Industrial products integrate highly confi … Industrial products integrate highly configurable safety-critical systems which must be intensively tested before being delivered to customers. This process is highly time-consuming and may require associations between product features and requirements demanded by customers. Machine Learning (ML) has proven to help engineers in this task, through automation of associations between features and requirements, where the latter are prioritized first. However, ML application can be more difficult when requirements are written in natural language (NL), and if it does not exist a ground truth dataset with them. This work presents SRXCRM, a Natural Language Processing-based model able to extract and associate components from product design specifications and customer requirements, written in NL, of safety-critical systems. The model has a Weight Association Rule Mining framework that defines associations between components, generating visualizations that can help engineers in prioritization of the most impactful features. Preliminary results of the use of SRXCRM show that it can extract such associations and visualizations. ract such associations and visualizations.