https://www.mdu.se/

Dataset generation is cumbersome yet of great importance for successful training of machine learning models. Collecting real-world data is expensive and sometimes prohibited, considering e.g. safety aspects or legal restrictions. By generating the bulk of training data by synthetic means it is possible to impose arbitrary and extensive scene randomization for increased data diversity.

Methods to quantify similarity between datasets on a statistical level are important tools to detect Out-of-Distribution (OOD) data and domain alignment. We have studied how such methods can be used to correlate model prediction accuracy drop when exposed to OOD-data.

Domain adaptation can be applied as an additional step to synthetic data, to decrease the gap to real world datasets, however it can introduce inadvertent label-flipping, a sort of semantic inconsistency between synthetic source and domain adapted output. Therefore, we pursuit another way of reducing the domain gap, by generating high-fidelity digital representations of real-world scenes and objects. We do this through the use of Neural Radience Fields and Gaussian Splats. These methods allow us to render objects of interest for a detection problem, with the perfect annotation of synthetically produced data, and a high degree of realism which we show improves detection accuracy compared to traditionally generated visual content.

The low-volume manufacturing industry is facing an ever increasing competitive market environment, where high degrees of flexibility, complex products and short time-to-market are important factors to remain competitive. These market dynamics are a strong motivator for companies to design and introduce new products to the market rapidly, thereby placing an increased emphasis on the new product development (NPD) process. The integration of design and manufacturing functions has been identified as a critical enabler to improve the NPD process and reduce lead times of products. A successful design manufacturing integration however requires a significant amount of data and information to be exchanged by both functions. To ensure this exchange, a robust product data infrastructure is necessary, along with the capability of systems to communicate seamlessly. The ability to exchange data relies heavily on technical interoperability, which is essential to enable a seamless dataflow and has been shown reduce the need for manual interventions and increases accessibility to data and enables better coordination of resources. Existing literature has primarily focused on technical interoperability from a high volume manufacturing perspective, noting that lacking technical interoperability can significantly impact the performance of manufacturing processes, including inefficiencies regarding workflows, data exchange and increased costs. However, less emphasis has been placed on technical interoperability from the perspective of low-volume manufacturing industry. 

The purpose of this thesis is therefore to explore how technical interoperability can be leveraged to facilitate design manufacturing integration within the context of low-volume manufacturing industry. This was achieved by performing an in depth single case study where an NPD process of a low-volume manufacturing company was analysed. The findings identify current challenges related to technical interoperability, their impacts on the NPD process, and finally the steps required to leverage technical interoperability to facilitate design manufacturing integration. The primary theoretical contribution of this thesis is its addressal of a notable research gap concerning the role of technical interoperability in facilitating design manufacturing integration, as well as providing further insights into the challenges and their impacts on NPD processes due to deficiencies in technical interoperability from the contextual perspective of the low-volume manufacturing industry. The practical contributions of this thesis include a comprehensive understanding of existing challenges related to technical interoperability and the impacts these challenges can have on the performance of an NPD process. Furthermore, this thesis provides a step-by-step approach detailing how technical interoperability can be leveraged and exemplifies the process.