Fracture fixation plates are a type of orthopaedic implant that is used to repair broken bones. They are either left in the body or removed after a bone healing period of 6-8 months. Commercially pure titanium (cp Ti) plates although possess good mechanical strength are not suitable for biodegradable applications. Different materials were tested for their biocompatibility to avoid revision surgery, demonstrating the potential of Fe-Mn as it promotes bone homeostasis. However, their rate of corrosion is very low, with reported values of 0.07 mm year^-1, whereas biodegradable fixation plate materials require a corrosion rate of 0.5 mm year^-1. Studies on porous alloys of Fe-Mn have revealed an increase in corrosion rate for up to 0.8 mm year^-1, but at the expense of mechanical strength making them suitable only for scaffolding of cancellous bones.

Triply Periodic Minimal Surface (TPMS) designs have the potential to create engineered pores with controlled porosity by adjusting the isovalue, which allows scaling of pore size. In this work, three designed samples of gyroid based TPMS cells with isovalues varying from 1.330 to 1.056 having porosity level ranging from 5-15% are developed, that mimic property of cortical bone. Ntopology software is used to perform Boolean intersection of gyroid cells and implant structure to develop porous fixation plate for Metal Additive Manufacturing (MAM).

This study evaluates the flexural strength and stiffness of the designed fracture fixation plate by simulating four point bending test on Ansys. A comparative FEA analysis is presented on mechanical properties for TPMS designed Fe-Mn alloy and cp Ti.

Additive Manufacturing (AM) of metals has emerged as a key discipline to produce parts that enable certain performance advantages. However, the list of available alloys for AM remains limited. Metal AM is challenging due to the high number of passes required, especially for alloys with poor weldability such as Inconel 738 (IN738). This nickel-based superalloy, with its high creep resistance, is very difficult to process commercially via AM due to cracking susceptibility. However, there is literature suggesting feasibility, although it may result in regions of high deformations and stresses. If such areas can be predicted in advance, they can be improved through design modifications. AM process simulation is a tool that enables such deformation predictions. In this work, we use a commercial simulation tool, Simufact Additive, to calibrate and predict the deformation of simple geometries produced in IN738 via laser powder bed fusion AM. Using the deformation of a printed geometry the set of inherent strains in simulation could be calibrated to match the actual printed deformation maximum. Using the calibrated inherent strains predictions on other geometries could be made and compared to their printed counterparts. Through this procedure, it was found that the inherent strain method cannot predict every form of deformation adequately. When the prediction geometries and the calibration geometries have similar main driving factors for their deformation, deformation deviations of <20% were achieved. The prediction is less reliable when using the calibration on a deformation stemming from stress relief occurring through the cutting of a printed part.

Topology optimization is a useful tool for exploiting the lightweight potential of structural components and is often used in design for additive manufacturing. In preparation of topology optimization, the design area is divided into a design space and a non-design space. The design space is the area available for optimization, and the non-design space is the area where the geometric and physical boundary conditions, such as support and loads, are defined. To identify the geometric boundary conditions of a component, the features of the component suitable for transmitting loads must be identified. This process is typically done manually and is therefore very time-consuming and costly.

This paper presents an automated identification of non-design features of CAD models for topology optimization using learning-based segmentation. The CAD files are taken from a large database of industry-relevant components. Based on the geometry and topology of the components, a graph structure is created and processed by a deep neural network. The results show good match with real cases and can be continuously improved by training with additional data.