Laird Research Group

Dr. Carl D. Laird
Assistant Professor and
William & Ruth Neely Faculty Fellow
Artie McFerrin Department of Chemical Engineering

Our expertise includes large-scale modeling, design and operations under uncertainty, parameter estimation, and inversion. Research applications include traditional chemical engineering processes, in addition to homeland security applications, infectious disease modeling and estimation, and network problems.

Pyomo - Optimization Modeling in Python 
(Springer Optimization and Its Applications)

by William E. Hart, Carl Laird, Jean-Paul Watson,
and David L. Woodruff

Hardcover Release Date: February 29, 2012

For more information, or to order from Springer, click here!
To purchase from Amazon, click here!

Today, vaccines play an important role in human lives. To address dynamic national requirements for various pharmaceuticals, there is a need for reconfigurable therapeutics manufacturing facilities that are capable of rapid product changeover. Limitations of the existing manufacturing infrastructure are made painfully clear by recurring drug shortages. This research focuses on developing a flexible modeling and nonlinear programming framework for optimizing virus-based production of therapeutics in a reconfigurable facility.

Continued gratitude goes to the National Science Foundation Faculty Early Career Development (CAREER) Award for their financial support.

Large-scale, nonlinear dynamic models frequently arise when describing the physics of important engineering and biological problems. In this research, we are developing new algorithms that exploit concurrent computing architectures to provide efficient parallel solution of large-scale parameter estimation problems, focusing on those with both temporal and spatial structure that can be exploited.

Continued gratitude goes to the National Science Foundation Cyber-Enabled Discovery and Innovation (CDI) for their financial support.

Current strategies to place gas sensors in industrial settings are based upon heuristics or semi-quantitative approaches. Optimal sensor placement is difficult due to the large number of unknown variables that influence the risks associated with gas leaks. Heuristic and semi-quantitative approaches often yield results that are far from the optimal sensor placements in terms of cost and risk-reduction. This research focuses on the use of numerical optimization techniques to fin optimal solutions to gas sensor placement problems. Additionally, this research investigates the effects of coverage-based constraints and Conditional-Value-at-Risk constraints to the sensor placement formulation. Further work examines the variability in solutions due to the number of scenarios used for data in solving for the placement.

Continued gratitude goes to the Mary K. O'Connor Process Safety Center for their financial support.

As the size of nonlinear optimization problems continues to increase, general purpose tools, coupled with standard desktop computing hardware, are not able to provide efficient solution. Furthermore, computer chip manufacturers are no longer focusing on increasing clock speed, but rather on hyper-threading and multicore architectures. This research focuses on the development of numerical solution strategies that can efficiently utilize emerging computing architectures like the GPU.

Continued gratitude goes to the National Science Foundation Faculty Early Career Development (CAREER) Award for their financial support.

This research focuses on the development of real-time algorithms for performing source inversion and response planning in the event of an accidental or intentional contamination in a water distribution system. Since these systems would work in a real-time setting, these algorithms must be able to solve efficiently and effectively for very large networks when given only limited measurement information due to limiting resources. To help meet these goals we have been developing efficient water quality models that can be used in an inversion context, as well as mixed-integer linear and quadratic programming formulations for source location identification and optimal sampling. These approaches have been shown to effectively find the source and extent of contamination given a real water network model.

Continued gratitude goes to Sandia National Laboratories and PUB Singapore for their financial support.

Reliable, mechanistic models for the spread of infectious disease are valuable for improving our understanding of important factors affecting disease dynamics and for helping guide public health decision making. Our work has focused on developing deterministic and stochastic discrete-time and continuous time models of childhood infectious diseases. We bring large-scale nonlinear optimization expertise to provide a framework for efficient estimation of long-term dynamic models. These estimation problems are similar to classic ill-posed inverse formulations, and effective regularization is required to obtain reliable seasonal estimates. Current research includes further refinement of model structure to improve their ability to extrapolate.

Continued gratitude goes to Sandia National Laboratories, and the ASCR program in the Office of Science of the Department of Energy for their financial support.

Concern regarding the accidental or intentional contamination in a water distribution system has led to the development of tools to perform source inversion and response planning for these systems. However, these tools are often difficult to install and then require a great deal of training to learn their proper use. To overcome these limitations, this research focuses on the development of a graphical, web-based user interface to serve as the front-end for the real-time algorithms for performing source inversion and response planning that have are being developed by other researchers.

Continued gratitude goes to Sandia National Laboratories and PUB Singapore for their financial support.

Recurring drug shortages remain a significant problem and demonstrate limitations in the current pharmaceutical manufacturing infrastructure. This work analyzes how the financial risks of pharmaceutical manufacturers can be reduced via a combination of governmental intervention and the availability of emergency production facilities. These resources allow for manufacturers to maintain a manufacturing presence for important pharmaceuticals with significant demand uncertainty. Additional research includes sensor placement in industrial settings and fast model predictive control.

Continued gratitude goes to the National Science Foundation Faculty Early Career Development (CAREER) Award for their financial support.

Current strategies to place gas sensors in industrial settings are based upon heuristics or semi-quantitative approaches. Optimal sensor placement is difficult due to the large number of unknown variables that influence the risks associated with gas leaks. Heuristic and semi-quantitative approaches can give results that are far from optimal in terms of cost and risk reduction. This research presents adaptations to a sensor placement formulation which incorporate sensor voting schemes and probability of sensor failure. In many instances, release events are not considered detected until multiple detectors acknowledge the release. Therefore, modified formulations must be developed to take into account this need for sensor voting. Additionally, this research considers adaptations to the original sensor placement formulation that account for the imperfect sensor, or the sensor that has some probability of failure.

Continued gratitude goes to the Mary K. O'Connor Process Safety Center for their financial support.