Fault Detection and Diagnostics for Rooftop Air Conditioners

More documents

Recommendations

Info

50 In order to extend the easily-implemented SRB fault diagnosis idea to handle multiple-simultaneous faults, Deliverable 2.1.5 proposed a decoupling-based method, which decouples the interactions between the transformed FDD features Z and faults X (see Equation (4-1)). That is, it makes each entry of the feature vector Z only correspond to unique fault entries of the fault vector X and vice versa. ⎡λ1 ⎢ Z = ⎢ ⎢ ⎢ ⎣ λ 2 O ⎤ ⎥ ⎥ X ⎥ ⎥ λn ⎦ (4-1) Based on the decoupled features, the SRB FDD technique can be applied to handle multiple-simultaneous faults. That is, the decoupled to be n n − dimensional FDD problem has been 1 − dimensional SRB FDD problems. In addition, the decouplingbased diagnosis method simplifies fault detection from a high-D problem to n 1-D ones. Equation (3-1) boils down to the following n 1-D equations, ω1: normal 2 ( − µ ) ≤ 1 z i σ i, normal 2 i, normal > ω2: faulty or 2 − ( χ ) {(1 − α ),1} (4-2A) z i − µ σ i, normal i, normal ω1: normal ≤ > ω2: faulty N −1 {(1 − α ),0,1} (4-2B) 2 −1 −1 where, ( χ ) {} , is the inverse chi-square cumulative distribution function, N {} , is the inverse normal cumulative distribution function, α is the false alarm rate, and i = 1,2, L,n . Fault diagnosis automatically is achieved without any extra computation immediately after fault detection is finished, so the fault diagnosis classifier is not required.
51 This approach overcomes the drawback of the SRB diagnosis method and handles multiple-simultaneous faults diagnosis and becomes more generic and systemindependent and does not require complicated rules, which depend on the system. The above approach is based on decoupled features. Mathematically, there exists an infinite number of decoupled features, but for HVAC systems only those with intuitive physical meaning and those that are readily available (low-cost) are practical. Deliverable 2.1.5 develops a methodology or guidelines to find these kinds of features. This methodology or guideline first classifies the RTU faults from two criteria. From microscopic and macroscopic points of view, the seven faults can be divided into two classes: component-level and system-level faults, which are shown in Figure 4-1. If classified from the view of fault cause, they can be divided into: operational and service faults. ROOFTOP UNIT FAULTS COMPONENT-LEVEL FAULTS SYSTEM-LEVEL FAULTS Compressor Valve leakage Condenser Fouling Evaporator Fouling Liquid-Line Restriction Non-Condensable Gas Low Refrigerant Charge Leakage Undercharge Refrigerant Overcharge OPERATIONAL FAULTS SERVICE FAULTS Figure 4-1 Taxonomy of Rooftop Faults
Page 1 and 2:
CALIFORNIA ENERGY COMMISSION Final
Page 3 and 4:
Acknowledgements Jim Braun and Haor
Page 5 and 6:
Abstract Project 2.1, Fault Detecti
Page 7 and 8:
TABLE OF CONTENTS LIST OF TABLES LI
Page 9 and 10:
LIST OF FIGURES Page 1 - Field Test
Page 11 and 12:
The report “Description of Labora
Page 13 and 14:
mounted heat pumps for heating and
Page 15 and 16:
The retail stores are in Southern C
Page 17 and 18:
Figure 1 - Field Test Sites Data Co
Page 19 and 20:
Gibson School (Cont’d) Woodland S
Page 21 and 22:
Table 1 - Data List for Modular Sch
Page 23 and 24:
BUILDING TYPE: Modular School Rooms
Page 25 and 26:
HEATING / AIR CONDITIONING EQUIPMEN
Page 27 and 28:
Sacramento Area McDonalds PlayPlace
Page 29 and 30:
Bradshaw Road (Sacramento Area) McD
Page 31 and 32:
TEST INSTRUMENTATION: Tables 2 and
Page 33 and 34:
Table 2 - Data List for Inland Rest
Page 35 and 36:
Page 37 and 38:
Page 39 and 40:
Castro Valley (San Francisco Bay Ar
Page 41 and 42:
Castro Valley McDonalds PlayPlace P
Page 43 and 44:
more or less custom design, publish
Page 45 and 46:
BUILDING TYPE: Retail Store ADDRESS
Page 47 and 48:
TEST INSTRUMENTATION: Similar test
Page 49 and 50:
• Temperature and humidity levels
Page 51 and 52:
November 2001 - December/January 20
Page 53 and 54:
Table of Contents 1. Introduction..
Page 55 and 56:
1. Introduction All the thermodynam
Page 57 and 58:
Table 1.1 Comparisons of Three Mode
Page 59 and 60:
solution procedure involves the non
Page 61 and 62:
independent of the moisture content
Page 63 and 64:
function, f ( X , y) , if it can be
Page 65 and 66:
Figure 2.1 Neural-Network Implement
Page 67 and 68:
prototype was developed by using th
Page 69 and 70:
3. Comparison of black-box modeling
Page 71 and 72:
which is more accurate than the exp
Page 73 and 74:
Similar to GRNN, RBF has very good
Page 75 and 76:
Table 3.2 RMS error (Polynomial,GRN
Page 77 and 78:
Table 3.4 Contrast of Black-box mod
Page 79 and 80:
Polynomial plus GRNN Training Desir
Page 81 and 82:
interpolation(poly+GRNN) extrapolat
Page 83 and 84:
used to build the steady-state mode
Page 85 and 86:
α , ρ, τ I t t s h o A t a Figur
Page 87 and 88:
Temperature (F) 82 79 76 73 Condens
Page 89 and 90:
0.050 0.045 Pressure = 101.3 [kPa]
Page 91 and 92:
T T − T − T W = W −W −W m =
Page 93 and 94:
Figure 4.8 Output of three steady-s
Page 95 and 96:
will be calculated to represent the
Page 97 and 98:
180 175 RMS error = 0.3984 (F) Pred
Page 99 and 100:
180 175 RMS error =1.1472 (F) Predi
Page 101 and 102:
51 50 RMS error=0.3860 (F) Predicte
Page 103 and 104:
4.4 California field site results S
Page 105 and 106:
105 100 RMS error =0.5982 (F) Predi
Page 107 and 108:
Since air conditioners always cycle
Page 109 and 110:
6. Conclusions and future work So f
Page 111 and 112:
Lee, W., House, J.M. and Shin, D.R.
Page 113 and 114:
Table of Contents 1 Introduction...
Page 115 and 116:
AHU α β 2 d i EER ∆ ∆ η v T
Page 117 and 118:
$ !! !! )
Page 119 and 120:
Paper statistics in HVAC FDD Number
Page 121 and 122:
011>" - , ?" 8?"
Page 123 and 124:
+ : 6336" ,
Page 125 and 126:
+ :: !!- :: !!
Page 127 and 128:
6 24 122 7) 6 3++, ) . !! /011
Page 129 and 130:
)
Page 131 and 132:
336 F / s $ S
Page 133 and 134:
N $ V v1
Page 135 and 136:
Figure 3-4 2-dimensional residual d
Page 137 and 138:
10 5 Normal and current operation p
Page 139 and 140:
7 N Ω f N N Ω = Ω X ,
Page 141 and 142:
0 + α 6 d χ ( )
Page 143 and 144:
Normal operation region Residual-2
Page 145 and 146:
Table 3-2 Refrigerant Leak at 20% l
Page 147 and 148:
ratio dist = P F 1 P F 1 2 1 9.0
Page 149 and 150:
$ , " - (
Page 151 and 152:
)(( 0( 04 ) 6330" /
Page 153:
Table 4-2 Polynomial plus GRNN mode
Page 157 and 158:
4.2 m r,predict (kg/min) 4 3.8 3.6
Page 159 and 160:
6 1)( ( ( 6 1)( ) , ( )
Page 161 and 162:
- )
Page 163 and 164:
? "J> *"J060 #"J083 *$"J670 #"J08<
Page 165 and 166:
Table 4-17 Detected (normal, fault)
Page 167 and 168:
c=1 c=10 c=20 1 0.8 Distance ratio
Page 169 and 170:
$ /) $ 0111" .. !!
Page 171 and 172:
Chen, Bin and Braun, J.E, 2000. Sim
Page 173 and 174:
Ventilating, Air-Conditioning and R
Page 175 and 176:
1.T amb 2.T ra 3.RH ra Plant Prepro
Page 177 and 178:
F) $
Page 179 and 180:
+: + $ 5
Page 181 and 182:
.44(' * 5 09( ( 0( F N ( M , Σ)
Page 183 and 184:
Load 20% 40% 60% 80% 100% 3 4 Fault
Page 185 and 186:
Load 20% 40% 60% 80% 100% 3 4 Fault
Page 187 and 188:
3 4 0 0 0 0 0.4173 0 0.0010 0 0.000
Page 189 and 190:
.44(' $ () ) ) - $
Page 191 and 192:
∆P Restrictio n Level=100%* ∆P
Page 193 and 194:
2 TABLE OF CONTENTS TABLE OF CONTEN
Page 195 and 196:
4 A1.2.3 Valve Position Expression
Page 197 and 198:
6 Figure 2-10 Decoupling refrigeran
Page 199 and 200:
8 LIST OF TABLES Table 1-1 Fault di
Page 201 and 202:
10 N = Number of generated sample p
Page 203 and 204:
12 with a fixed-orifice as the expa
Page 205 and 206:
14 (residuals) should have a zero m
Page 207 and 208:
16 expected distribution of the res
Page 209 and 210:
18 operation. Another advantage is
Page 211 and 212:
20 1.1.2.1 Original SRB Fault Diagn
Page 213 and 214:
22 Corresponding to the SRB fault d
Page 215 and 216:
24 integration of the probability d
Page 217 and 218:
26 1. Simplifies fault detection fr
Page 219 and 220:
28 Step i Do FDD on fault i . After
Page 221 and 222:
30 ROOFTOP UNIT FAULTS COMPONENT-LE
Page 223 and 224:
32 has two possible causes: refrige
Page 225 and 226:
34 1.2.3 Decoupling of Component Fa
Page 227 and 228:
36 1.2.3.2 Condenser-Related Faults
Page 229 and 230:
38 mixture, χ ref is known as the
Page 231 and 232:
40 Table 1-3 also lists the refrige
Page 233 and 234:
42 Refrigerant Property T cond, pre
Page 235 and 236:
44 as an independent feature only f
Page 237 and 238:
46 evaporator air flow rate reducti
Page 239 and 240:
48 5. 2 Pll ∆ deviates drasticall
Page 241 and 242:
50 System-Level Faults RefUnder Ref
Page 243 and 244:
52 1.2.5 Summary of Decoupling Sche
Page 245 and 246:
54 noise, system disturbances and m
Page 247 and 248:
56 compressor data. When there is a
Page 249 and 250:
58 Figure 2-5 Decoupling evaporator
Page 251 and 252:
60 Total Pressure Drop of Liquid- L
Page 253 and 254:
62 Figure 2-11 Illustration of Demo
Page 255 and 256:
64 bypass the compressor. At this t
Page 257 and 258:
66 for T dis because there is no ov
Page 259 and 260:
68 Figure 2-14 Outputs of the FDD d
Page 261 and 262:
70 the system requires more refrige
Page 263 and 264:
72 recommended that: the system req
Page 265 and 266:
74 Figure 2-20 Outputs of the FDD d
Page 267 and 268:
76 Figure 2-21 Histogram bar plot o
Page 269 and 270:
Page 271 and 272:
Page 273 and 274:
82 2.3.2 Summarized Results for Oth
Page 275 and 276:
84 3 CONCLUSIONS AND RECOMMENDATION
Page 277 and 278:
86 Breuker, M.S. and Braun, J. E.,
Page 279 and 280:
88 Lee, W., House, J.M. and Shin, D
Page 281 and 282:
90 APPENDIX 1 PHYSICAL MODELS OF EX
Page 283 and 284:
92 A1.1.2 Short-Tube Models Many re
Page 285 and 286:
94 m& = CA ρ P up − P ) (A1-5) (
Page 287 and 288:
96 The abrupt change of CA for the
Page 289 and 290:
98 These equations can be combined
Page 291 and 292:
100 H h T sh , max opening T sh , s
Page 293 and 294:
102 A1.2.5 Parameter Estimation Met
Page 295 and 296:
104 T (2 T sh, ratingopening , sh,m
Page 297 and 298:
106 Harms’Result Harms plotted al
Page 299 and 300:
108 Mass flow rate Local linearizat
Page 301 and 302:
110 temperature on the RTD is unifo
Page 303 and 304: 112 in cold water application, it i
Page 305 and 306: 1 ACKNOWLEDGEMENTS The research tha
Page 307 and 308: 3 2.2.3 Todd Harms’ Data ........
Page 309 and 310: 5 LIST OF FIGURES Figure E-1. FDD d
Page 311 and 312: 7 LIST OF TABLES Table E-1 FDD resu
Page 313 and 314: 9 IA = Independence Assumption λ i
Page 315 and 316: 11 EXECUTIVE SUMMARY Packaged air c
Page 317 and 318: 13 current values of the fault indi
Page 319 and 320: 15 Figure E-4 Histogram of the EER
Page 321 and 322: 17 2. Operational cost savings, whi
Page 323 and 324: 19 Table E-5 Conservative Lifetime
Page 325 and 326: 21 constraints. Economic constraint
Page 327 and 328: 23 Paper statistics in HVAC FDD 35
Page 329 and 330: 25 on directional changes to identi
Page 331 and 332: 27 fault levels at different operat
Page 333 and 334: 29 was concluded that the method wa
Page 335 and 336: 31 However, several improvements ar
Page 337 and 338: 33 point sensor placement is genera
Page 339 and 340: 35 2 DATA SOURCES USED FOR EVALUATI
Page 341 and 342: 37 The fives types of faults are re
Page 343 and 344: 39 Occupation Type Climate Location
Page 345 and 346: 41 The SRB FDD method determines wh
Page 347 and 348: 43 The following sections describe
Page 349 and 350: 45 points rather than on a distribu
Page 351 and 352: 47 current operation point P 1, P F
Page 353: 49 4 A DECOUPLING-BASED FDD TECHNIQ
Page 357 and 358: 53 ⎡ ∆T ⎢ ⎢ ∆T 2 ⎢ ∆
Page 359 and 360: 55 improving the compressor model p
Page 361 and 362: 57 Condenser Air Mass Flow Rate (lb
Page 363 and 364: 59 Evaporator Air Mass Flow Rate (l
Page 365 and 366: 61 Liquid-Line Pressure Drop (PSI)
Page 367 and 368: 63 4.2.2 Purdue Field Emulation Sit
Page 369 and 370: 65 1. Evacuated the system, and the
Page 371 and 372: 67 normal _ value − current _ val
Page 373 and 374: 69 Figure 4-16 shows one frame afte
Page 375 and 376: 71 Figure 4-18 shows one frame afte
Page 377 and 378: 73 valve position can be seen from
Page 379 and 380: Figure 4-22 Outputs of the FDD demo
Page 381 and 382: 77 Figure 4-23 Histogram bar plot o
Page 387 and 388: 83 4.2.3.2 Summarized Results for O
Page 389 and 390: 85 5 ECONOMIC ASSESSMENTS Since the
Page 391 and 392: 87 inspection savings would be $2,0
Page 393 and 394: 89 Q & Cap = W & × EER (5-2) The e
Page 395 and 396: 91 4. electricity costs ( C e ). Ut
Page 397 and 398: 93 Table 5-3 lists estimates of equ
Page 399 and 400: 95 5.4 Smart Service Schedule Savin
Page 401 and 402: 97 7. A 6-ton RTU having a cost of
Page 403 and 404: 99 Location North California South
Page 405 and 406:
101 5. Three case studies were inve
Page 407 and 408:
103 REFERENCES Aaron, D. A., and P.
Page 409 and 410:
105 Davis, Coby. 1993. Comparison o
Page 411:
107 Rossi, T.M., 1995. Detection, D
show all

Fault Detection and Diagnostics for Rooftop Air Conditioners

Create successful ePaper yourself

Delete template?

Save as template?