Fault Detection and Diagnostics for Rooftop Air Conditioners

49
4 A DECOUPLING-BASED FDD TECHNIQUE
Although the improved SRB FDD method has good performance for individual
faults, it requires measurements over a wide range of conditions for training reference
models. The development of these models can be time consuming and costly.
Furthermore, SRB FDD methods can only handle individual faults. This section
summarizes a new method, termed the decoupling-based FDD method, which reduces
engineering and installed costs for FDD and handles multiple-simultaneous faults. This
methods are evaluated using both laboratory and field data. Additional details of the
method and evaluation are given in Deliverable 2.1.5.
4.1 Approach
To handle multiple-simultaneous faults, the interactions among different faults
should be decoupled (refer to Figure 3-1). That is, if one independent feature, which is
impacted only by one fault, can be found for each individual fault, then multiplesimultaneous
faults are decoupled. For a linear system or some special nonlinear systems,
a transformation can be found to diagonalize a transfer function matrix to decouple the
system if a detailed system physical model is available. However, to obtain such a
detailed physical model taking faults into account for a rooftop unit system is extremely
difficult. Another way to decouple the system is to unfold the black-box representing the
rooftop unit system to view it from a microscopic point of view and find some
independent features with physical meaning for component-level faults, and isolate
service faults from operation faults immediately after service is done and when the
system stops. There is an important and practical restriction for the independence features.
They should be able to be expressed as functions of low-cost measurements such as
temperature and pressure.