The Industry 4.0 smart factories allow both optimization and integration of internal processes, utilizing the predictability of failure elements/components in a manufacturing process to prevent reprovals at the end of the process for quality control. The Supervised Machine Learning (ML) methods could be useful to detect anomalies and gain even more value throughout the entire supply chain. The ML approaches face barriers since it demands a changing in the production plant mindset to a more digital production and in the organization's structure for a more advanced data security. The paper aims to propose a smart inconsistency and fail prediction system for manufacturing systems of an automakers supplier assembly process based on the applications of ML techniques. The data provided for the training showed significant deviations and non-linearity allied to only 5 attributes as input variables, which is considered a small number of features for similar problems in the literature. The trained model was then applied to the assembly line with unobserved data of new products, with its result compared with similar previous productions. The results of the tests showed that the proposed stacking model lessens the possibility of rework in the next stages of assembly and creates a more precise process control for the supervisor. The implementation's results pointed out the potential of the stacking model proposed to be a useful tool in the context of Industry 4.0 since the reductions mean greater availability of production time and lower costs with quality control.