Research

      Extreme Event Simulation      

BUILDING FIRE EVACUATION SIMULATION

A Heuristic Optimization Based Methodology For Fire Evacuation Simulation Incorporating Human Behaviors

- In the design of fire safety issue for a building, two problems are critical:
1) inadequate evacuation enabling infrastructures and
2) Life Safety Codes are not sufficient to ensure fire safety.
The framework of a Particle Swarm Optimization based evacuation model Vacate is developed to solve these two problems. G. Taygi, C. Bloebaum, P. DesJardin (read more)

A Particle Swarm Optimization Based Behavioral And Probabilistic Fire Evacuation Model

- We are incorporating the fire hazard model, critical human behaviors, and the probabilistic decision-making system into Vacate to help fire safety engineers predict the evacuation scenarios more confidently. Z. Xue, C. Bloebaum, P. DesJardin (read more)

Multi-Floor Building Fire Evacuation Simulation

- We are interested in how stairways can impact the evacuation efficiency in fire emergency. Currently a simplified 3D multi-floor building model is set up and Vacate is being modified to adapt the stairways. A large scale evacuation simulation for extreme events which involve multi-floor, outside door, and vehicle evacuation is identified as our final goal. Z. Xue, C. Bloebaum (read more)

EMERGENCY VEHICLE EVACUATION SIMULATION

Emergency Vehicle Evacuation Simulation

-The purpose of the project is to help road network designers and engineers test the safety and evacuation efficiency of road networks under the extreme conditions like fire, earthquake, tsunami, tornado and man-made explosions. S. Klump, C. Bloebaum (read more)

PEDESTRIAN EVACUATION SIMULATION

Pedestrian Evacuation Simulation

- To face the increase of the large scale emergencies in today’s world, we need to be well-prepared before those disasters happen. Vacate-Out gives us the confidence as prepared by simulates critical outside door human behaviors like evacuating in pre-existing groups, waiting for other agents, investigating, taking alterative paths, and etc. B. Ries, C. Bloebaum (read more)

VOLCANIC RISK MITIGATION

Volcanic Risk Mitigation

- Using the simulations, users can estimate how fast and how far lava flows from a volcanic eruption would travel and in which direction. Users can feed that data into computer models, from which users can calculate the probability that sliding material will destroy towns and roads. Factors such as turbulence and viscosity of the flow, the coefficient of friction and the flow's starting velocity are taken into account. E. Winer, C. Bloebaum (read more)

      Visualization of n-Dimensional Data for Design Selection

Hyper-space Pareto frontier visualization with Hyper-Space Diagonal Counting (HSDC) method

-This visualization approach is based on the Hyper-Space Diagonal Counting (HSDC) method. An n-dimensional Pareto frontier can be intuitively visualized in a lossless 2-D fashion.

Hyper-Radial Visualization (HRV) method for N-dimensional data visualization.

-This new n-Dimensional visualziation method is based on a hyper-radial representation. The HRV is being used for concept selection in multi-attribute design.

Graph Morphing Representation.

-Graph Morphing Representation can help designers to obtain a better understanding of their complex problems.

     Web-based Insfrastructures for Design and Collaboration

Web-Based Visualization Environment For Decision-Making In Multidisciplinary Design Optimization

- This research demonstrates a methodology for combining a physical representation of a design artifact with the associated design process and sensitivities with the express purpose of facilitating trade-off decisions in a design process. The methodology provides designers a capability to simultaneously consider the impact a design change will have both on the product being designed and the analyses necessary to evaluate the impact of any changes in the product design. G. Agrawal, C. Bloebaum (read more)

Decision Support Tool For Multidisciplinary Design Optimization (MDO)

- The three different decomposition domains of complex product development, and how a design change in any of the decomposed elements in one of the domains would affect the decomposed elements of that and other domains as well as the overall design process are investigated thoroughly. The web-based, interactive decision support tool can simultaneously display and capture the dependencies within and between the DSM representations. S. Parashar, C. Bloebaum (read more)

Visual Dependency Structure Matrix (VDSM)

- Due to the high levels of inter-connectivity encountered in complex systems, a method of evaluating coupling strengths is important to finding ways of reducing computational costs in the analysis and, subsequently, optimization process. Designers now have the capability to visually process such issues as analysis module time and cost, local coupling sensitivity, and total derivative-based sensitivity metrics. K. English, C. Bloebaum (read more)

Geographic Independent Virtual Environment (GIVE)

- This research demonstrates a methodology for combining a physical representation of a design artifact with the associated design process and sensitivities with the express purpose of facilitating trade-off decisions in a design process. The methodology provides designers a capability to simultaneously consider the impact a design change will have both on the product being designed and the analyses necessary to evaluate the impact of any changes in the product design. G. Agrawal, C. Bloebaum (read more)

     Multiobjective Optimization Methods for MDO

Multi-Objective Range/Target Concurrent Subspace Optimization (MORTCSSO) method.

- The Multi-Objective Range/Target Concurrent Subspace Optimization (MORTCSSO) method is a hybrid approach based on the MORCSSO and MOTCSSO methods. MORTCSSO method can express the designer’s preferences through a combination of range and target specifications.

Multi-Objective Genetic Algorithm Concurrent Subspace Optimization (MOGACSSO) method.

- The Multi-Objective Genetic Algorithm Concurrent Subspace Optimization (MOGACSSO) method is a Genetic Algorithm-based heuristic solution strategy, and this method can handle multi-objective problems in coupled multidisciplinary design.

Multi-Objective Pareto Concurrent Subspace Optimization (MOPCSSO) method

- The Multi-Objective Pareto Concurrent Subspace Optimization (MOPCSSO) method Is based on MDCSSO method, and it is developed to concurrently handle the conflicting objectives that exist in multi-objective MDO problems. In MOPCSSO, the subspace optimizations deal with multiple individual objective functions instead of using an aggregate objective function.

An MDO Test Suite

- An MDO Test Suite is under construction that will be available to the MDO community for testing and validation of a variety of MDO methodologies, including meta-model development (i.e. replacement analysis), optimization method development, and sensitivity analysis, amongst others.

     Optimization For Medical Imaging

Optimization in Medical Imaging

- Medical Imaging procedures frequently require inter- or intra-modality 3D object alignment. This occurs most often in diagnostic imaging involving structural and functional imaging or cancer treatment surveillance imaging with large time gaps between acquisitions.
The alignment procedure consists of two main components, the similarity metric and the optimization routine used to achieve the optimal alignment. Despite large efforts and the use of many different approaches form both mathematical (Gradient Descent, Newton-Rhapson, Simplex) and heuristic (Genetic Algorithms, Simulated Annealing) are all methods still very computationally expensive.
This project focuses on using particle swarm optimization as a basis for a fast convergence to a global optimal object alignment. The method benefits from efficient solution space traversing and its prevention of being trapped into local optima. Starting from a single swarm approach we venture into parallel swarm global optimization with inter-swarm knowledge and different swarm characteristics.
Our evaluations are based on mouse-jaws acquired in collaboration with the UB Dental School for a study measuring jaw bone decay. The jaws are from different mice and have been acquisitioned with translational and rotational differences, making any evaluation approaches impossible. Initial results showed a fast and good convergence to a near global optimal result , bringing both jaws in very good alignment and build the foundation for more detailed evaluations using different 3D objects and algorithm variations.