28.10.2014 Views

Fault Detection and Diagnostics for Rooftop Air Conditioners

Fault Detection and Diagnostics for Rooftop Air Conditioners

Fault Detection and Diagnostics for Rooftop Air Conditioners

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

4.4 Cali<strong>for</strong>nia field site results<br />

Since there are some problems with data collection in Cali<strong>for</strong>nia, there is not enough data<br />

to train models <strong>for</strong> every site. In this section, results are presented <strong>for</strong> one site, Milpitas.<br />

The rooftop unit has one stage <strong>and</strong> is installed in a McDonalds. This rooftoop unit is a<br />

little different from the unit at Purdue field site. First, there is no mixed air humidity<br />

sensor available, so the damper position reading is used to calculate the mixed air humidity.<br />

Second, from the data collected during August 28 to September 26 (Figures 4.24 <strong>and</strong> 4.25),<br />

this unit kept cycling very frequently <strong>and</strong> seems that it never reached steady-state, so the<br />

steady-state detector parameters were modified to get some “steady-state” data (Figure<br />

4.26). A moving window steady-state detector with a fixed window width of 2 was used.<br />

Figures 4.27 <strong>and</strong> 4.28 show the interpolation <strong>and</strong> extrapolation per<strong>for</strong>mance <strong>for</strong><br />

condensing temperature, respectively. The interpolation per<strong>for</strong>mance was evaluated by<br />

r<strong>and</strong>omly selecting two-thirds of the total data during August 28 to September 26 to train<br />

the model <strong>and</strong> one-third of the total data to test the model. For figure 4.28, the data<br />

collected during August 28 to September 15 was used to train the model <strong>and</strong> the data of<br />

September 15 to 26 was used to test the extrapolation per<strong>for</strong>mance. These two figures<br />

show that condensing temperature models worked reasonably well. Evaporating<br />

temperature models had RMS errors of 0.7528 F <strong>for</strong> interpolation <strong>and</strong> 1.1152 F <strong>for</strong><br />

extrapolation. However, discharge line temperature models worked poorly <strong>and</strong> their RMS<br />

errors were 2.345 F <strong>for</strong> interpolation <strong>and</strong> about 4 F <strong>for</strong> extrapolation. The discharge line<br />

temperature takes the largest time to reach steady-state.<br />

90<br />

Evaporating Temperature (F)<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

8/28/2001<br />

0:00<br />

9/1/2001<br />

15:20<br />

9/6/2001<br />

18:50<br />

9/11/2001<br />

12:25<br />

9/16/2001<br />

2:31<br />

9/20/2001<br />

19:31<br />

9/26/2001<br />

8:56<br />

Time<br />

Figrue 4.24 Frequent cycling of Milpitas site during August 28 to September 26<br />

52

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!