Optimization of nickel-based superalloys by an efficient hot consolidation method and machine learning

Nickel-based superalloys are widely used in aerospace and energy industries owing to their superior high-temperature strength and creep resistance. This study proposes an optimized hot consolidation approach for these alloys, integrating powder metallurgy with machine learning to enhance mechanical properties. The research systematically investigates the effects of ball milling time, hot consolidation temperature, strain rate, and alloy composition on hardness, providing a data-driven approach to process optimization. Experiments were conducted using mechanically alloyed nickel powders, which were consolidated under varying thermal conditions and assessed for mechanical properties. The results indicate that ball milling significantly enhances hardness, while hot consolidation temperature and strain rate are key factors governing final material properties. Machine learning analysis further identifies the critical processing parameters affecting hardness, enabling precise control over material performance. By leveraging data-driven models, the study improves predictive accuracy, reducing the reliance on traditional trial-and-error experimentation, while enhancing efficiency in alloy design.

By combining machine learning models with powder metallurgy techniques, this study overcomes traditional barriers in alloy fabrication, enabling precise prediction and optimization of mechanical properties, particularly hardness.

Workflow overview:

Preparation (panel a): Nickel-based alloy powders are prepared via ball milling and consolidated through a hot consolidation process. This step ensures the samples are uniformly prepared for further testing and modeling. Experimentation (panel b): Consolidated samples are encased in stainless steel casings and subjected to Vickers hardness testing to evaluate mechanical properties. This experimental data forms the foundation for training machine learning models. Prediction (panel c): Machine learning models (multilayer perceptron [MLP], gradient boosting decision tree [GBDT], support vector regression [SVR], and ridge regression [RR]) predict hardness on the basis of key input parameters: alloy type, ball milling time, consolidation temperature, and strain rate. These predictions guide the selection of optimal processing conditions. Feedback and optimization (panel d): Experimental data validates machine learning predictions, refining the process parameters iteratively. Optimized conditions—identified as 24 h of ball milling, 1050 °C consolidation temperature, and a strain rate of 5 s⁻¹—maximize alloy hardness and improve the overall workflow.