AlloyLens: A Visual Analytics Tool for High-throughput Alloy Screening and Inverse Design

TL;DR

AlloyLens tackles the challenge of inverse design in high-dimensional alloy spaces by integrating a coordinated SPLOM with dynamic sliders, gradient-based sensitivity analysis, and nearest-neighbor recommendations within a Jupyter-based visual analytics workflow. The approach uses a lightweight MLP surrogate to model nonlinear composition–property relationships, enabling real-time predictions, partial derivatives for local sensitivities, and immediate exploration of tradeoffs. Key contributions include (1) a tightly integrated SPLOM-surrogate interface, (2) gradient-based local sensitivity analysis for distortion-free exploration, and (3) domain expert-driven evaluation across structural, thermal, and electrical alloy design tasks. The results demonstrate effective hierarchical exploration, near-miss discovery via soft-margin brushing, and actionable candidate materials for downstream validation, illustrating practical impact on hypothesis-driven materials design in automotive, aerospace, and heat-exchanger contexts.

Abstract

Designing multi-functional alloys requires exploring high-dimensional composition-structure-property spaces, yet current tools are limited to low-dimensional projections and offer limited support for sensitivity or multi-objective tradeoff reasoning. We introduce AlloyLens, an interactive visual analytics system combining a coordinated scatterplot matrix (SPLOM), dynamic parameter sliders, gradient-based sensitivity curves, and nearest neighbor recommendations. This integrated approach reveals latent structure in simulation data, exposes the local impact of compositional changes, and highlights tradeoffs when exact matches are absent. We validate the system through case studies co-developed with domain experts spanning structural, thermal, and electrical alloy design.

Figures (9)

• Figure 1: Workflow overview. (1) The user begins with an overview of all candidate materials, shown as chemical compositions mapped to various material properties. (2) They interactively explore the design space by: (2a) applying constraints to filter for desired property ranges; (2b) balancing tradeoffs when constraints conflict; and (2c) analyzing local sensitivity to decide whether further simulation or experimentation is needed. Finally, users can (3) export selected candidates for (4) further analysis or experimental validation.
• Figure 2: SPLOM overview for hierarchical exploration (R1). The interface starts with a scatterplot matrix showing pairwise relationships between elements and properties, allowing users to identify trends and focus areas for further filtering.
• Figure 3: Interactive feature selection. Users define target properties via checkboxes, enabling task-specific filtering and real-time exploration of alloy candidates.
• Figure 4: Gradient-based sensitivity analysis (R3, R4). Sensitivity curves derived from an MLP surrogate reveal nonlinear input-output relationships and local sensitivity as users adjust sliders, enabling local tradeoff analysis.
• Figure 5: Brush with soft-match tolerance margin. Blue points satisfy all active slider bounds; purple points fall within $\pm 5 \%$ of specified bounds.