Periodic Reporting for period 1 - CoBRAIN (Integrated Computational-Experimental material Engineering of Thermal Spray coatings)
Période du rapport: 2023-01-01 au 2024-06-30
Protective coatings against wear and corrosion play a critical role in many industrial fields. However, extant technologies have considerable sustainability drawbacks. A prime example is electroplated Cr, a safe material whose deposition however involves highly toxic and carcinogenic Cr6+ compounds included in the REACH list of Substances of Very High Concern (Annex XIV). Thermal spray processes are intrinsically “cleaner”: they do not employ baths with large amounts of chemicals. However, they also use materials like WC-Co-based hardmetals and Stellite alloys that contain carcinogenic elements (Co) and/or critical raw materials (CRMs) – Co, W.
Therefore, more sustainable coating formulations are needed, while still exploiting the versatility and low environmental impact of thermal spray. However, almost one century of efforts in this respect have been met with scarce success. Therefore, CoBRAIN aims to harness the as-yet largely unexplored compositional space of High-Entropy and Multi-Principal Component Alloys (HEAs, MPEAs) to devise novel compositions. The HEA/MPEA concept offers unique features, precluded to conventional alloys based on only one main element. Further, the sheer vastity of the HEA/MPEA compositional space offers more latitude to optimise the coating properties.
The CoBRAIN project specifically seeks to develop HEAs/MPEAs with “simple” metallic elements like Cr, Mn, Fe, Ni, Cu, Al, Ti, Si, to minimize the amount of expensive and/or critical materials. Some of these elements are also considered critical (e.g. Mn) or strategic (e.g. Ni) by the EU, but their criticality is less significant than Co and W, and in a HEA none of them will ever be the sole main constituent, thus reducing the alloy's sensitivity to the criticality of one constituent. Hardmetals are formulated by adding TiC, a hard phase obtained from much more widely available elements than WC.
The number of possible combinations in the HEA/MPEA space is extremely wide and cannot be probed by experiments alone; even physical models would not suffice. Therefore, CoBRAIN uses high-throughput physical modelling to create databases onto which machine learning (ML) is trained (Figure 1: WP2), with validation from experimental testing (Figure 1: WP3). In particular, after creating a first iteration of the ML model, its “weak areas” are identified and explored systematically with further experimental activities (Figure 1: WP4). Integration of data(and metadata) from multiple sources is enabled by creating a semantic representation through an overarching ontology of thermal spraying (Figure 1: WP1).
The final ML model is integrated, together with a life cycle performance analysis (LCPA) module, into a sustainable decision support system (SDSS) - Figure 1: WP5. The SDSS system and its outputs will be tested in real case-studies (Figure 1: WP6).
Therefore, more sustainable coating formulations are needed, while still exploiting the versatility and low environmental impact of thermal spray. However, almost one century of efforts in this respect have been met with scarce success. Therefore, CoBRAIN aims to harness the as-yet largely unexplored compositional space of High-Entropy and Multi-Principal Component Alloys (HEAs, MPEAs) to devise novel compositions. The HEA/MPEA concept offers unique features, precluded to conventional alloys based on only one main element. Further, the sheer vastity of the HEA/MPEA compositional space offers more latitude to optimise the coating properties.
The CoBRAIN project specifically seeks to develop HEAs/MPEAs with “simple” metallic elements like Cr, Mn, Fe, Ni, Cu, Al, Ti, Si, to minimize the amount of expensive and/or critical materials. Some of these elements are also considered critical (e.g. Mn) or strategic (e.g. Ni) by the EU, but their criticality is less significant than Co and W, and in a HEA none of them will ever be the sole main constituent, thus reducing the alloy's sensitivity to the criticality of one constituent. Hardmetals are formulated by adding TiC, a hard phase obtained from much more widely available elements than WC.
The number of possible combinations in the HEA/MPEA space is extremely wide and cannot be probed by experiments alone; even physical models would not suffice. Therefore, CoBRAIN uses high-throughput physical modelling to create databases onto which machine learning (ML) is trained (Figure 1: WP2), with validation from experimental testing (Figure 1: WP3). In particular, after creating a first iteration of the ML model, its “weak areas” are identified and explored systematically with further experimental activities (Figure 1: WP4). Integration of data(and metadata) from multiple sources is enabled by creating a semantic representation through an overarching ontology of thermal spraying (Figure 1: WP1).
The final ML model is integrated, together with a life cycle performance analysis (LCPA) module, into a sustainable decision support system (SDSS) - Figure 1: WP5. The SDSS system and its outputs will be tested in real case-studies (Figure 1: WP6).
The status of the project’s activities and achievements is as follows:
- Five sets of HEAs and five sets of HEA+TiC hardmetals were identified from the literature. A high-throughput procedure for powder production by HEBM was set up (Figure 2), which, coupled with sieving and classification procedures, provides size fractions suitable for various thermal spray processes (HVOF, HVAF, CGS) simultaneously from a single milling batch. Using these powders, high-quality HVOF, HVAF and CGS samples were produced (Figure 2), also in a repeatable, high-throughput way.
- After creating characterization data (CHADA) workflows for each method of interest (WP1, deliverable D1.3) a complete workflow for high-throughput experimental characterization is being established in WP3 (Figure 3). All deposited samples are subjected to a basic set of tests to establish their quality and the main microstructural and micromechanical features; selected samples are further investigated for functional and more in-depth micromechanical assessments. The eventual workflow will be detailed in deliverable D3.3.
- In parallel, the modelling data (MODA) structures and modelling workflows were established (WP1, deliverable D1.2). Physical and process models run in WP2 are thus providing data to train the first iteration of the machine learning (ML) architecture. These models include: CalPhaD calculations of phase composition and physical properties (Figure 4); Density Functional Theory (DFT) calculations of stacking-fault energies (SFEs) and matrix/hard phase wetting; phase-field simulation of the the microstructural and phase evolution during solidification (Figure 5); crystal plasticity models to predict micro-mechanical responses; macro- and meso-scale FE models to simulate the large-scale stress, strain and temperature distributions under functional testing conditions; computational fluid dynamics (CFD) modelling of the thermal spray processes. Each of these models is verified against experiments.
- In WP1, all the concepts related to thermal spraying and their mereocausal relationships were defined. The project’s data and metadata are being mapped into the resulting knowledge graph, implemented using the OWL2 language in a free triplestore (OntoText GraphDB), to create the complete knowledge base (KB) of the project.
- The architecture of the SDSS was established (WP5, Figure 1), with its browser-based user interface; links to the ML and parametric LCPA modules; and the visualization module to provide the decision support.
- Five sets of HEAs and five sets of HEA+TiC hardmetals were identified from the literature. A high-throughput procedure for powder production by HEBM was set up (Figure 2), which, coupled with sieving and classification procedures, provides size fractions suitable for various thermal spray processes (HVOF, HVAF, CGS) simultaneously from a single milling batch. Using these powders, high-quality HVOF, HVAF and CGS samples were produced (Figure 2), also in a repeatable, high-throughput way.
- After creating characterization data (CHADA) workflows for each method of interest (WP1, deliverable D1.3) a complete workflow for high-throughput experimental characterization is being established in WP3 (Figure 3). All deposited samples are subjected to a basic set of tests to establish their quality and the main microstructural and micromechanical features; selected samples are further investigated for functional and more in-depth micromechanical assessments. The eventual workflow will be detailed in deliverable D3.3.
- In parallel, the modelling data (MODA) structures and modelling workflows were established (WP1, deliverable D1.2). Physical and process models run in WP2 are thus providing data to train the first iteration of the machine learning (ML) architecture. These models include: CalPhaD calculations of phase composition and physical properties (Figure 4); Density Functional Theory (DFT) calculations of stacking-fault energies (SFEs) and matrix/hard phase wetting; phase-field simulation of the the microstructural and phase evolution during solidification (Figure 5); crystal plasticity models to predict micro-mechanical responses; macro- and meso-scale FE models to simulate the large-scale stress, strain and temperature distributions under functional testing conditions; computational fluid dynamics (CFD) modelling of the thermal spray processes. Each of these models is verified against experiments.
- In WP1, all the concepts related to thermal spraying and their mereocausal relationships were defined. The project’s data and metadata are being mapped into the resulting knowledge graph, implemented using the OWL2 language in a free triplestore (OntoText GraphDB), to create the complete knowledge base (KB) of the project.
- The architecture of the SDSS was established (WP5, Figure 1), with its browser-based user interface; links to the ML and parametric LCPA modules; and the visualization module to provide the decision support.
The most important results beyond the state-of-the-art achieved so far are:
- Production of HEA and HEA+TiC feedstock powders and coatings free of cobalt. In fact, most experiences with thermal spray deposition of HEAs up until now still employed Co-containing formulations, and there is no experience on thermal spraying of hardmetals with HEA matrix.
At least one of these TiC-based hardmetals with HEA matrix provided specific wear rates under high-stress abrasion conditions comparable to HVOF Cr3C2+25wt.%(Ni-20wt.%Cr), which the past attempts with HVOF-sprayed TiC-based hardmetals in non-HEA matrices never achieved.
- ML models trained onto databases of thermodynamic properties and SFEs of HEAs. There are other ML models for thermodynamic phase prediction of HEAs, but for the first time this is purposefully aimed at coatings. ML models trained on SFE values from DFT have not been reported up until now.
- Generation of the first ever ontology of thermal spraying for the systematic mapping of data and metadata.
- The ontology creation exercise was also instrumental for the development of the soon-to-be-released EMMO 1.0.0 update and the novel EMMO-LITE tool.
More specific provisions with respect to needs for market uptake can be provided only at a later stage, when completion of the development activities shall have provided a clearer picture of the full application potentials of the novel solutions. The market update prospects will also be affected by evolutions in the normative framework, e.g. decisions on the reauthorization of hexavalent chromium usage for electrodeposition, possible restrictions on cobalt, and future trends of the raw materials’ market.
- Production of HEA and HEA+TiC feedstock powders and coatings free of cobalt. In fact, most experiences with thermal spray deposition of HEAs up until now still employed Co-containing formulations, and there is no experience on thermal spraying of hardmetals with HEA matrix.
At least one of these TiC-based hardmetals with HEA matrix provided specific wear rates under high-stress abrasion conditions comparable to HVOF Cr3C2+25wt.%(Ni-20wt.%Cr), which the past attempts with HVOF-sprayed TiC-based hardmetals in non-HEA matrices never achieved.
- ML models trained onto databases of thermodynamic properties and SFEs of HEAs. There are other ML models for thermodynamic phase prediction of HEAs, but for the first time this is purposefully aimed at coatings. ML models trained on SFE values from DFT have not been reported up until now.
- Generation of the first ever ontology of thermal spraying for the systematic mapping of data and metadata.
- The ontology creation exercise was also instrumental for the development of the soon-to-be-released EMMO 1.0.0 update and the novel EMMO-LITE tool.
More specific provisions with respect to needs for market uptake can be provided only at a later stage, when completion of the development activities shall have provided a clearer picture of the full application potentials of the novel solutions. The market update prospects will also be affected by evolutions in the normative framework, e.g. decisions on the reauthorization of hexavalent chromium usage for electrodeposition, possible restrictions on cobalt, and future trends of the raw materials’ market.