Reducing the redundancy in high dimensional data and finding the most relevant features is an important task in any data-driven approach. Especially, when the data consists of several datasets recorded from multiples sources. In fact, with such configuration, the redundancy can be within one source or even between different sources. This work explores a case study from manufacturing production process, in which, each step of production is considered as a source of data and contains many parameters (features). To reduce the dimensionality, an unsupervised feature selection has been applied using a genetic algorithm as search strategy.