Near Miss Archive: a Challenge to Share Knowledge Among Inspectors and Improve Seveso Inspections

Abstract

In European Seveso Legislation for the control of the hazard of major accidents (Directive 2015/12/UE), the Safety Management System SMS is an essential obligation for managers and the authorities are required to periodically verify its adequateness through periodical inspections at Seveso sites. One of the pillars of the SMS is the collection and analysis of documents on accidents, near misses and possibly anomalies, in order to identify weaknesses and implement continuous improvement. In Italy, for a few years, the documents, gathered from all Italian Seveso sites by the inspectors, have been archived and used for research purposes. The archive currently contains some 4000 reports, collected in five years by some 100 inspectors throughout Italy. The paper discusses in the detail the challenges faced to extract the knowledge hidden in the documents and make it usable through the design of a robust model. For this aim, Machine Learning techniques have been used as a preprocessing of the reports for extracting the concepts and their relations, organized into an entity-relation model. The effectiveness of this methodology and its potentiality are pointed out by investigating a few hot topics, exploiting the information contained in the repository.

Modelling the relationships between the barriers to implementing machine learning for accident analysis: the Indian petroleum industry

“…, 2021). Data privacy and security issues have also been reported in numerous studies (Ansaldi et al. , 2021; Khatri et al.…”

“…Data collection and inconsistent data formats: Following an accident, an investigation team analyse multiple data sources and prepare a report that is then sent to concerned stakeholders (i.e., internal departments, government agencies, regulatory bodies). A major flaw with this process is the use of different data taxonomies which have inconsistent data formats (Kaisler et al, 2013;Kim et al, 2003) , as well as curating and validating data (Ansaldi et al, 2021).…”

“…Data privacy, security and access: Access to accident datasets has been reported as a barrier for implementing ML to analyze accident data and develop predictive models (George & Renjith, 2021;Dixit et al, 2021). Data privacy and security issues have also been reported in numerous studies (Ansaldi et al, 2021;Khatri et al, 2020;Wang et al, 2018). Due to these issues, oil companies are highly susceptible to cyber-attacks (Whitworth & Suthaharan, 2014;Agrafiotis et al, 2018).…”

“…The implementation of new ML initiatives is frequently blocked by organizational resistance to change (Ansaldi et al, 2021;Scholkmann, 2021;Schuetz & Venkatesh, 2020;Côrte-Real et al, 2019). Organizational resistance to change occurs when organizations lack readiness and willingness to change their current technologies and underlying processes (Dennehy et al, 2021;Dubey et al, 2015;(Raut et al, 2021;Alharthi et al, 2017).…”

Towards Learning Causal Representations of Technical Word Embeddings for Smart Troubleshooting

“…In terms of application, a direct implementation of this developing approach could be driven by a retrieval-augmented generation system for work orders to advise the maintenance team by identifying the most probable underlying root cause to a given problem, and reduce both the time to action and asset downtime while increasing the safety of the railway service (Ansaldi, S. M., Agnello, P., Pirone, A., and Vallerotonda, M. R., 2021). This enhanced troubleshooting system would equip a model that combines pre-trained parametric memory (i.e., the causality-contextualized word embedding) and non-parametric memory (i.e., a classic data retrieval-based engine) for language generation (Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-t-, Rocktäschel, T., Riedel, S., and Kiela, D., 2020).…”