Active-learning-assisted optimization and HIP induced interfacial strengthening of atmospheric plasma-sprayed ta coatings on tungsten

Atmospheric plasma spraying (APS) is an effective technique to fabricate tantalum (Ta) protective coatings on tungsten substrates in harsh nuclear environments. However, their application is impeded by the inherent lamellar microstructure, inter-splat pores, and weakly bonded Ta/W interface. Herein, a data-efficient optimization strategy integrating sequential orthogonal experiments, Gaussian process regression (GPR)-based active learning, and Pareto screening was developed to optimize the APS process parameters for simultaneously reducing coating porosity and improving tensile bonding strength. Using only 24 experimental samples, the optimized GPR models achieved predictive coefficients (Q 2 ) of 0.889 and 0.894 for porosity and bonding strength, respectively. A balanced Pareto-optimal process condition (43 kW plasma power, 48 L·min −1 Ar flow rate, 120 mm spray distance, and 22.5 g·min −1 powder feed rate) was experimentally validated, yielding a coating porosity of 3.487 ± 0.185% and a tensile bonding strength of 19.353 ± 0.379 MPa, with prediction errors below 3%. Compared with the optimum result obtained from the initial orthogonal design, the optimized coating exhibited a 27.4% reduction in porosity and a 28.2% increase in bonding strength. Subsequent capsule-free hot isostatic pressing (HIP) at 1400 °C/200 MPa for 3 h further reduced the porosity to 0.472 ± 0.107% and increased the bonding strength to 31.874 ± 4.985 MPa through pore closure, interlamellar consolidation, and enhanced coating/substrate bonding. Microstructural characterization revealed that HIP promoted significant Ta/W interdiffusion, broadening the elemental transition zone from approximately 30 nm to 200 nm, while transforming the initially narrow amorphous-dominant interface into a graded amorphous–nanocrystalline transition region. Consequently, the fracture path shifted from preferential propagation along weak interlamellar and interfacial regions to partial failure within the W substrate, indicating substantially improved interfacial load transfer. This work demonstrates that integrating data-driven process optimization with HIP-induced interface engineering provides an efficient strategy for producing high-integrity APS refractory metal coatings with enhanced mechanical reliability.