Leveraging Human Experience

Human-Informed Topology Optimization

News

News from the Group

About Us

About

The carstensen group is an interdisciplinary engineering lab at MIT focusing on developing new rigorous methods and algorithms that can improve structural design.

WELCOME
The digitalization of manufacturing is radically changing how we fabricate and construct structures. Today we can make things that were too complex for manufacture just a few years ago. To fully leverage the new manufacturing possibilities and make our structures lighter, faster, safer, or any other objective we seek, we need to re-think our approach to structural design. In the carstensen group we work with developing structural- and topology optimization frameworks that improves design. We work with design at the conceptual stage, develop algorithmic details for implementation of new manufacturing considerations and take optimized designs out of the computational space to experimentally investigate their behavior. We are based at MIT in the Department of Civil and Environmental Engineering. The
structures we are interested in can have any size; it can be a high-rise building, a part or component of a larger structure or the architecture of a porous material.

Work

Featured Projects

Speeding up Design Time

When designing with topology optimization, the human user spends significant time manually selecting and tuning algorithmic hyperparameters to achieve an acceptable design. Drawing inspiration from hyperparameter optimization in machine learning, this paper presents a new approach that speeds up the design time by automating the tuning process. We show that for a simple design case, the new framework reduces the number of times a topology optimization algorithm is executed by 90% without sacrificing the structural performance.

Barriers and Opportunities to Embodied Carbon Reductions in Structural Engineering

Through a survey of structural engineers primarily in the northeast US, we identified the primary barriers to embodied carbon reduction, in particular barriers to the use of tools like parametric design and optimization, and opportunities to overcome these barriers. The formalization of these barriers is key to understanding what changes need to be made, both in industry and academia, to enable further reductions in the embodied carbon emissions of construction.

Human-Informed
Topology Optimization

This work presents a new topology optimization framework in which the design decisions are made by humans and machines in collaboration. In Human-Informed Topology Optimization (HiTop), we enable the design engineer to actively use their experience to locally alter the minimum feature size requirements. The user can identify potential members of concern and the algorithm responds by eliminating, merging, or thickening them. HiTop's ability to achieve performance improvement with few computational resources is demonstrated on buckling and stress concentration examples.

Waterjet Cut
Reinforced Concrete

This work presents an experimental investigation of the effect of ribbing patterns overlayed on waterjet cut steel plates when using them as reinforcement for concrete. A key takeaway is that we can get within 90% of the pull out force of regular rebar just by varying the ribbing pattern. This opens up opportunities for prototype and scale testing of new civil structural designs.

Timber-Steel Trusses with Minimized Embodied Carbon

There is an increasing need for automated design processes that can help guide structural design towards lower embodied carbon solutions. This research presents a two-material truss topology optimization algorithm that aims at reducing the Global Warming Potential (GWP) of the designed structure. The framework is demonstrated on truss designs with a mix of glue-laminated timber (GLT) and steel elements for both 2D and 3D design examples.

Cellular Materials with Maximized Energy Absorption

Topology optimization is increasingly being used to design the architecture of porous cellular materials with extreme elastic properties. This project seeks to extend the design problem to the nonlinear regime and aim to maximize the energy absorption. Designs are obtained, fabricated and tested with bulk metallic glass as the base solid.

Topology Optimization of Rigid Interlocking Assemblies

Traditionally, interlocking assemblies are used in carpentry and associated with a set of quintessential shapes that are governed by the traditional cutting tools for wood. With today’s manufacturing, both ease of fabrication and design options are dramatically changed to give this type of connection renewed relevance and development potential as it can lock components in place without using of adhesives or fasteners such as mortar, glue, bolts, nails or screws.

Topology optimized design is tailored to fabrication by material extrusion processes by incorporating a discrete nozzle size constraint. Material extrusion processes are used in a range of 3D printing technologies; from bio printing for tissue engineering, over Fused Filament Fabrication to Big Area Manufacturing and Concrete 3D printing.

Topology Optimization with Nozzle Size Restrictions for 3D Printing

Two Phase Minimum and Maximum Feature Size Control in Topology Optimization

Manufacturability of topology-optimized designs is improved by providing the design engineer with minimum length scale control of both the solid and void phases of a design. We also propose a design framework that enables maximum length scale control of a single phase.

Reinforced Concrete (RC) beams are designed using an existing hybrid bi-linear topology optimization algorithm for design of Strut-and-Tie models. The reinforcement layouts are water jet cut from steel plates and placed into the concrete during casting. The beams are experimentally tested to investigate the performance of the designs. Behavioral improvements are found even at low levels of complexity.

Topology-Optimized Reinforced Concrete Beams

Topology-Optimized Concrete

Concrete beams are designed using existing topology optimization algorithms and cast into falsework that is fabricated with digital tools. The beams are experimentally tested to investigate the performance of the design algorithms outside the computational domain.

Contact

Clients

Contact

Address

77 Massachusetts Avenue

MIT Room 1-153, Attn: Josephine Carstensen

Cambridge MA 02139

United States

Leveraging Human Experience

Human-Informed Topology Optimization

News from the Group

About Us

The carstensen group is an interdisciplinary engineering lab at MIT focusing on developing new rigorous methods and algorithms that can improve structural design.

Featured Projects

Speeding up Design Time

Barriers and Opportunities to Embodied Carbon Reductions in Structural Engineering

Automated Design with Reclaimed Partially Disassembled Truss Components

HiTop with User-Drawn Image Input

Multiple Feature Size Controls in Human-Informed Topology Optimization

Topology Optimization for Extrusion-Based AM

Improved Manufacturability for Restricted Design Scenarios

Human-Informed Topology Optimization

Waterjet Cut Reinforced Concrete

Timber-Steel Trusses with Minimized Embodied Carbon

Cellular Materials with Maximized Energy Absorption

Topology Optimization of Rigid Interlocking Assemblies

Topology Optimization with Nozzle Size Restrictions for 3D Printing

Two Phase Minimum and Maximum Feature Size Control in Topology Optimization

Topology-Optimized Reinforced Concrete Beams

Topology-Optimized Concrete

Contact

Address

Get in Touch

HiTop with User-Drawn
Image Input

Topology Optimization for
Extrusion-Based AM

Human-Informed
Topology Optimization

Waterjet Cut
Reinforced Concrete