Digital Health and Safety Risk Library Use Case

Dr Carlos A. Osorio-Sandoval; Dr William H. Collinge; Dr Clara Cheung; Dr Mojgan Hadi Mosleh; Dr Patrick Manu; Dr Andre Freitas; Dr Zili Zhou

risk management
Construction
Download the full report as a pdf file

This report presents the work undertaken during the Phase 2 – from June 2020 to June 2021 – of the BIM Risk Library Use Case of the Discovering Safety programme. Continuing with the work reported in the BIM Risk Library Final Report (Phase 1), the new version of SafetiBase, which was developed by 3D Repo based on collaboration with the Discovering Safety team, moved from a prototype to a piloting phase. The Risk Library, implemented within this platform, was deployed across multiple projects from the industry partners that joined the Community of Practice established by the Use Case team.


These industry partners piloted the new version of SafetiBase in a range of construction projects of different types, including residential, commercial, industrial, and infrastructure projects. New risk scenarios and treatments were retrieved from the pilot projects to improve the Risk Library knowledgebase.


Concurrently, risk scenario were also retrieved from the UK Health and Safety Executive’s (HSE) Archive data, for which a Natural Language Inference (NLI) approach was proposed to automatically classify free text into target categories.


This Phase of the Use Case evaluated the implementation of the Risk Library within a 3D BIM environment, as well as the data structure that underpins it, through qualitative research methods. Furthermore, the Use Case also conducted an investigation into the applicability of these concepts in a 4D BIM environment, focusing on how 4D can be exploited to enhance the safety risk identification process. The main findings of this Use Case are outlined below:

  1. Upon completion of the pilot project phase, the Risk Library contain 401 treatments for 31 different risk scenarios related to 11 risk categories.
  2. The NLI experimental work demonstrated that the approach is a viable way for automatically classifying accident records into an accident kind category, which is related to the Risk Library schema.
  3. Evaluation of the new version of SafetiBase using expert survey and individual interviews showed that the tool has practical utility in industry. Suggestions to further enhance the practical utility of the tool were also offered by the participants in the evaluation activities.
  4. Overall, the Phase 2 study shows the benefits that the Risk Library can bring to the implementation of design for safety in the construction industry. Design professionals should be encouraged to engage with it to promote its wide use in industry and its continuous development.

Aims and objectives

Phase 2 of the Risk Library Use Case had several key aims. These were: (1) to expand the Risk Library developed during Phase 1 by including more risk scenarios and treatments; (2) to evaluate its implementation within a BIM environment through a series of Pilot projects with industry; (3) to explore implementation of the risk scenario/treatment ontology and Risk Library within a 4D BIM environment; (4) to explore methodologies for capturing risk scenarios from the HSE archive using automated data processing techniques.

To achieve the above aims the following objectives were established:

  1. Expand the Risk Library from HSE Archive data
  2. Engage with industry partners to pilot the implementation of the Risk Library within the new version of 3D Repo’s SafetiBase
  3. Prepare a user manual for pilot project participants
  4. Expand the Risk Library from pilot projects’ data
  5. Hold a series of 4D workshops with industry to explore implementation of the Risk Library within a 4D BIM environment
  6. Evaluate the new version of 3D Repo’s SafetiBase from practitioners’ perspective

Key findings

The work done and evidence gathered from Phase 2 of the BIM Risk Library Use Case validates the original conceptual risk scenario/treatment ontology and the premise that health and safety in construction is open to improvement via proactive use of a tool linked to a digital BIM model. The implementation of the Risk Library within an existing commercial cloud-based BIM tool, namely 3D Repo – SafetiBase, was well received by industry practitioners and enabled the exploitation of both 3D and 4D functionalities.

Piloting the tool with a wide variety of projects in industry proved that the concept of a digital tool to assist in safety management is sound. Visualising safety information within a BIM environment provides context to the identified risks, which enhances communicating risks to other stakeholders. Subsequently, this approach provides an opportunity to mitigate these risks with improved design, construction methods, and planning.

The Pilot Projects also enabled further data to be collected from industry regarding real risk scenarios and treatments from live projects: this data feeding into a growing risk library database of potential value for all industry to use and share.

The authors believe the Risk Library is a promising means to leverage knowledge and regulations on construction safety across multiple projects and organisations – the methods for expanding the Risk Library also facilitating the continuous improvement of the knowledge base. In particular, automatic classification of free text into the Risk Library categories allows gaining semantic information from documented past events.

In addition to the general conclusions, there are several issues of note of relevance for the further use/development of the tool and further development of a risk library:

  • The ontology used to characterise risk scenarios provides information consistency and a way for structuring information uniformly across projects and across companies. This has the potential to create a uniformity of approach in the future. Several pilot project participants commented on the potential positive outcomes of structuring risk data uniformly, in design, maintenance and operations of buildings.
  • The creation of an embryonic risk library from Phase 1 and Phase 2 work holds great potential for further development that connects with wider societal ambitions to create open, trustworthy data eco-systems for the benefit of all (c.f. the Open Data Institute: https://theodi.org/). The open sharing of data related to construction risks and their treatment would potentially advance national and international practices.
  • The use of the digital health and safety tool was noted as a valuable learning tool opportunity for future designers. The learning potential of the Tool could be researched in the future.
  • Several Pilots noted how the Tool is very useful for more unusual, unique risks encountered on projects, rather than routine risks.
  • The use of the Tool to assist on projects in 4D generated much interest from industry (see Section 3.5). Use of a “dynamic” risk register in a 4D environment was noted as particularly exciting and innovative. Use of the Tool challenged and changed working practices and the role of designers in the risk/treatment identification process. Whilst different pilot projects approached use of the Tool differently (i.e. team approach / individual designer approach), there were some commonalities. For example, the use of Excel spreadsheets to capture and communicate risks to contractors was questioned, as well as the formatting/structuring of the spreadsheets.
  • The contractual relationships and organisational set-up between clients and design teams on different pilot projects effected use of the Tool. Very positive use of the Tool and feedback was received from pilots where the client managed designers themselves in-house; more challenging situations arising where clients/designers were separated by contracts and distance.
  • One pilot project advanced the innovative use of the tool by integrating it with their own procurement and tendering process: assessing tendering parties through their use of the digital tool in identifying solutions to design questions.
  • The risk scenario and treatment validation process was time consuming. It is suggested that this process needs to be streamlined and made more efficient moving forward. One potential solution would be to have 3 treatments per scenario. Consultation with the Community of Practice on this issue is advised.
  • Working at scale challenge: Phase 2 work proved the veracity of the BIM risk library concept and the value of a digital tool to assist designers working on projects with health and safety on a small number of projects from diverse sectors. A challenge now is to scale up usage across industry. This is a multi-faceted challenge requiring attention on several issues: communication/publicity; senior management support and drive; software issues (extension from 3D Repo/SafetiBase); more explicit value data generation from tool use at company/project level (metrics); tackling contractual/organisational challenges with use of tool.
  • Potential HSE benefits: exploration of potential benefits to HSE from the Tool and the NLP approach to interrogating the HSE archive could be investigated. For example, the ontology concepts employed for the risk scenario definition could be integrated with HSE investigator work sheets/reports so there is an integration of concepts and more integrated workflows at multiple levels. The NLI approach employed in Phase 2 opens up options to do statistical analysis of HSE resources (e.g. RIDDORs analysis for % type of accidents, projects, etc).
  • The NLP work has laid groundwork for further HSE archive interrogation: the NLI coding could be matured into a tool, with further methodological refinements made if required (annotation of data and sampling/checking for heterogeneity and accuracies).

Recommendations

On the basis of the work done in Phases 1 and 2 of the Risk Library use case, future work could include:

  • Leveraging the contextual information within BIM models to streamline risk scenario identification.
  • Aligning the proposed NLI classifier to the Risk Library ontology to identify recurring risk scenarios in documented past events in order to address them with appropriate treatment prompts that can be appended to the Risk Library.
  • Developing a training tool (e.g. serious game) to familiarise new users with the categories used in the Risk Library to facilitate risk characterisation.
  • Pilot testing the implementation of the Risk Library within a 4D BIM environment in a live construction project to better understand how this approach could leverage the time element of 4D models.
  • Adopting a data structure approach to characterise non-safety related risks, such as business risks and occupational health risks.