A systematic literature review and taxonomy proposition of machine learning techniques in smart manufacturing

The purpose of this paper is to analyse the use of machine learning in smart manufacturing, describing techniques, technologies, industries, and purposes associated with industrial applications. We conducted a systematic literature review using Scopus, in which 26,032 documents were found. After applying quality criteria, 107 articles were analysed. The main findings show that machinery was the industry subsector with the major implementations regarding machine learning; process improvement is the main concern (interest) of all implementations; random forest was the most specific machine learning technique used; and diverse technologies associated with this context were identified such as: the industrial internet of things, digital twin, sensor technologies (soft, optical, barometric, ultrasonic), software technologies (Python, MATLAB, LabView, Google AutoML Platform) and equipment technologies (robotic, PLC, CNC). Most fault detection machine learning applications were focused on predictive maintenance, specifically in mechanical equipment (bearings, machines in general, and assembly lines). This study presents a novel taxonomy that identifies 85 specific machine-learning techniques used in smart manufacturing.