Leaping into the Future: Artificial Intelligence for Health, Safety and Environmental Management

Abstract Performing complex activities on industrial workshop floors carries the risk that incidents and injuries might occur. Traditional health, safety and environment (HSE) monitoring solutions require significant manual intervention, are prone to biases and inaccuracies, and are reaching their limits. A company's technology lifecycle management (TLM) group realized there was a need for a proactive and innovative approach and developed an automated system to detect unsafe conditions and acts, provide timely alerts for prevention or mitigation, and influence human behavior. A unique software application, named Digital Workshop, was developed within the company's TLM asset performance management platform. The software leverages advancements in artificial intelligence (AI) using on-site workshop camera feeds coupled with customized vision analytics to automatically detect, categorize, and report HSE non-compliance events. Examples of events that can be automatically detected include zone intrusions, restricted areas access, solo worker activities, personal protective equipment (PPE) adherence and general housekeeping as well as dynamic high-risk scenarios including mechanical lifting, forklift operations, lathing, grinding, welding, and the use of rotating machinery. HSE non-compliance notifications are sent to mobile phones and recorded in the company's system of records enabling corrective actions to be taken. This AI-based safety approach was successfully implemented in the company's workshop environments, on shop floors and in operational facilities. The application generates HSE-related statistics and insights that can greatly enhance location performance, by helping to avoid incident recurrence, and improve the working environment for personnel. The application can identify HSE unsafe behaviors or practices, such as PPE non-compliance, forklift proximity, mechanical lifting non-compliance, zone intrusion, working at heights and other high-risk activities. Early detection enables timely intervention and incident prevention. The system has been particularly beneficial to area managers and supervisors who conduct pre-operational and regular safety meetings. The information is valuable to teams tasked with reviewing and addressing non-compliance events and implementing continuous improvement actions. Recorded footage can be used for learning and coaching moments and helps identify areas where additional control measures or personnel training may be needed. This has enabled detection, analyses and corrective actions associated with undesirable behaviors and unsafe practices to be implemented. Post-deployment benefits have ranged from safety behavioral changes to creative solutions being proposed by personnel. It has been found to be especially beneficial for contractors and new hires, who typically experience a higher rate of incidents. The technology is scalable to different work environments, including while driving and in non-company operated facilities. This innovative technology uses proprietary machine learning algorithms, developed in-house, to detect a wide range of HSE non-conformance scenarios. The AI model has been trained for various camera positioning and data from different locations to account for varying conditions and environments in the company's facilities. The system has been taught to identify the HSE non-conformances and automatically issue alerts or notifications to end users.