Biomimetic Machine Learning approach for prediction of mechanical properties of Additive Friction Stir Deposited Aluminum alloys based walled structures

Aug 4, 2026

This study presents a biomimetic machine learning framework for predicting mechanical properties of Additive Friction Stir Deposited (AFSD) aluminum alloy walled structures. Finite element analysis simulates the AFSD process for five alloys (AA2024, AA5083, AA5086, AA7075, AA6061), generating 200 simulation samples capturing coupled thermal and mechanical effects. Decision Tree and Random Forest regression models optimized by genetic algorithms forecast von Mises stress and logarithmic strain. The GA-RF model achieves R²=0.9676 for stress and R²=0.7201 for strain, supporting process understanding and parameter optimization across alloys.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages26

Reading Time1 h 6 min

AccessFree to download

1. Introduction

AFSD process overview and key parameters

Motivation for machine learning in AFSD

2. Materials and Methods

# Biomimetic Machine Learning approach for prediction ofmechanical properties of Additive Friction Stir DepositedAluminum alloys based walled structures Akshansh Mishral,* 'School ofIndustrial and Information Engineering,Politecnico Di Milano,Milan,Italy Abstract:This study presents a novel approach to predicting mechanical properties ofAdditiveFriction Stir Deposited(AFSD)aluminum alloy walled structures using biomimetic machinelearning.The research combines numerical modeling of the AFSD process with geneticalgorithm-optimized machine learning models to predict von Mises stress and logarithmicstrain.Finite element analysis was employed to simulate the AFSD process for five aluminumalloys:AA2024,AA5083,AA5086,AA7075,and AA6061,capturing complex thermal andmechanical interactions.A dataset of 200 samples was generated from these simulations.Subsequently,Decision Tree(DT)and Random Forest(RF)regression models,optimizedusing genetic algorithms,were developed to predict key mechanical properties.The GA-RFmodel demonstrated superior performance in predicting both von Mises stress(R²=0.9676)and logarithmic strain (R²=0.7201).This innovative approach provides a powerful tool forunderstanding and optimizing the AFSD process across multiple aluminum alloys,offeringinsights into material behavior under various process parameters. Keywords:Additive Friction Stir Deposition;Additive Manufacturing;Machine Learning;Hybrid Algorithms ## 1.Introduction Additive Friction Stir Deposition(AFSD)is a friction stir-based additive manufacturingprocess that involves the layer-by-layer deposition of material using feedstock,substrate,andtools [1-5].This procedure is based on friction stir welding (FSW),in which the materialexperiences intense thermoplastic deformation without melting,resulting in a thin,equiaxedmicrostructure in the finished product.There are three main variants of friction stir-baseddeposition additive manufacturing technologies:Friction Surface Deposition AdditiveManufacturing(FSD-AM),Friction Extrusion Additive Manufacturing(FEAM),and AdditiveFriction Stir Deposition(AFSD).Friction Surface Deposition Additive Manufacturing(FSD-AM)involves using metal rods as feedstock.These rods are fixed to a spindle that rotates andpresses down,generating heat through friction stirring.The plasticized material is then layeredonto the substrate to form the additive component.As the spindle moves along the processtrajectory,the component is formed,although the material experiences unconstrainedexpansion in both radial and axial directions,often resulting in curled edges around the rod.Friction Extrusion Additive Manufacturing(FEAM)uses metal rods which are transformed into a plastic state through friction with a rotating die driven by axial force.The plasticizedmetal is then extruded from the die outlet and fills the gap between the substrate and the tool,forming the component as the spindle moves.This method,however,tends to produce a poorlybonded layer due to the frictional interaction between the feedstock and rotating die.AdditiveFriction Stir Deposition(AFSD),which is shoulder-assisted,generally employs rods,wires,orpowders as raw materials [6-10].These materials are introduced into a hollow,non-consumabletool and,under extrusion,friction,and stirring effects,become thermoplasticized and migratedownwards to the substrate.The mechanical mixing of the softened substrate and plasticizedraw material creates a robust bond,followed by component formation as the spindle traversesa predefined path.Compared to the other two techniques,AFSD offers more precise controlover material flow and forming morphology.Several essential parameters influence the AFSDprocess,such as tool rotating speed,feed rate,and layer height.The rate of heat generation isprimarily determined by the tool's spinning speed,whereas the feed rate or axial force governsthe rate of material deposition.The tool traverse velocity influences the geographicaldistribution of he

FAQs

What is the proposed approach for predicting AFSD mechanical properties?

The study integrates numerical modeling of the AFSD process with genetic algorithm-optimized machine learning models to predict von Mises stress and logarithmic strain.

Which aluminum alloys and prediction targets are used in the work?

Five alloys are simulated (AA2024, AA5083, AA5086, AA7075, AA6061), and the models predict von Mises stress and logarithmic strain for deposited walled structures.

How were the training data and models built?

Finite element analysis produced a dataset of 200 samples, then Decision Tree and Random Forest regression models were trained and optimized using genetic algorithms.

Aug 4, 2026

Upload by Genevieve