reduced rule bases. The starting point for this research project was with the analysis of the environment of aircraft ground deicing operations, which constitutes a complex and dynamic sociotechnical system. The optimal performance of the deicing procedures relies on many factors such as environmental conditions, technology, organizational aspects, and the human factor. The main objective of this research project is to introduce a new approach for safety assessments of complex socio-technical systems, specifically the context of aircraft deicing. To ensure the integrity of deicing/anti-icing procedures and provide the desired level of safety, an assessment of influential factors is necessary. The specific objectives are to model the system in place and perform a systemic assessment applying the Functional Resonance Analysis Method (FRAM) to determine the influential factors that affect the performance of deicing operations; to introduce a new systematic approach combining fuzzy logic with FRAM to allow for a more standardized representation of performance variability; and thirdly, to integrate the RST method into the framework to provide the tools to classify large datasets and automatically generate comprehensible rule bases. Each phase in this project provided an application scenario inspired by actual airplane accidents related to deicing operations and the obtained results were compared eventually to draw conclusions and validate the results from a theoretical angle of view. The results obtained from each application revealed new findings and illuminated some understudied areas concerning aircraft deicing. The analysis of the Scandinavian crash SK751 allowed for modelling the system of deicing providing an explanation for the development of the accident and linking events from a functional perspective. The integration of fuzzy logic facilitated the computation of the output’s variability in a more precise way and provided a systematic and standardized framework to account for performance variability. The RST approach as declared in the objectives enabled the automatic generation of reduced rule bases without sacrificing accuracy using ideal datasets. From a technology-readiness perspective, the proposed model is still in the early stages and requires further validation using real world data and further applications and optimization. The scope and timeline of this project did not allow to dig deeper; however, it is hoped that this project shall initiate further research activities in the future addressing the tackled topics and issues. Furthermore, the adoption of frameworks as fuzzy logic and RST has been very limited in the field of risk and safety management. Such tools offer innovative and different approaches to cope with uncertainty and problems related to data classification. This in turn could assist in improving safety measures and minimizing risks to provide better aircraft ground deicing/anti icing operations and better protection for humans and machines. The new perspectives offered by these tools could eventually reflect positively on the economic, technological, and organizational aspects of the deicing industry and any other industrial context concerned with safety management. In the end, it is hoped that this research shall provide some answers and open the door for more arising questions to initiate as a conclusion further future studies.

Date	14 Dec 2020
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Sylvie Nadeau (Supervisor)