ACS PRF | ACS
All e-Annual Reports
45603-AC9
Instrumentation Network Design and Retrofit
PROJECT ACTIVITIES
The original goals of the project were:
a) New network performance measures: residual reliability, and accuracy of estimators, which will help overcome practical problems in implementing current design methods.
b) An unconstrained optimization model, which relates all sensor network properties to the value of the information they produce and contrasts it with cost.
c) Inclusion of financial risk management in the problem.
Revised Scope of Work (Due to reduction from three years to two years)
The original proposal included:
- New Robustness measures (Accuracy and residual reliability): Of these, we will leave out the residual reliability, but will keep the work on accuracy.
- Unconstrained Optimization and Financial Risk: This part will consider the linear cases only. The non-linear systems will be taken out.
===============================================================
REPORT
Although the projects suffered from some problems, it is now well underway. The major outcomes are:
a) A paper outlining stochastic accuracy calculations is in press:
-Bagajewicz M. and DuyQuang Nguyen. Stochastic-Based Accuracy of Data
Reconciliation Estimators for Linear Systems. Computers and Chemical Engineering.
b) The code for incorporating accuracy calculations to the existing sensor network design code is underway. Results are anticipated to be available at the end of September and we expect to send a paper for publication by November. We are also happy to inform that this will cover the non-linear case as well as the linear case. This is more than the declared goal.
Summary of Project Findings
• Stochastic accuracy calculations were made for linear cases and will be extended to non-linear cases. Stochastic accuracy, which is more realistic than the definition of accuracy advanced by the PI prior to the start of this project\, is also showing smaller and more adequate predictions.
• The code to incorporate stochastic accuracy to sensor network design is underway. We expect some difficulties in computation time, but at least the conceptual part is taken care of.
• Reliability of sensor networks and the ability to calculate it is now well understood by Sanjay Kumar, the new student and will also be incorporated into the code.
• We will attempt to also introduce residual reliability.
Personnel
As it is explained below, the project employed one existing student (DuyQuang Nguyen) , funded from other sources (the author of the paper cited above), hired one student (Aravinda Buddhala), who after 3 months left, and now hired a student (Sanjay Kumar), who started in August 15, 2007.
The project was programmed to last between September 1, 2006 and August 31, 2008. At the time the project was approved, it was too late to hire a student. Classes start on Aug 15 at the University of Oklahoma, and recruiting usually takes place in April may, June at the latest. The PI immediately started the recruiting process and found a student (Aravinda Buddhala) who joined in January 2007. Although she signed up to take classes, she was immediately put to work on the project, but left it a few months later. Although this created a major setback for the project in terms of project results not being accomplished (when she was ready to start coding, she left), it has not, fortunately created a big financial problem, since only a portion of her stipend was paid by this grant. In addition, another student was assigned to finish the accuracy code corresponding to this project. In May, the PI started a new search and we have now a new student, Sanjay Kumar who is taking very few classes and is supposed to pick up where Aravinda Buddhala left.
Financial aspects: Only the PI summer salary was paid and some travel money was used. The student's stipends have not been used yet. In view of the fact that we had to bring a new student, and that his timing for graduation is August 2009, we are likely to request a non-cost extension next year because the funding will be present. Fortunately, the project is enjoying more support than the original ACS PRF commitment.
TRAINING AND DEVELOPMENT
The project is contributing to graduate two graduate students who will possess skills related to the monitoring of processes and the associated instrumentation upgrade problems.
OUTREACH ACTIVITIES
The PI and his group engaged in a number of dissemination activities.
- We made a presentation at the Texas Tech University Control Consortium in April 2007.
- A paper was presented at the Inter American Conference of Chemical Engineering held in Buenos Aires, October 2006. The paper laid out all the ideas regarding accuracy that are now being incorporated into the code.
- We submitted papers to the AIChE Annual Conference to be held in November 2007 in Salt Lake City.
