Evaluation of AI-driven Generative Design and Redesign of a MINI-LINK Mounting Kit
Examensarbete för masterexamen
Product development (MPPDE), MSc
This master’s thesis is a partial fulfillment for the degree of Master of Science in Product Development and aimed to assess the viability of incorporating AI-driven generative design within Ericsson’s mechanical design department. In this study, Ericsson’s specific needs and requirements for generative design were explored and identified through a series of interviews and a comprehensive user study. Subsequently, the identified needs were transformed into benchmarking criteria for evaluating the capabilities of different software in performing generative design or topology optimization. The primary objective of the benchmarking phase was to evaluate the extent to which various software options aligned with the benchmarking criteria and their proficiency in executing generative design or topology optimization tasks. Following evaluation against the benchmarking criteria, PTC Creo Parametric emerged as the highest scoring software and was consequently employed in the redesign of an existing mounting kit for a MINI-LINK radio. The outcomes of the redesign phase revealed promising advancements in the form of improved design that surpassed the performance of the pre-existing solution in terms of weight reduction, increased stiffness, and a lower total cost. As the complexity of the model, load cases and constraints increased in the redesign of the mounting kit, limitations with the current version of Creo were revealed. A potential explanation is the difficulty to combine a generic method as the contextual complexity and detail imposes specific constrains. Concluding the thesis report, a revised and improved workflow proposal for product development process within the mechanical design department was presented, in addition with the insights and findings obtained throughout the study.
Artificial Intelligence, Benchmarking, Deep Learning, Generative Design, Needs Identification, Redesign, Topology Optimization