Solving the ADAPT Benchmark Problem - A Student Project Study

Erik Almqvist, Daniel Eriksson, Andreas Lundberg, Emil Nilsson, Erik Frisk, and Mattias Krysander
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_10_114.pdf408.12 KBSeptember 20, 2010 - 12:56am

This paper describes a solution to the Advance Diagnosis and
Prognostics testbed (ADAPT) diagnosis benchmark problem. One main
objective were to study and discuss how engineering students, with
no diagnosis research background, would solve a challenging
diagnosis problem. The study was performed within the framework of
a final year project course for students studying control
engineering. A main contribution of the work is the discussion on
the development process used by the students. Key steps process
include a local view, resulting in component-local residual
generators/fault detectors. Also, a diagnosability analysis method
were used interactively during development to guide test
selection, i.e. which residual generators that were needed to
meet diagnosis performance requirements. The component-local view
proved beneficial, not only with respect to diagnosis performance
but also with respect to the development process, allowing
parallel development and reusable designs. The solution is based
on physical models of the components and includes common
techniques from control theory, like observers and parameter
estimators, together with established algorithms for consistency
based fault isolation. The system is fully implemented in C++ and
evaluation, using the DXC software platform, shows good diagnosis
performance.

Publication Control Number: 
114
Submission Keywords: 
diagnosis
residual generation
fault isolation
adapt benchmark
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed