Automated Design of Additive Manufactured Flow Components

Abstract

Many prior works demonstrate the potential of additive manufacturing (AM) for flow components. Examples include nozzles, flow distributors, hydraulic manifolds, and heat exchangers. Compared to conventional manufacturing methods, such as milling and drilling, AM offers a high degree of design freedom and enables the fabrication of organic-shaped and functionally optimized flow structures. Such parts can be produced without additional tooling at reduced lead times, allowing the rapid and iterative testing of many design variants. In addition, AM enables the cost-efficient production of customized parts that can be tailored to the individual needs of specific customers or applications. Despite the potential of AM, a key challenge is to design complex parts for AM. In practice, a common approach is to manually create the 3D geometry of parts using computer-aided design (CAD) tools. However, a manual design process can lead to several challenges. First, it requires considerable expert knowledge of AM and skills with CAD tools, which can be a critical barrier for novice designers. Second, the design of organic-shaped parts can be time-consuming and require the creation of hundreds of design features. Third, designers must consider process-related restrictions of AM (e.g., prevention of critical overhangs) and may require several manual loops to analyze and modify design features for manufacturability (e.g., adaption of circular flow channels to droplet shapes). Fourth, the part development can involve many design changes, e.g., to include feedback from simulations and iterative tests, compare different production scenarios, and customize parts to specific needs. If such frequent design changes are performed manually, the manual effort, labor costs, and development time can significantly increase. This work aims to automate the design of complex flow components fabricated using AM. For this purpose, this work follows a knowledge-based engineering approach and implements rule- and knowledge-based design algorithms for specific flow components. In particular, this work focuses on multi-flow nozzles and hydraulic manifolds produced using the AM process of laser powder bed fusion. This work presents three studies, each focusing on a specific design challenge. Study I automates the design of complex AM multi-flow nozzles. The basic modeling idea is to decompose nozzles into a set of design elements that are used as the basic building blocks of nozzles and include recurring features, such as different cross-section shapes, flow channels, channel branches, guiding vanes, and reinforcement ribs. These design elements are organized using a hierarchical structure. This modeling approach allows to capture the hierarchical nature of complex nozzles and automate the design creation and nesting of multiple flow channels. Study II focuses on the automated consideration of the AM overhang constraint during the design generation of AM parts, such as hydraulic manifolds. For this purpose, the study models the dependency between geometric parameters (e.g., inclination of flow channel cross-sections) and process-related parameters of AM (e.g., build direction, minimal build angle, and maximum diameter of horizontal cross-sections). Based on these relations, this study demonstrates how to automatically create flow channels without critical overhangs inside the channels by locally modifying the shape of circular cross-sections to adapted shapes (e.g., droplet). In addition, this study shows how to generate integrated and sacrificial supports. The result is a production-ready 3D part design that can be used to fabricate prototypes or conduct simulations. Study III automates the routing of multiple flow channels for AM flow components, such as hydraulic manifolds. For this purpose, the study models flow channels as virtual cables defined by a chain of particles (= centerline of flow channels) and collision spheres (= required space of each flow channel). These cables are iteratively subjected to geometric-based constraints in order to impose different functional part requirements (e.g., minimizing the length of flow channels and preventing overlaps between different channels). In addition, the adaption of channel cross-sections for AM is taken into account by iteratively updating the radii of the collision spheres during the automated routing of flow channels. Based on the presented studies, a key conclusion is that a rule- and knowledge-based approach can be applied successfully to automate the design of complex AM flow components, such as multi-flow nozzles and hydraulic manifolds. Potential future research directions include transferring the results to different applications, further simplifying the automated design process, and integrating machine learning techniques.