Predictive Maintenance of Spray Equipment Using Machine Learning in Paint Application Services

Paint application services require adequate preparation to accommodate the enormous capital investment and rigorous quality requirements associated with conventional spray applications. The low volume-to-mass ratio of spray equipment components can result in costly downtime if not correctly maintained. Predictive maintenance utilizing machine learning and computer vision can mitigate potential failures through better planning and allocation of resources. This proof of concept work focused on the paint group components of a multi-purpose mobile spray unit, which serves to store the paint and propel it through an atomizing nozzle at pressure. A Rugged Edge AI computer was installed on a multi-purpose mobile spray unit fitted with paint group components, including a paint pump and hose, heat exchanger, paint heated chamber, and an air compressor. The data were collected from the mobile spray unit while conducting normal spraying activities in real-time with the wireless sensors united through Industrial Internet of Things architecture. The assessment of the functioning of the paint group components was done based on two types of data acquired from separate sources: 1) images of various locations, interactions, and conditions of a paint rubber hose when affected by different situations throughout the entire process of normal spray application and 2) data from wireless sensors assessing electrical consumption data from the input of the hydraulic pump while propelling paint through the other components. The analysis of the images was done using a machine vision application. The analysis indicated that there are detectable changes in the electrical consumption patterns generated by the data acquisition system sensors, pointing to some abnormal situations when executing the spray application, according to the inspection of the images.