# Machine Learning and RSM for Lattice Structure Optimization

## TL;DR

This paper uses machine learning and RSM to optimize the design of lattice structures for a motorcycle throttle cam, aiming to achieve the best stiffness-to-weight ratio.

## Contribution

The novel aspect is combining RSM and ANNs to identify optimal lattice design parameters for functional performance.

## Key findings

- RSM and ANNs both identify optimal design points for lattice structures based on geometric parameters.
- The study confirms the non-linear behavior of lattice structures through both RSM and NN analysis.
- The approach aims to improve upon incomplete practical applications of lattice structures in industry.

## Abstract

This study concerns the analysis of lattice structures printed with EPAX resin for the manufacturing of a motorcycling throttle cam with Response Surface Methodology (RSM) and Artificial Neural Networks (ANNs). The design of the pattern core in the lattice structure is defined parametrically to identify optimal design points (best stiffness to weight ratio in particular). Some geometric parameters used as input in RSM and in the NN analysis include the origin of the lattice structure and its spatial orientation, cell dimensions, and thicknesses. The dataset obtained with this approach is used for an RSM analysis of variance (ANOVA) to highlight the most important inputs. NN analysis is performed on the same RSM dataset to confirm the results. Both methodologies identify in-domain points of optimal design due to the typical non-linear behavior of these structures. The literature and industrial experience already provide numerous references to studies characterizing lattice structures. However, related practical applications are often incomplete and only achieve functional rather than optimal models. The approach described also aims to overcome this limitation. The software used for the design is nTop 5.0.4.