Experimental Assessment of Environmental and Economic Impacts of Dampers’ and Filters’ Faults in a Typical Air-Handling Unit in Southern Italy

About 36% of the world’s energy demand and 37% of greenhouse gas emissions are attributed to buildings. Heating, Ventilation and Air-Conditioning (HVAC) systems play a major role, accounting for between 50 and 60% of building energy demand and between 10 and 20% of overall energy consumption [1, 2]. The scientific literature points out that HVAC systems are frequently operated under faulty conditions due to lack of proper maintenance, components’ failure or incorrect installation, significantly affecting their energy, environmental and economic performance [3, 4]. Companies typically use two different types of maintenances (reactive or preventive maintenance) to ensure the reliability of HVAC units [5, 6]. The Automated Fault Detection and Diagnosis (AFDD), which automatically detects and diagnoses deviations from normal operation problems, could address some critical points that both approaches have [2, 7, 8]. According to several studies [6, 8, 9], AFDD tools can provide a number of advantages, including enhancing indoor comfort and equipment lifespans, as well as reducing greenhouse gas emissions, energy costs, maintenance and operating expenses [10, 11]. The techniques used for conducting AFDD analyses can be categorized into: (i) quantitative model-based, (ii) qualitative model-based, and (iii) data-driven methods [7, 9, 12]. The data-driven methods are gaining more and more interest mainly thanks to those facts the (i) they can enhance the accuracy of detection and diagnosis, as well as (ii) they can be adopted even when there’s limited physics-based knowledge [7, 8, 13]. However, data driven-based AFDD tools can be exploited to identify faults by analyzing changes in specific variable patterns and looking for causes based on the comparison between normal and faulty field operation. [7, 8, 13]. Therefore, they necessitate labelled, dependable, cost-effective, and scalable experimental data; however, acquiring such data is really challenging and time-consuming [7, 8, 14]. It’s worth noting that only a limited number of studies performed quantitative analyses by using real-world data assessing how different types and severities of faults can impact users’ comfort, key parameter patterns, energy usage, greenhouse gas emissions, operational and maintenance expenses, and equipment lifespan [4, 15-17]. Only few studies have been focused on single-duct dual-fan constant air volume air-handling units (AHUs) [17-19]. Even if the occurrence of dampers’ faults in AHUs has been analysed in a number of scientific works [20], faults on filters of AHUs have been studied only in very few papers [20]; in particular, none of them has been carried out with reference to Italian climatic conditions. In light of these observations, further efforts could be crucial in shaping the future advancement of AFDD technologies with reference to the occurrence of faults on dampers and filters in AHUs. In this study, the operation of a typical single-duct dual-fan constant air volume AHU, integrated in the SENS i-Lab of the Department of Architecture and Industrial Design of the University of Campania Luigi Vanvitelli (located in Aversa, south of Italy), has been experimentally characterized on the basis of daily tests performed under both normal and faulty conditions during both Italian summer and winter seasons. In particular, the operation of the AHU has been analysed while artificially introducing the following 7 different typical faults: (1) return air damper kept always closed (stuck at 0%), (2) fresh air damper kept always closed (stuck at 0%), (3) fresh air damper kept always open (stuck at 100%), (4) exhaust air damper kept always closed (stuck at 0%), (5) fresh air filter partially clogged at 50%, (6) supply air filter partially clogged at 50% and (7) return air filter partially clogged at 50%. The experimental data have been investigated comparing the ones obtained under faulty conditions with data acquired under fault-free operation under similar boundary conditions. The aim of this investigation is to assess the impact of the identified faults in terms of CO₂ emissions and operating costs to support the development of data-driven AFDD tools. Section 2 describes the experimental set-up, while Section 3 details the experimental tests. Section 4 compares the boundary conditions of the experiments, while Section 5 discusses the effects associated with the faults’ occurrence.

Twenty-eight daily fault free and faulty experiments are carried out (from 9:00 a.m. up to 6:00 p.m.) in order to investigate the AHU behaviour under Italian summer and winter seasons. In particular, 14 tests (SN1, SN2, SN3, SN4, SN5, SN6, SN7, WN1, WN2, WN3, WN4, WN5, WN6 and WN7) have been performed under normal (N) conditions; in more detail, the tests (SN1, SN2, SN3, SN4, SN5, SN6 and SN7) have been carried out during the summer 2022 (S), while the other tests (WN1, WN2, WN3, WN4, WN5, WN6 and WN7) have been developed during the winter 2022/23 (W). The remaining 14 experiments (SF1, SF2, SF3, SF4, SF5, SF6, SF7, WF1, WF2, WF3, WF4, WF5, WF6, and WF7) have been performed during both the summer 2022 (S) and the winter 2023 (W) while manually forcing the positions of the dampers or partially clogging filters (by covering the filters with a single-wall corrugated cardboard, where the covered area (CP) represented the fault severity). Table 1 reports the operating conditions associated to the faulty experiments in terms of opening percentages of the return air damper (OP_DRA), the outdoor air damper (OP_DOA) and the exhaust air damper (OP_DEA), as well as the CP of the return air filter, the outdoor air filter and the supply air filter (CP_RAFil, CP_OAFil and CP_SAFil, respectively).

Table 1. Operating conditions of the faulty experiments

All tests, conducted under both normal and faulty conditions, were performed with TSP,Room and RHSP,Room set to 26℃ (with a DBT of 1℃) and 50% (with a DBRH of 5%), respectively, during summer. Conversely, during winter, TSP,Room and RHSP,Room were set to 20℃ (with a DBT of 1℃) and 50% (with a DBRH of 5%). The supply and return air fan velocities were maintained at 50%, regardless of the test conditions.

In this section, the comparison between the fault free tests and the corresponding faulty tests has been performed, from different points of view, to highlight the impact of each investigated fault on the AHU performance. In particular, the Section 5.1 summarizes the effects of the selected faults on both comfort and electric energy demand; the Section 5.2 focuses on global equivalent CO₂ emissions, while the Section 5.3 shows the impacts on operating costs.

5.1 Effects of fault on comfort and electric energy demand

The above-mentioned normal and faulty tests have been compared in terms of comfort and electric energy demands. In particular, Figure 2 reports the percentage comfort time difference (%DCT) between each faulty and corresponding normal calculated as follows:

$\% \Delta C T=C T_{\text {Faulty }}-C T_{\text {Normal }}$           (1)

where, CT_Faulty and CT_Normal represent the percentage of time under thermal or hygrometric comfort conditions (i.e., the percentage of time during which indoor air temperature or relative humidity is within the given deadbands) corresponding to faulty and normal tests, respectively. A negative value of this parameter indicates that the hours of comfort associated with the faulty test are lower than the corresponding normal test.

2.png

Figure 2. Values of %DCT as a function of the tests

The comparison in terms of %DCT highlights that the worst fault in terms of hygrometric comfort time is the fault F4 (exhaust air damper stuck at 0%) during winter, which corresponds to a considerable reduction by about 29% with respect to the test WN5.

Figure 3 reports the parameter %DEE indicating the percentage difference between the faulty and the corresponding normal test in terms of daily electric energy consumption:

$\% \Delta E E=\frac{\left(E E_{ {Faulty }}-E E_{ {Normal }}\right)}{E E_{ {Normal }}} \times 100$            (2)

where, EE_Faulty and EE_Normal are, respectively, the electric energy consumption of AHU during the faulty test and the corresponding normal test. A positive value of this parameter means that the electric energy demand associated with the faulty test is larger than the corresponding normal test.

3.png

Figure 3. Values of %DEE as a function of the tests

This figure underlines that the occurrence of the faults is not really relevant during summer, with values of %DEE ranging between -6.33% and +3.87%; in particular, the most adverse fault during summer in terms of %DEE is F7. Conversely, the faults are characterized by a significant impact in terms of electric demand during winter; in more detail, the faults F1, F3, and F4 cause a greater consumption in comparison to the normal operation by 13.79%, 10.15%, and 12.63%, respectively.

5.2 Effects of faults on global equivalent CO₂ emissions

The environmental assessment is performed through an energy output-based emission factor approach [26]. According to this approach, the mass m_x of a given pollutant x emitted while producing the energy output EE, can be worked out as:

$m_x=u_x^{E E} \times E E$          (3)

where, $u_x^{E E}$ is the energy output-based emission factor, that is the specific emissions of x per unit of EE. This factor depends upon several operating and structural variables, such as the specific equipment, partial load operation, age, state of maintenance, outdoor conditions, and pollutant abatement systems. CO₂ emissions prove to be quantitatively much more important than other greenhouse gas emissions [27]. In particular, the CO₂ emission factor associated to the electricity generation in Italy $u_{C O_2}^{E E}$ depends on day as well as the time of the day. Figure 4 reports this factor for each couple of corresponding normal and faulty tests as a function of the time derived from the study [28] according to the day during which the faulty test has been performed (by assuming that each faulty test and its corresponding normal test take place on the same day).

4.png

Figure 4. CO₂ emission factor associated to electricity consumption as a function of both tests and time [28]

The AHU environmental performance associated to the faulty tests in comparison with those of the corresponding normal tests have been assessed as follows:

$\begin{aligned} & \% \Delta m_{C O_2}=\frac{\left(m_{\mathrm{CO}_2, {Faulty}}-m_{\mathrm{CO}_2, { Normal }}\right)}{m_{C O_2, { Normal }}} \times 100 =\frac{\left(u_{C O_2}^{E E} \times E E_{{Fuulty}}-u_{\mathrm{CO}_2}^{E E} \times E E_{ {Normal }}\right)}{u_{C O_2}^{E E} \times E E_{ {Normal }}} \times 100\end{aligned}$          (4)

where, $m_{{CO}_2, { Faulty }}$ and $m_{{CO}_2, { Normal }}$ are, respectively, the global equivalent CO₂ emissions of AHU components in the case of faulty and normal conditions, EE_Faulty and EE_Normal are the electric energy consumptions during faulty tests and corresponding normal tests, respectively. A positive value of this parameter indicates that the global equivalent CO₂ emissions associated with the faulty test are larger than those corresponding to the normal test.

Figures 5 and 6 indicate the parameter $\% \Delta m_{C O_2}$ as a function of the tests, with instantaneous values calculated time-step by time-step and reported in descending order, for the summer and winter tests, respectively.

5.png

Figure 5. $\% \Delta m_{C O_2}$ calculated time-step by time-step in descending order as a function of the summer tests

6.png

Figure 6. $\% \Delta m_{C O_2}$ calculated time-step by time-step in descending order as a function of the winter tests

Specifically, with reference to the summer tests (Figure 5), positive values of $\% \Delta m_{\mathrm{CO}_2}$ are observed during a period of about 240 minutes (corresponding to 40% of the entire tests' durations), whatever the couple of tests under investigation is (except the case of SF4 vs. SN4); values of $\% \Delta m_{C O_2}$ close to zero can be recognized for about 40 minutes, while the remaining parts correspond to negative values of $\% \Delta m_{\mathrm{CO}_2}$. The maximum value of $\% \Delta m_{C O_2}$ is about 1358%, while the minimum corresponds to a value of about - 493%; the average value is equal to about 75% considering all the summer tests. With reference to the winter tests (Figure 6), positive values of $\% \Delta m_{C O_2}$ are recognized during a period of about 210 minutes (corresponding to 30% of the entire tests' durations), whatever the couple of tests under investigation is; values of $\% \Delta m_{C O_2}$ close to zero are obtained for about 20 minutes, while negative values of $\% \Delta m_{\mathrm{CO}_2}$ correspond to the remaining parts of the tests. The maximum value of $\% \Delta m_{\mathrm{CO}_2}$ is about 1302%, while the minimum corresponds to a value of about -90%; the average is equal to about 85% considering all the winter tests.

Table 2 reports the daily values of $\% \Delta m_{C O_2}$ as a function of the tests.

Table 2. Daily values of $\% \Delta m_{C O_2}$ as a function of the tests

The values in this table indicate that: (i) the fault 4 (i.e., exhaust air dumper kept always closed) is characterized by a relevant effect under both summer and winter conditions, with daily values of $\% \Delta m_{C O_2}$ corresponding to about 110% and 13%, respectively; the fault 1 (i.e., return air damper kept always closed) significantly impacts the CO₂ emissions during winter only, with a notable daily increment by about 14% in comparison to the normal case. The faults 2 and 5 correspond to negative daily values of $\% \Delta m_{C O_2}$ during both summer and winter; this means that they reduce the global equivalent CO₂ emissions with respect to the fault free tests, whatever the season is. However, the reduced environmental impact (due to a reduced electric energy consumption, as indicated in Figure 3) is compensated by a lower daily thermal/hygrometric comfort time during the summer period in the cases of both the faults 2 and 5 and during the winter period for the fault 5 only. The results associated to the occurrence of the fault 2 during winter can be explained by taking into account that the absence of outdoor air allows smaller fluctuations of indoor air humidity as well as more stable hygrometric conditions in addition, it can be noticed that a relevant reduction of humidifier operation time during the winter faulty test can be recognized with respect to the corresponding normal test. In the case of the fault 1, Table 2 indicates that, with respect to the normal test, the faulty operation is characterized by lower daily global equivalent CO₂ emissions during summer only, together with reduced electric energy demand (Figure 3) and negligible variations in terms of thermo/hygrometric comfort time; this result is because the utilization of hot outdoor air is more than counterbalanced by the significantly decreased supply airflow rate elaborated by the AHU.

5.3 Effects of faults on operating costs

In this paper, the operating costs associated to the electric energy consumption are evaluated. Figure 7 reports the unit cost of electricity for each couple of faulty and normal tests according to the Italian scenario [29]. Similar to the environmental analysis, no difference in terms of unit cost is assumed between the faulty test and the associated normal test.

The operating costs of AHU associated with the occurrence of the faults are compared with those corresponding to the fault free tests by using the following formula:

$\begin{gathered}\% \Delta O C=\frac{\left(O C_{{Faulty }}-O C_{ {Normal }}\right)}{O C_{ {Normal }}} \times 100 =\frac{\left(U C_{E E} \times E E_{ {Faulty }}-U C_{E E} \times E E_{ {Normal }}\right)}{U C_{E E} \times E E_{ {Normal }}} \times 100\end{gathered}$          (5)

where, OC_Faulty and OC_Normal are, respectively, the operating costs of AHU under faulty and normal conditions, EE_Faulty and EE_Normal are the electric energy consumptions during faulty and normal tests, respectively, and UC_EE is the unit cost of electricity purchased from the grid according to the values in Figure 7. A positive value of %DOC indicates that the operating costs associated with the faulty test are larger than those of the corresponding normal test.

7.png

Figure 7. Unit cost of electricity purchased from the grid as a function of both the tests and the time [29]

Figures 8 and 9 show the parameter %DOC as a function of the tests, with values calculated time-step by time-step and reported in descending order, for the summer and winter tests, respectively. Specifically, with reference to the summer tests (Figure 8), positive values are observed during a period of about 250 minutes (corresponding to 45% of the entire tests’ durations), whatever the couple of tests under investigation is (except the case of SF4 vs. SN4); values of %DOC close to zero can be recognized for about 40 minutes, while the remaining parts correspond to negative values of %DOC. The maximum value of %DOC is about 3094%, while the minimum corresponds to a value of about -83% as well as the average is equal to about 133 %. With reference to the winter tests (Figure 9), positive values of %DOC are recognized during a period of about 230 minutes (corresponding to 35% of the entire tests’ durations), whatever the couple of tests under investigation are; values of %DOC close to zero are obtained for about 130 minutes, while negative values of %DOC correspond to the remaining parts of the tests. The maximum value of %DOC is about 1302%, while the minimum corresponds to a value of about -90% as well as the average is equal to about 85%.

Table 3 indicates the daily values of %DOC as a function of the tests. The values in this table indicate that: (i) the fault 4 (i.e., exhaust air dumper kept always closed) is characterized by a relevant effect under both summer and winter conditions, with daily values of %DOC corresponding to 110% and 12%, respectively; the fault 1 (i.e., return air damper kept always closed) and the fault 3 (i.e., fresh air damper kept always open) significantly impact the daily operating costs during winter only, with notable daily increments by about 13% and 10%, respectively, in comparison to the normal cases.

8.png

Figure 8. Values of %DOC calculated time-step by time-step in descending order as a function of the summer tests

9.png

Figure 9. Values of %DOC calculated time-step by time-step in descending order as a function of the winter tests

Table 3. Daily values of %DOC as a function of the tests

The faults 2 and 5 correspond to negative daily values of %DOC during both summer and winter; this means that they reduce the operating costs with respect to the fault free tests, whatever the season is. However, the reduced economic impact (due to a reduced electric energy demand, as indicated in Figure 3) is balanced by a lower thermal/hygrometric comfort time during the summer period in the cases of both the faults 2 and 5 and during the winter period for the fault 5 only. The results associated with the occurrence of the fault 2 during winter can be explained by taking into account the absence of outdoor air causes reduced fluctuations in terms of indoor air humidity as well as more stable hygrometric conditions; in addition, it can be noticed that a relevant reduction of humidifier operation time during the winter faulty test can be recognized with respect to the corresponding normal test. In the case of the fault 1, Table 3 indicates that, with respect to the normal test, the faulty operation is characterized by lower operating costs during summer only, together with reduced electric energy demand (Figure 3) and negligible variations in terms of thermo/hygrometric comfort; this is due to the fact that the utilization of hot outdoor air is more than counterbalanced by the significantly decreased supply airflow rate elaborated by the AHU.

This study has been performed within the activities of the research project titled “UTMOST FDD: an aUToMated, Open, Scalable and Transparent Fault Detection and Diagnosis process for air-handling units based on a hybrid expert and artificial intelligence approach. From experimental open data to transfer model learning for the enhancement of energy management and indoor environmental quality in buildings” funded by the call “PRIN 22” (CUP: E53D23002840006).