Review of interpretable machine learning for process industries

This review article examines recent advances in the use of machine learning for process industries. The article presents common process industry tasks that researchers are solving with machine learning techniques. It then identifies a lack of consensus among past studies when selecting an appropriate model given a prescribed application. Furthermore, the article identifies that relatively few past studies have considered model inter-pretability - a "black-box" challenge holding back machine learning's implementation in more high-risk in-dustrial applications. This interdisciplinary field of engineering and computer science is still reasonably young. Additional research is recommended to standardize methods and establish a strategic framework to manage risk during adoption of machine learning models.