Fine blanking, a high precision metal forming technique that enables smooth sheared surfaces, without the need for expensive post processing, is susceptible to defects like cracks, burrs, roll-over or sheared edges. Direct quality measurement of mass-produced parts is time-consuming and costly, thus indirect quality assessment based on process data can help to significantly reduce costs and ensure part quality. Process data has been proven to contain various information about physical phenomena that occur during forming operations, allowing for the indirect measurement of target values like product quality or tool wear. However up to date it remains a challenge to assure accurate and reliable model performances in ever changing non-optimal production environments, where data is noisy, incomplete and filled with redundancies. The fusion of multiple complementary data sources has the potential to reduce model uncertainty due to incomplete, inaccurate or siloed information, leading to overall more informed decisions. In past research, signal-, feature- and decision-level fusion techniques have been investigated for some manufacturing processes, but the challenges, necessities and potentials are very much dependent on the specific use. This paper therefore investigates the potential of data fusion for the fine blanking process, using machine learning and feature fusion techniques for the prediction of cracks based on force and acoustic emission data.