Predicting Human-System Interaction Risks Associated with Autonomous Systems in Mining

Abstract

Industry is implementing increasing amounts of automation into operations. The Australian mining industry is no exception as it is introducing autonomous mining vehicles and trains, remote controlled processing plants and the use of drones and robots to do survey and inspection work. Often these technologies are adopted to improve operational efficiencies and to reduce workers' exposure to high risk situations. However, in most mining environments, the adoption of automated technologies has not completely removed humans from the operation. Humans still need to interact with the technology to clean, service and maintain it. Humans also have to perform other tasks in the automated mining environment such as inspection of ground conditions, mapping mining and dump areas, maintaining roads and infrastructure etc. Thus, introducing automation into mining environments has the potential to introduce new and significant human-system interaction safety risks. The emergence of these new safety risks are evident in recent accidents in the mining industry as well as in other industries that have introduced automation. Traditionally, risk based approaches have been used in the Australian mining industry and other industries to identify and treat safety related risks. Such approaches include the use of hazard identification techniques (HAZID), Workplace Risk Assessment and Control (WRAC), Failure Mode and Effects Analysis or Failure Modes and Effects Criticality Analysis (FMEA or FMECA), and Process or Job based Hazard Analysis (PHA or JHA). These traditional techniques have helped reduce fatal and catastrophic incidents in the mining industry but deficiencies in their application has also been highlighted in a number of major accident investigation reports. In addition, recent research has suggested that that traditional risk identification techniques by not be effective for new, software-enabled technologies that are embedded in socio-technical systems with complex or dynamic human-system interactions. In response new socio-technical risk assessment approaches have been develop such as System Theoretic Process Analysis (STPA) and Strategies Analysis for Enhancing Resilience (SAfER). However no publications could be found that seek to understand from a end-user perspective the efficacy of the traditional and new techniques in assessing human-system interaction risks associated with the introduction of autonomous and automated technologies in mining environments.To begin to address this gap, research was conducted that sought to answer the question - What combination of risk assessment techniques delivers the most effective means of identifying risks associated with human-system interactions in remote and autonomous mining operations? The research method involved have mining industry professionals trial four techniques - Preliminary Hazard Analysis (HAZID), Failure Mode and Effects Criticality Analysis (FMECA), Strategies Analysis for Enhancing Resilience (SAfER), and System Theoretic Process Analysis (STPA) (Systems-theory Method) - in a workshop environment. Three different workshops were conducted each of which focused on a different automated technology. The first focused on identifying human-system interaction safety risks in surface mine automated haulage areas. The second focused on identifying human-system interaction safety risk associated with autonomous longwall mining operations underground. The third focused on human-system interaction safety risks associated with remote controlled operation of ore processing plants. After the workshop trialed each technique, the participants were survey to collect their perceptions of the usability and usefulness of each technique. Results from the participant feedback suggest that each techniques was able to identify potentially hazardous human-system interactions but that each had strengths and weaknesses depending on whether risks were being assessed risks pre or post implementation. A hybrid or combination approach was suggested with further testing of the proposed approach being recommended.