AIXD - AI eXtended Design
AIXD is an open-source Python toolbox that integrates AI into computational design.
Try out AIXD
AIXD simplifies handling the development of machine learning models for inverse design, surrogate modeling, and sensitivity analysis, enabling domain experts to rapidly explore diverse solutions with minimal coding.
Practical Application of AIXD in Design Scenarios
These are example projects which illustrate how to effectively integrate the machine learning models into your parametric design projects.
Inverse Design for a Vertical Garden
AI can augment structural design by providing a diverse set of high-performing alternatives tailored to project constraints and design priorities. Instead of beginning from scratch or relying solely on heuristics, design tables, or past project experiences, the framework equips designers with viable starting points that respect project conditions and reveal trade-offs across multiple competing objectives. By embedding quantitative evaluation into the earliest design decisions, the approach has the potential to reduce rework, accelerate iteration cycles, and support more efficient outcomes.
Publication: external page Augmented Intelligence for Architectural Design with Conditional Autoencoders: Semiramis Case Study
Inverse Design for a Pedestrian Bridge & Sensitivity Analysis
Assessment of Aging Infrastructure - Surrogate Model for Existing Concrete Frame Bridges
Test the toolbox on your parametric design problem
AIXD is an open-source toolkit for data-driven forward and inverse design and is tailored to design-related domains like architecture, civil engineering, product design and mechanical engineering. With few high-level commands, designers can harness their parametric models to generate project-specific datasets, or import data generated externally. This data is used then to train custom machine learning models, and finally to deploy them for the desired downstream tasks like design exploration, sensitivity analysis or data visualization. It follows a simple few-steps workflow.
Through AIXD, we intend to make the power of AI methodologies more accessible to domain experts, and allow them to integrate these approaches in their design pipelines. Then, the implemented features, such as accelerated forward and inverse design, can be used in a human-in-the-loop fashion, enabling the designer to carry out a faster and more exhaustive exploration of the problem, and reach a better understanding of the design problem.