In this report, we present a prototype assembly support system for hydraulic components, tailored to a mechanical industry production plant. The case study for the project is Komatsu Forest, a leading provider of forestry machines. The system uses multimodal data analysis to understand the assembly process and detect errors. In particular, it uses a TensorFlow network to identify hydraulic components, a projective computer vision model to map a CAD drawing against the partial assembly, and natural language processing techniques to recognise patterns in the non-conformance reports.