- 上一篇：复合材料高强度混凝土的试验研究桥梁英文文献和中文翻译
- 下一篇：齿轮箱外壳综合设置和夹具规划系统英文文献和中文翻译
Thespecifications of the AHU and the operating range of temperatures and air flow conditions aredepicted in Table IV.In the FDD experiments conducted the supply air temperature was varied as a function ofroom load and the room air temperature was maintained at a chosen set-point by regulating theair flow rate to the room. 3. FAULT DIAGNOSIS THEORYThe methods for diagnosing the faults in HVAC systems comprise of both fault detectionand fault diagnosis. To detect faults, it is first necessary to define and classify what constitutes afault. This is usually referred as fault pattern classification. Several methods are used toanalyse and diagnose faults. These are: rule based diagnosis, statistical pattern recognition,neural networks and fuzzy logic. In this paper we use a model-based residual technique todetect faults. The residual is the difference between value of the state with no fault and its valuewith fault. These residuals are normalized and the pattern of the normalized residuals is used to detect the faults. In this study, the following residuals were defined and used in the faultdiagnosis analysis:ROD ¼ ODN ODF ð1ÞRTI ¼ TIN TIF ð2ÞRSFR ¼ SFRN SFRF ð3ÞRST ¼ STN STF ð4ÞRSF ¼ SFN SFF ð5ÞEquations (1)–(5) show residuals relative to normal states and failed states (outdoor airdamper, indoor temperature sensor, supply flow rate sensor and supply fan). The normalizedresiduals were computed by piding the residuals with the highest residual in its class using thefollowing equation:RT ¼ TN TFTN TF jjMAXð6ÞIn the above equations, the subscripts are defined as follows: OD=outdoor air damper,IT=indoor temperature sensor, SFR=flow rate sensor, ST=supply temperature, SF=fanoutput, N=no fault state, and F=fault state. In this paper we have used the normalizedresidual patterns as inputs to train neural networks. The trained neural network is used toidentify the residual patterns and recognize the fault according to a predefined faultclassification scheme.4. EXPERIMENTAL METHODSeveral elements of HVAC systems often fail during day-to-day operation. Among thesesensors, actuators, and fans are the most common elements that fail or give erroneous signals. Inorder to develop a robust fault diagnosis and detection technique, we have conducted severalexperiments in which faults were simulated and diagnosis analysis were performed to verify ifthe faults can be identified accurately.To this end, we consider a VAV–AHU operation by focusing attention on the following fourelements: outdoor air damper (OD), indoor temperature (IT) sensor, supply flow rate (SFR) andsupply fan (SF). During typical operation of HVAC systems in buildings these elements may failone at a time (single fault) or more likely they will fail concurrently in several combinations(multiple faults). The tests were conducted by simulating winter weather conditions in the EC.The outdoor temperature simulated was 58C corresponding to a typical mild day in winter. TheVAV–AHU was controlled by a central energy management control system. The operatingconditions of the VAV system were set as follows: supply air pressure set-point=440 Pa, supplyair temperature=368C; room temperature set-point=228C; outdoor damper open posi-tion=30%.
A series of experiments were conducted in which a single fault was applied one at a time. Thesingle fault classification scheme used in the tests is defined in Table V. The other sets ofexperiments conducted were to simulate multiple faults. These faults were characterized by simultaneous failures of two elements. These are also defined in Table V. An energymanagement control system was programmed to generate the phenomena similar to actualfaults. The elements were set to cause fault magnitudes of 10% from their original no faultsettings. The faulty signals with 10% error were applied to outdoor air damper and temperaturesensor. Also the output signals of supply flow rate sensor and the supply fan were increased byabout 10% in order to replicate fault state.5. RESULTS AND DISCUSSIONS5.1. Performance characteristics of the systemFigure 4 shows the outputs of each element such as supply temperature, heater powerconsumption, flow rate and damper opening that were monitored during the tests involvingsingle and multiple faults. The time evolutions of the outputs under single and multiple faultswere compared. Figures 4(a) and 4(b) show the outputs from the ST sensor operating in singlefault mode (Figure 4(a)) and multi-fault mode (Figure 4(b)). The multiple fault modes weregenerated using different combinations of ST+OD, ST+SFR or ST+SF elements.