Dr.
Som Karamchetty
Sole Member, SomeTechnologies, LLC
Program Manager, Mechanical Technology Inc.,
Managed advanced technology research programs in gas turbine
systems, organic Rankine cycles, and other thermal
systems, mechanical systems, and advanced software applications. Invented
several thermal systems and devices and received one
A Brief Description of Straight Stack Algorithm and Software
Development Program for Air Force:
Program managers at Mechanical Technology Incorporated (MTI),
I will briefly explain the technical problem. The gas turbine
engines under investigation typically had sixteen compressor disks that were
assembled together into the compressor by means of long bolts that passed
through sixteen sets of bolt-holes in each rotor. When they were
remanufactured, the disks have a small eccentricity and a small skew. Of
course, an ideal rotor will have no eccentricity and is parallel on both faces.
Therefore, a practical assembly of a compressor was never a true cylinder but
had a curved profile for the final rotor. When such a rotor was spun at the
typically high gas turbine speeds, the eccentricities caused severe vibrations.
In cases where the effects of the eccentricities and skews accumulated in a
positive fashion, the resulting vibrations were intolerable and occasionally
destroyed the rotor assembly. Dynamic analysis of the rotor revealed
permissible profiles of eccentricity. The assembly process was designed to take
advantage of these permissible limits by aligning the bolt-holes on each rotor
with its neighbors.
The final compressor could be assembled into sixteen to the power
fifteen combinations. Therefore, a manual assembly and testing process took
forever. So MTI engineers proposed a computer modeling of the assembly. But
even on a computer, the combinations were so numerous that brute-force search with realistic time limit was impractical. I
developed a novel method, called the stacking algorithm.
I started with measuring and characterizing the rotors by means of
a laser coordinate measurement machine (LCMM). From the coordinate
measurements, eccentricity and skew for each rotor were obtained. I developed
the stacking algorithm by combining the “branch and bound” technique used in operations research and search techniques common
in artificial intelligence (AI). I led a group of programmers in elegant
computer storage techniques to handle the combinations as the search for a
solution progressed. I recognized that in most practical situations, the
solution space was large. Finding any single solution was sufficient. There was
no need to discover all possible solutions. In most practical cases, my
opportunistic method provided a solution, as verified by simulation programs. I
prescribed physical time limits on the computer to terminate a search. The
stacking algorithm and computer system worked exceptionally well. This is an
example of an affordable engineering solution to a military problem utilizing
techniques available in the academic and industrial research communities.
Renewable Energy Resources
Proposal for
sought proposals to identifying renewable energy resources for the State.
Mechanical Technology Incorporated (MTI) decided to bid and appointed me as the
proposal manager. I analyzed the request for proposal (RFP) and assessed the
science and technology talents required. I assembled a team of outside
consultants and subcontractors.
The proposal team consisted of several companies with specialists
and experts in solar, wind, biomass, and energy conservation technologies. We
added engineers to carry out systems analysis and economists to conduct
economic viability analyses. We developed a work breakdown structure (WBS) and
estimated resource requirements and arrived at a bid cost. We wrote a proposal,
presented a briefing, and conducted negotiations. Our technical proposal was
judged the best among the bidders.
HARVEST:
I developed an inference engine using IQLisp.
Harvest enabled experts to input fault trees and users to obtain diagnostics.
Harvest (advanced version) had connections to sensors on machines in order to
conduct on-line diagnostics. Harvest was used in demonstrating a prototype
diagnostics program for the