Detection, diagnosis, and evaluation of faults in vapor compression equipment
Abstract
This thesis develops techniques for automated detection, diagnostics, and evaluation of faults in vapor compression equipment. Fault evaluation was added to the more common steps of fault detection and diagnostics to consider the special aspects of performance degradation faults over abrupt faults. A model for testing these techniques in a simulation environment was developed. The model is described and experimental validation results are presented in this thesis. The model has a modular design that enables easy adaptation to different equipment configurations. It can be easily tuned with a few simple measurements and can simulate faults in the cycle. The fault detection technique described in this thesis evaluates the impact of measurement errors on the confidence that current measurements are different from the predictions of a normal performance model. If the statistical confidence exceeds a predetermined threshold, then a fault is indicated. Diagnostics are performed by statistically evaluating a generic set of rules indicating the direction change of each measurement. This diagnostic technique does not require a learning phase for each piece of equipment, is capable of detecting a 5% refrigerant leak, and can distinguish between refrigerant leaks, condenser fouling, evaporator fouling, liquid line restrictions, and compressor valve leakage. Four fault impact evaluation criteria were developed to determine if the fault is severe enough to justify the service cost. These criteria are: comfort, economics, safety, and environmental hazard. Evaluating these criteria evolved into a constrained optimization problem to minimize lifetime service and energy costs while maintaining the other criteria as constraints. This problem was solved exactly using dynamic programming to create a minimum cost baseline for comparison with a simplified near-optimal scheduler, regular interval maintenance, and comfort constrained only maintenance. It was found that optimal service scheduling reduced lifetime operating costs by as much as a factor of two over regular service intervals and 50% when compared to constrained only service. The near-optimal algorithm gave operating costs that were within 1% of the optimal results.
Degree
Ph.D.
Advisors
Braun, Purdue University.
Subject Area
Mechanical engineering|Industrial engineering|Operations research
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.