This thesis work, conducted in collaboration with an industrial company, focuses on the development of an automated tool for the preliminary sizing of planetary gearboxes used as final drives in agricultural machinery. In this field, the requirement to transmit high torques within extremely compact volumes, while simultaneously guaranteeing high standards of reliability and durability under severe operating conditions, makes the traditional design process complex and unsystematic. The nature of the problem is configured as a multi-objective and multi-constrained optimization, characterized by the presence of heterogeneous decision variables — discrete (normalized modules, pressure angles), integer (number of teeth), and continuous (face width) —and by conflicting objectives, such as the minimization of the overall volume and the maximization of surface pitting resistance. To overcome the limitations of conventional approaches, a multi-objective genetic algorithm based on the NSGA-II (Non-dominated Sorting Genetic Algorithm II) logic has been implemented in the MATLAB® environment. This meta-heuristic methodology allows for a global exploration of the wide space of admissible solutions, identifying a set of optimal configurations that define the Pareto Front of the problem. Such an approach provides the designer with a structured view of engineering trade-offs, allowing for a selection of the most suitable solution for the specific application. The operational core of the tool consists of a modular analytical code that integrates the principles of international ISO standards with specific corporate design guidelines. The gear geometry is described through a mathematical model that rigorously considers the effects of profile shift coefficients on the operating kinematics. Research robustness is guaranteed by a filtering system based on 11 critical constraints, organized into three macro-categories: geometric congruence and assembly (coaxiality, assembly condition, and clearance between planets), profile integrity (contact ratio, minimum tip thickness, and no-interference), and mechanical resistance (Hertzian pressure and tooth root stress calculated via the Lewis approximation optimized with interpolation functions). The tool development was completed by the realization of an advanced graphical user interface (GUI), conceived as a control and diagnostic center for decision support. The GUI allows for the monitoring of dominance rank evolution, analysis of population diversity, and the automatic identification of the optimal front compromise through the calculation of the minimum Euclidean distance from the origin. The effectiveness of the tool was validated by analyzing representative case studies. Thanks to automated export functionalities, the optimized geometric data were integrated into the finite element analysis (FEA) workflow using the Transmission3D software, confirming the consistency of the results obtained and the validity of the entire methodological work. Finally, in the concluding chapter of the discussion, the main results obtained from the analyzed case studies and the possible future developments of the project are discussed.
Il presente lavoro di tesi, svolto in collaborazione con una realtà industriale, si focalizza sullo sviluppo di uno strumento automatizzato per il dimensionamento preliminare di riduttori epicicloidali impiegati come final drive in macchine agricole. In questo ambito, la necessità di trasmettere elevate coppie in volumi estremamente contenuti, garantendo al contempo elevati standard di affidabilità e resistenza in condizioni operative gravose, rende il processo di design tradizionale complesso e poco sistematico. La natura del problema si configura come un’ottimizzazione multi-obiettivo e multi-vincolo, caratterizzata dalla presenza di variabili decisionali di natura eterogenea — discrete (moduli normalizzati, angoli di pressione), intere (numero di denti) e continue (larghezza di fascia) — e da obiettivi in forte contrasto tra loro, quali la minimizzazione dell’ingombro volumetrico e la massimizzazione della resistenza superficiale a pitting. Per superare i limiti degli approcci convenzionali, è stato implementato in ambiente MATLAB® un algoritmo genetico multi-obiettivo basato sulla logica NSGA-II (Non-dominated Sorting Genetic Algorithm II). Questa metodologia meta-euristica permette di esplorare in modo globale l’ampio spazio delle soluzioni ammissibili, individuando un insieme di configurazioni ottime che definiscono il Fronte di Pareto del problema. Tale approccio fornisce al progettista una visione strutturata dei trade-off ingegneristici, consentendo una selezione consapevole della soluzione di compromesso più idonea alla specifica applicazione. Il cuore operativo dello strumento è costituito da un motore analitico modulare che integra i principi della normativa internazionale ISO con specifiche linee guida progettuali aziendali. La geometria degli ingranaggi è descritta attraverso un modello matematico che considera rigorosamente gli effetti dei coefficienti di spostamento sulla cinematica di funzionamento. La robustezza della ricerca è garantita da un sistema di filtraggio basato su 11 vincoli critici, organizzati in tre macro-categorie: congruenza geometrica e montaggio (coassialità, condizione di divisibilità, clearance tra satelliti), integrità del profilo (grado di ricoprimento, spessore di testa minimo, assenza di interferenze) e resistenza meccanica (pressione hertziana e stress al piede dente calcolato tramite l’approssimazione di Lewis ottimizzata con funzioni di interpolazione). Lo sviluppo è stato completato dalla realizzazione di un’interfaccia grafica (GUI) avanzata, concepita come centro di controllo e diagnostica per il supporto decisionale. La GUI permette di monitorare l'evoluzione dei rank di dominanza, analizzare la diversità della popolazione e identificare automaticamente il compromesso del fronte ottimale mediante il calcolo della distanza Euclidea minima dall'origine. L'efficacia dello strumento è stata validata analizzando casi studio rappresentativi. Grazie a funzionalità di esportazione automatizzata, i dati geometrici ottimizzati sono stati integrati nel flusso di calcolo agli elementi finiti (FEA) tramite il software Transmission3D, confermando la coerenza dei risultati ottenuti e la validità dell'intero workflow metodologico. Infine, nel capitolo conclusivo della trattazione, vengono discussi i principali risultati ottenuti dai casi studio analizzati e i possibili sviluppi futuri del progetto.
Sviluppo di uno strumento di ottimizzazione multi-obiettivo automatizzato basato sull'algoritmo genetico NSGA-II per il dimensionamento preliminare di riduttori epicicloidali
NICOSIA, CHIARA
2024/2025
Abstract
This thesis work, conducted in collaboration with an industrial company, focuses on the development of an automated tool for the preliminary sizing of planetary gearboxes used as final drives in agricultural machinery. In this field, the requirement to transmit high torques within extremely compact volumes, while simultaneously guaranteeing high standards of reliability and durability under severe operating conditions, makes the traditional design process complex and unsystematic. The nature of the problem is configured as a multi-objective and multi-constrained optimization, characterized by the presence of heterogeneous decision variables — discrete (normalized modules, pressure angles), integer (number of teeth), and continuous (face width) —and by conflicting objectives, such as the minimization of the overall volume and the maximization of surface pitting resistance. To overcome the limitations of conventional approaches, a multi-objective genetic algorithm based on the NSGA-II (Non-dominated Sorting Genetic Algorithm II) logic has been implemented in the MATLAB® environment. This meta-heuristic methodology allows for a global exploration of the wide space of admissible solutions, identifying a set of optimal configurations that define the Pareto Front of the problem. Such an approach provides the designer with a structured view of engineering trade-offs, allowing for a selection of the most suitable solution for the specific application. The operational core of the tool consists of a modular analytical code that integrates the principles of international ISO standards with specific corporate design guidelines. The gear geometry is described through a mathematical model that rigorously considers the effects of profile shift coefficients on the operating kinematics. Research robustness is guaranteed by a filtering system based on 11 critical constraints, organized into three macro-categories: geometric congruence and assembly (coaxiality, assembly condition, and clearance between planets), profile integrity (contact ratio, minimum tip thickness, and no-interference), and mechanical resistance (Hertzian pressure and tooth root stress calculated via the Lewis approximation optimized with interpolation functions). The tool development was completed by the realization of an advanced graphical user interface (GUI), conceived as a control and diagnostic center for decision support. The GUI allows for the monitoring of dominance rank evolution, analysis of population diversity, and the automatic identification of the optimal front compromise through the calculation of the minimum Euclidean distance from the origin. The effectiveness of the tool was validated by analyzing representative case studies. Thanks to automated export functionalities, the optimized geometric data were integrated into the finite element analysis (FEA) workflow using the Transmission3D software, confirming the consistency of the results obtained and the validity of the entire methodological work. Finally, in the concluding chapter of the discussion, the main results obtained from the analyzed case studies and the possible future developments of the project are discussed.| File | Dimensione | Formato | |
|---|---|---|---|
|
Nicosia.Chiara.pdf
embargo fino al 07/04/2029
Dimensione
13.05 MB
Formato
Adobe PDF
|
13.05 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14251/5259