Saving Lives with Statistics - An Introduction to Data Science in Workplace Safety

Abstract

Workplace safety is a main objective of any company working in the oil and gas business. The processes have been developed and established over the past decades based on individual experiences and causal pathways. The exhaustion of technical and administrative barriers has led to the introduction of behavioral safety. Recent advances in data technology and machine learning have disrupted many businesses and processes and can lead to a new paradigm in workplace safety as well. In this case study we demonstrate the application of data science and predictive analytics to aid the HSE function and prevent accidents. We have analyzed operational and accident data from the past 10 years at a leading oil and gas company to quantify the effectiveness of their safety programs. We have determined how many accidents each program actually prevents, and is able to prevent in an optimal setting. We have determined the optimal level of engagement for each program, and at what level diminishing returns set in. We have further developed a predictive model to forecast the occurrence of accidents one month ahead of time. In this way the HSE function is able to focus on 15% of locations to control 69% of the accidents. The forecast was also able to predict accidents at locations where one would traditionally not expect accidents to happen, such as locations with low activity. This paper shows the potential for improvement that is possible with the emerging big data, artificial intelligence and machine learning tools specifically in the field of workplace safety.