Digital Framework Solutions for Enhancing Additive Manufacturing Workflows

Additive manufacturing (AM) has transformed the manufacturing landscape providing decentralised production options while opening new opportunities for manufacturing highly customised and complex parts. As opposed to the tried and tested conventional manufacturing (CM), the data pipelines of AM are vulnerable and the ease with which intellectual property (IP) can be compromised by theft or malicious attacks, creates a significant challenge, and needs to be addressed. AM also comes with some intrinsic process shortcomings, such as the interactions between feedstock materials, the need for post-AM processes, and the lack of process robustness, which have hampered its widespread adoption in the industry. Additional AM challenges include repeatability and quality consistency, low build rates, high cost of materials, the need for continuous process monitoring and control, as well as lack of standardisation in many parts of the AM workfl ow. These challenges are being investigated and addressed in this thesis through the process of reviewing and analysing the data from various phases of the AM workflow. The research focuses on a series of ideas that are discussed below.
One of the primary objectives of this work is to review those product data and lifecycle management (PD and PLM) approaches in the AM workflow that facilitate the sharing of product data among engineering and manufacturing parties. AM is forcing companies to rethink their strategies to account for its implications across the entire product lifecycle. Next, the AM workflow was investigated from an AM resources capacity perspective in order to develop an agent-based decision support platform, which would be capable of proposing alternative additive manufacturing resources and process configurations to design engineers, while reducing the number of communication steps among engineering teams and organisations. In the next phase, the AM workflow was further investigated in order to automate engineering processes with the help of robotic process automation (RPA) technologies. A commercial RPA platform is employed to create an automated workflow, which can consider multiple AM process configurations. Furthermore, the need for AM tools that can track variations and increase consistency withing the AM workflow was identified. With that in mind, the research also proposed a modular framework that enables engineers to update and optimise design changes as well as the AM process configuration considering a series of design features. The framework is capable of considering and aligning 3D data related to different phases of the workflow, within an integrated data management model. The developed alignment algorithm was based on the Bounding-Box technique whilst the data model for structuring the framework was built on top of the Hierarchical Data Format (HDF5). The framework then utilises a multi-step classification system to consider additional information related to mechanical properties of past builds, such as porosity, which can then be utilised to calculate the distribution of the same mechanical properties across new builds. Using a defect-centric (Biased) and uniform discretisation (Unbiased) methods, the framework then utilised machine learning methods with discretisation techniques to predict defective and non-defective zones in printed parts.
The outcomes of this research provide researchers and practitioners with a variety of novel approaches and tools that can be used for improving cost, time and quality performance as well as process robustness. In addition, engineers may take advantage of the proposed approaches in order to consider the available AM resources and process configurations during the product design phase.