A decoupling-based unified fault detection and diagnosis approach for packaged air conditioners

Haorong Li, Purdue University


Existing methods addressing automated FDD for vapor compression air conditioning systems (1) require measurements over a wide range of conditions for training reference models, development of which can be time consuming and costly, and (2) can not deal with multiple faults that occur simultaneously. This thesis presents new methods that reduce engineering and installed costs for FDD, improve overall sensitivity for detecting and diagnosing faults, and handle multiple-simultaneous faults. ^ The mathematical formulation of model-based FDD techniques implied that decoupling is the key to handling multiple-simultaneous faults. To eliminate cost-prohibitive overall system models, an alternative physical decoupling methodology to mathematical decoupling was developed, which led to the decoupling-based FDD technique. During the mathematical development, a previously developed FDD method, termed the statistical rule-based (SRB) method, was re-examined and cast within the general mathematical framework, which led to two new detection and diagnosis classifiers with better performance and low implementation costs. Various component models and virtual sensors were proposed to generate decoupled features. These models are low-cost in that they exploit manufacturers' performance rating data and only require limited and readily available data for training. To justify fault service after diagnostics and evaluate the economic benefit associated with application of the proposed FDD technique, new economic evaluation approaches were developed, which involve a new overall economic performance degradation index termed EPDI. ^ Finally, the proposed techniques were validated using both laboratory and field data. The SRB FDD technique with improved components was shown to have better sensitivity than the original SRB method. A prototype was made to demonstrate the application of the decoupling-based FDD technique. Sensitivity tests showed that all the individual faults can be identified before they cause a 5% of degradation in cooling capacity, EER and SHR, and their EPDIs reach 10%. Robustness tests of forty-one multiple-simultaneous-fault combinations showed that there were only two false alarms and sensitivity losses for a liquid-line restriction. Preliminary application for sites in California showed that faults happen very frequently at the field sites: 71% are significantly impacted by faults, 38% have more than two simultaneous faults, and 43% justify service immediately. FDD evaluation showed that $108/ton-year, around 70% of the original service costs, can be saved, and the operation cost savings ranged from $20 to $180/ton-year. The savings are significant and the payback period for the technique is less than one year. ^




Major Professor: James E. Braun, Purdue University.

Subject Area

Engineering, Mechanical

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server