Artificial Intelligence Optimisation of Laser Sintering of Ceramics: Towards a New Era of Additive Manufacturing

Cette offre de thèse est proposée dans le cadre du cluster IA ENACT et de ses partenaires.

Informations complémentaires : https://cluster-ia-enact.ai/

Context : 

The use of in-situ resource utilization (ISRU) is a key concept in lunar or Martian construction intended for long-duration human exploration. Additive manufacturing (3D printing), which uses lunar soil as the primary resource, can provide human shelters or tools for various applications. The main agreed-upon technique consists of building a rigid shell from sintered regolith in order to protect against micrometeorites, temperature gradients, and cosmic radiation. Once the shelter is built, it can accommodate a deployable or inflatable structure. These structures are lighter and take up less space, making them easier and less costly to transport aboard current launch vehicles. Selective laser melting additive manufacturing is a rapidly emerging technique that makes it possible to create ceramic parts with complex geometries capable of withstanding extreme environments. However, many technical challenges remain, particularly the major issue of controlling dimensional shrinkage and porosity during the critical sintering stage. Solving this problem is essential to ensure the mechanical reliability required by advanced industries such as aerospace, aeronautics, medical, and electronics. Led by the Instrumentation and Photonic Processes team at the ICube Laboratory, this research topic is aligned with the current strategy of integrating AI tools into photonic processes. This development is based on exploring the use of AI for predicting machining parameters, monitoring and real-time process correction, as well as beam shaping. This project will also strengthen our collaboration with the Institut Clément Ader by leveraging their expertise in the thermal, optical, and mechanical characterization of ceramics. In addition, this project will foster closer ties with the Polytechnic University of Milan, working with Barbara Previtali, an expert in laser processing and additive manufacturing, on building the database necessary for the success of this project.

Abstract

Selective Laser Sintering (SLS) enables the direct manufacturing of complex parts without material waste, which is common in conventional manufacturing methods. This PhD project explores the integration of artificial intelligence (AI) into the laser sintering process of ceramics such as alumina, silicon carbide, calcium phosphate, zirconia, and in this study, lunar dust simulants — regolith.

The objective is to optimize powder bed fusion additive manufacturing parameters using selective laser melting/sintering in order to improve the quality and reproducibility of produced parts, reduce production time, and accelerate the search for new processing parameters. Indeed, the composition of regolith — mainly composed of silica, alumina, iron oxide, and titanium oxide — varies depending on location. This variability is well represented by the wide range of simulants developed by space agencies and academic laboratories.

The Instrumentation and Photonic Processes (IPP) research team at the ICube laboratory has developed a powder bed SLS prototype using a 100 W CW 1090 nm laser and a galvanometric scan head for high speed and precision. The height and compression of added layers are controlled by a Z-stage, a roller, and a blade. Following a joint study conducted by INSERM and CNRS, and thanks to the work of recent PhD students and interns, this machine complies with the latest standards regarding protection against nanoparticle exposure.

The PhD will begin with the consolidation of a database of previously tested sintering parameters on regolith simulants. This database will be enriched with results from collaborators at the Polytechnic University of Milan and the Institut Clément Ader. A thorough literature review on laser sintering of ceramics — including density variations, cracking, and deformation — will further expand the database with new materials.

The doctoral candidate will then manufacture parts from different lunar dust simulants to feed the database. Sample characterization will mainly rely on compression testing and ceramography. These tests will be conducted in collaboration with the Institut Clément Ader, which works closely with IRAP (Institute for Research in Astrophysics and Planetology) in Toulouse on the development of a new French lunar simulant, BPY (derived from volcanic rock from the Pic d’Ysson in Auvergne).

In parallel, AI algorithms developed in previous studies on laser processes will be adapted and applied to predict and adjust optimal parameters such as laser power, scanning speed, and preheating temperature. Predictive algorithms, such as the promising new foundation model TabPFN, will be used to optimize and generate new samples.

The software developed during this PhD allows an operator to input the desired weld dimensions and material. Using a hybrid approach combining multiphysics modeling and machine learning, it outputs the parameters that produce a weld closest to the desired specifications. The operator can use these parameters as a starting point, significantly reducing optimization time.

Following earlier research, the software has been adapted for sintering lunar dust simulants and is available in open access. This test version allows the operator to either select an existing simulant or create a new one by specifying its composition.

This PhD aims to further improve the software by modeling the complex interdependencies between machine parameters and part quality using graph-based approaches (knowledge graphs or causal graphs). This will move beyond correlation-based machine learning toward decision guidance based on coherent and robust configurations.

One of the previously identified scientific challenges, for which the research team has begun multiphysics modeling, is the balling effect.

This work aligns with the laboratory’s strategy of integrating AI tools into photonic processes. The doctoral candidate will work closely with another PhD student developing real-time process monitoring with AI-driven feedback loops.

This research will go beyond parameter optimization, paving the way for automation and production personalization in line with Industry 4.0 requirements. It will also explore the use of AI for designing new ceramic materials and integrating advanced functionalities such as thermal or electrical conductivity.

In conclusion, this PhD lays the groundwork for a new era in ceramic additive manufacturing, where AI plays a central role in innovation and production efficiency.

The doctoral candidate will work across two sites: Icam Strasbourg-Europe (Schiltigheim) and the ICube laboratory (Illkirch). They will maintain regular collaboration with the Institut Clément Ader for sample characterization and with the Polytechnic University of Milan.

ICube Laboratory

The Engineering science, computer science and imaging laboratory

Created in 2013, the laboratory brings together researchers of the University of Strasbourg, the CNRS (French National Center for Scientific Research), the ENGEES and the INSA of Strasbourg in the fields of engineering science and computer science, with imaging as the unifying theme.

With around 650 members, ICube is a major driving force for research in Strasbourg whose main areas of application are biomedical engineering and the sustainable development.

The thesis topic is highly interdisciplinary. To succeed, the ideal candidate must possess a hybrid profile, capable of bridging the gap between Materials Science and Data Science.

1. Academic Background

The candidate must hold a Master’s degree (M2) or an engineering degree in one of the following fields:

Materials Science / Mechanics: With a strong aptitude for digital technology and programming.

Physics / Photonics: Specialization in laser-matter interaction or laser processes.

Data Science / Applied Computer Science: Provided there is a marked interest in experimental physics and manufacturing processes.

2. Technical Skills

A. Artificial Intelligence and Computer Science

As the core of the topic involves AI-based optimization, the candidate must master:

Machine Learning: Knowledge of prediction and optimization algorithms, including tabular foundation models like TabPFN.

Graph Theory: A key skill will be modeling parameter/quality interdependencies in the form of Knowledge Graphs or Causal Graphs.

Software Development: Ability to develop software tools for the operator interface and model integration.

B. Materials Science and Laser Processes

Since the candidate will be handling machines and powders, they must understand:

Additive Manufacturing: Understanding of the Selective Laser Sintering/Melting (SLS) process on a powder bed.

Ceramics Physics: Knowledge of materials such as alumina, silicon carbide, zirconia, and regolith simulants.

Characterization: Ability to conduct and analyze compression, density, and ceramography tests.

3. Experimental Skills

This is not a purely theoretical or numerical thesis. The candidate will be required to:

Operate an industrial prototype: Use an SLS machine equipped with a 100W laser.

Manage safety: Work with nanoparticles while adhering to strict, established safety standards.

Produce samples: Physically manufacture ceramic parts to populate the database.

4. Soft Skills & Personal Qualities

Collaboration Skills: The PhD student will work across two sites (Schiltigheim and Illkirch) and must interact regularly with:

The Institut Clément Ader for characterization.

The Polytechnic University of Milan for the database.

Another PhD student working on real-time monitoring at the ICube laboratory.

Fluent English: Essential for scientific literature and communication with Italian partners.

Scientific Curiosity: Interest in Industry 4.0, aerospace (lunar/regolith applications), and automation.

In summary: The ideal candidate is a materials/mechanical engineer who codes proficiently in Python, or a Data Scientist looking to work with physical materials, ready to "get their hands dirty" manufacturing ceramic parts while developing cutting-edge AI models.