Accurate characterization of material properties is critical for modeling and optimizing advanced systems, yet conventional experimental and simulation-based approaches remain costly and data-intensive. As artificial intelligence evolves from data-driven modeling to physics-informed and knowledge-guided paradigms, this perspective article highlights the role of physics-informed machine learning (PIML), specifically physics-informed neural networks (PINNs), as a key enabler of data-efficient, physically consistent inference. PINNs embed governing equations into the learning process and have demonstrated strong capability in recovering constitutive and transport parameters from sparse or noisy data while preserving physical fidelity. This paper examines the fundamental structure, workflow integration, and recent advances of PINNs in the context of inverse material characterization. It also discusses open challenges in computational cost, training stability, and uncertainty quantification. Looking forward, integration with digital twins, generative modeling, and autonomous experimentation presents a pathway toward interpretable, adaptive, and automated characterization for next-generation intelligent manufacturing.

The widespread adoption of bulk metallic glasses (BMGs) in aerospace and biomedical industries requires topology-optimized architectures that conventional manufacturing cannot achieve. In response, BMGs have been investigated for powder bed fusion (PBF), but the process remains challenging due to narrow thermal windows, expensive feedstock, and limited data. This study introduces a constrained multi-objective Bayesian optimization framework to optimize key PBF printing parameters, including laser power and scan speed, to maximize hardness while preserving the amorphous state of the printed BMG. Hardness is optimized as the primary objective with density incorporated in the scalarization to regularize the search space, and amorphous retention is enforced through a feasibility probability learned by a logistic classifier. Surrogate models are compared, including Gaussian process, Bayesian additive regression trees, Bayesian multivariate adaptive regression splines (BMARS), and a Bayesian attention neural network. Acquisition scores are computed with constrained expected improvement and are maximized on a uniform grid over power and velocity. Superior predictive accuracy is obtained with BMARS, and 95% credible intervals are calibrated to the measurements. A high-hardness region at high power and low velocity is localized by the surrogates. A fully amorphous sample at 60 W and 1300 mm/s is produced, and a hardness of 1010.4 HV is measured in agreement with the predicted high-hardness band. In conclusion, the study establishes a data-efficient process-window discovery method for BMG PBF, produces an interpretable process map, and demonstrates a screening framework suitable for constrained experimental budgets.

With the development of inkjet-printed electrodes, artificial intelligence-based quality control is essential for classifying inkjet-printed electrodes in a quality control environment. The quality of printed structures can be significantly affected by defects such as cracks, smudging, and misaligned deposits, which can degrade electrical performance and overall device reliability. Traditional quality control methods, including manual inspection and electrical testing, are time-consuming, subjective, and invasive, and they are unsuitable for high-throughput manufacturing environments. This work explores the application of computer vision and deep learning, specifically Convolutional Neural Networks (CNNs) and Feedforward Neural Networks, to automate defect detection and quality classification of inkjet-printed electrodes. To demonstrate the accessibility of deep learning techniques, Neural Architecture Search was implemented, showing the importance of automated model design in achieving high performance without extensive manual tuning or the need for expertise. The CNN models proved to be the most suitable approach for this image classification task, achieving a testing accuracy of 90.9% and a precision of 88.9% for a dataset of 2,406 electrode images containing both high-quality (1,020) and low-quality (1,386) prints.

Friction stir spot welding (FSSW) has gained increasing attention over the last decade due to its promising performance compared to conventional joining methods for similar metals. However, the thermal and tensile responses in this process are highly nonlinear. This study aims to explore the thermal and tensile performance of aluminum joints welded by FSSW using an innovative method based on exploratory data analysis (EDA), followed by several machine learning (ML) approaches. The welding parameters investigated in this study were tool rotational speed, dwelling time, and aluminum sheet thickness. The ML methods included linear and nonlinear regression models for welded joints at different welding parameters. We evaluated Bayesian ridge, elastic-net, support vector regression (SVR), random forest, polynomial regression (nonlinear), and robust regression. The random forest algorithm provided accurate predictions for lap-shear fracture load (R² = 0.96, mean squared error [MSE] = 0.01, and mean absolute error [MAE] = 0.07) in tensile performance, whereas the elastic net performed worst. Model-to-model differences were smaller for thermal performance, with the random forest model yielding the most accurate predictions (R² = 0.97, MSE = 26.51, and MAE = 3.86) while the SVR yielded the least accurate predictions. The study indicated that using EDA to address anomalies in welding conditions provides valuable insights into the best ML methods for predicting the thermal and mechanical performance of welding joints.

This study explores artificial intelligence (AI)-mediated participatory design integrating biomaterials and digital fabrication to co-create speculative artifacts grounded in lived experiences. The present study involves experimentation with biomaterials, exploring the intersection of image-based generative AI, participatory, and co-creative methodologies within a design framework that reimagines lived experiences shaped by identity-based exclusionary processes. Rather than pursuing AI-driven discovery of new materials, this study positions design as a mediating process among human experience, critical reflection, biomaterial exploration, and digital fabrication. The research introduces a three-stage workflow (co-creation, fabrication, and materialization) that employs AI as a mediating tool between subjective narratives and tangible speculative artifacts. During the co-creation stage, participants shared their personal experiences through open-ended surveys, text-to-image generative AI visualization, and algorithmic three-dimensional (3D) modeling. This process enabled participants to speculatively reimagine lived experiences of social exclusion, demonstrating how AI can support new modes of participatory and social engagement. During the fabrication stage, digital models were translated into physical counter-molds through 3D printing and subsequently cast in silicon, reaffirming the reciprocal relationship between digital and craft-based production. The materialization stage explored biomaterial compositions informed by participants’ narratives and materialities, incorporating hair, wood ash, and plastic waste into biomaterial compositions grounded in circular economy principles. The resulting artifacts function as speculative objects that incite interpretation beyond fixed symbolic representation. This study contributes to ongoing discussions in digital fabrication, material design, and critical craft by demonstrating how AI-mediated participatory co-creation can foster ethically conscious, socially engaged, and materially grounded design practices. Future work may extend this approach to larger collective settings and further explore the integration of biomaterials and AI within ecological and inclusive design frameworks.