At a time when the competitive market is operating rapidly, manufacturing industries need to stay connected, have interchangeability and interoperability in their factories, ensuring that there is heterogeneous communication between sectors, people, machines and the client, challenging the manufacturing industry to discover new ways to bring new products or improve their manufacturing process. Precisely because of the need to adjust to these new market demands, factories pursue complex and quick decision-making systems. This work aims to propose applications of Machine Learning techniques to develop a decision-making platform applied to a manufacturing line reducing scrap. This goal will be achieved through a literature review in the fields of Artificial Intelligence (AI) and Machine Learning to identify core concepts for the development of a failure prediction system. This research has demonstrated the problems and challenges faced by manufacturing daily, and how, through the application of AI techniques, it is possible to contribute to assist in these problems by improving quality, performance, scrap rates and rework, through connectivity and integration of data and processes. This paper contributes to evaluate the performance of machine learning ensembles applied in a real smart manufacturing scenario of failure prediction.