Evanghelos Zafiriou - US grants

This award is to provide research support to Dr. Zafiriou under the National Science Foundation's Presidential Young Investigator Awards (PYIA) Program. The objectives of the PYIA Program are to provide support to the Nation's most outstanding and promising young science and engineering faculty. The awards are intended to improve the capability of U.S. academic institutions to respond to the demand for highly qualified science and engineering personnel for academic and industrial research and teaching. Dr. Zafiriou's field of esearch is process control theory and applications. In particular, he will to work in three areas: (1) Model Predictive Control (MPC): MPC algorithms use a model to predict future values of the process outputs. Modeling error creates the need for designing controllers which are robust with respect to model/plant mismatch. The problem is more complex when hard constraints are present either as physical limitations or as performance and safety specifications, because in that case the overall closed-loop system is nonlinear, even if the plant dynamics are assumed linear. His goal is development of computer-aided design and tuning methodologies that directly account for modeling error. (2) Batch Control: Optimal control theory is often used to determine the input profiles for fed-batch (semi-batch) processes. These profiles, however, may not perform well when applied to the actual process, since they are only optimal for the model that was used. The PI is looking at the development of methods that appropriately modify the input profile during the course of successive batches, by utilizing both the model and plant information from the previous batches. This work is applicable to both polymerization and biochemical systems. (3) Neural Computing in Process Control: The PI's focus will be on the use of neural networks for sensor/actuator failure detection. Significant interactions exist between the operation of a control system and that of the diagnostic module. The main source of these interactions is the presence of model-plant mismatch. He will be examining the use of neural networks for distinguishing between patterns characteristic of control system failure and those corresponding to performance deterioration caused by modeling error.

; R o o t E n t r y F W g C o m p O b j b W o r d D o c u m e n t O b j e c t P o o l 3 g 3 g - . / 0 1 2 3 4 5 6 7 F Microsoft Word 6.0 Document MSWordDoc Word.Document.6 ; 9527576 Krishnaprasad This award is the University of Maryland with a sub-contract to North Carolina State University at Raleigh. The overall goal of this effort is to demonstrate a methodology for sensor-integrated control of rapid thermal chemical vapor deposition (RTCVD) of polycrystalline silicon (poly Si) from silane with focus on controlling deposition thickness and across-wafer uniformity. The project exploits advances in real-time sensors, including pyrometry for temperature, thermal imaging for temperature uniformity, and sampling mass spectrometry for thickness metrology and process ambient monitoring. Reduced-order process models constructed from high fidelity heat and fluid flow simulations, together with physically-based dynamic equipment, process, and sensor simulations, are the basis for control models. Resulting run-to-run control methodologies for controlling deposition thickness and across-wafer uniformity are being developed and validated experimentally, and real-time control approaches are being explored. These run-to-run control approaches will be extendible to real-time c ontrol. An architecture to support a basic supervisory control component is being demonstrated, using physically-based dynamic simulation to determine sensor signatures of specific equipment failure modes, together with advanced algorithms as interference tools for detecting sensor signal correlations and identifying indicated equipment/process malfunction. The investigators at the University of Maryland provide the effort on simulation and control, while the investigators at North Carolina State University provide the effort on sensors and on rapid thermal chemical vapor deposition of polycrystalline silicon. The experimental proof of concept of the control system will be performed in the cluster tool deposition apparatus at North Carolina State University. *** 0 0 Oh +' 0 $ H l D h , \\CLM15\SMURPHY$\WWUSER\TEMPLATE\NORMAL.DOT S u m m a r y I n f o r m a t i o n ( , 9527576 SHERONDA MURPHY SHERONDA MURPHY @ X g @ @ X g @ Microsoft Word 6.0 2 ; e = e d d l l l l l l l 1 % D T G 9 l l l l l l l l l s 9527576 Krishnaprasad This award is the University of Maryland with a sub-contract to North Carolina State University at Raleigh. The overall goal of this effort is to demonstrate a methodology for sensor-integrated control of rapid thermal chemical vapor deposition (RTCVD) of polycrystalline silicon (poly Si) from silane with focus on controlling deposition thickness and across-wafer uniformity. The project exploits advances in real-time sensors, including pyrometry for temperature, thermal imaging for temperature uniformity, and sampling mass spectrometry for thickness metrology and process ambient monitoring. Reduced-order process models constructed from high fidelity heat and fluid flow simulations, together with physically-based dynamic equipment, process, and sensor simulations, are the basis for control models. Resulting run-to-run control methodologies for controlling deposition thickness and across-wafer uniformity are being developed and validated experimentally, and real-time control approaches are being explored. These run-to-run control approaches wi

The long range goal of this project is to develop a computational formalism, built on stochastic Petri nets (SPN) that will enable accurate and dynamic calculation of all system variables and their variances. In particular, computational models of two Escherichia coli genetic circuits (emergence and decay of heat shock transcription factor, s?32, upon heat shock, and cell-to-cell communication or "quorum sensing") will be constructed, validated, and combined, suggesting the incremental assembly of predictive models that will ultimately predict system-wide behavior. Quorum sensing is known to determine the virulence of Pseudomonas aeruginosa (causative agent for cystic fibrosis), E. coli O157:H7 (Enterohemorrhagic E. coli), and Salmonella typhimurium (food poisoning). The focus is on the identification of signature genes that contribute to the synthesis and perception of autoinducer-2 (the signal molecule for quorum sensing), as well as the interplay between this circuit and the ?s32 circuit. By assembling combined circuits and by creating an optimization formalism that captures the stochastic variance, the Principal Investigators (PIs) will create an approach by which phenotype can be predicted, manipulated, and ultimately optimized. The specific optimization objective for the proposed work is to streamline the synthesis process of biologically active recombinant proteins in E. coli and it provides significant motivation for this work. It is expected that stochastic variance information is important in understanding these circuits. However, the approach is intended to evaluate this assertion at each stage of the work.