Keywords

1 Introduction

The energy consumed for cooling contributes the most to a typical building energy consumption in tropical region like Malaysia. Faults and performance degradation in the air conditioning and mechanical ventilation/heating, ventilation, and air conditioning (ACMV/HVAC) system lead to energy wastage. As the cost of ACMV operation is already high, any efficiency losses would increase the cost further. In addition, electricity used to power the system is mainly generated from non-renewable sources in Malaysia and Singapore, which contributes to greenhouse gases when burned. As a result, there is a need for fault detection and diagnosis to minimize the impacts when a fault occurs. Faults can occur at any of the sub-systems of the ACMV and due to its size and complexity, it is a challenge to diagnose them accurately at low severity levels.

The ACMV or HVAC is a system that controls the temperature, humidity, fresh air, filtration, and movement of air through space. In Malaysia region, heating is not required, thus the system is known as ACMV, while regions that require heating use HVAC. The goal of such a system is to provide thermal comfort and acceptable indoor air quality for the occupants or to provide cooling for equipment. The ACMV can achieve this by conditioning the air through [1,2,3,4,5]: cooling, dehumidifying, cleaning, and ventilating. All these can be done based on the principles of thermodynamics, fluid mechanics and heat transfer. There are three subsystem that works together to perform the functions mentioned above. These three sub-systems are cooling tower, chiller plant, and AHU. Figure 1 depicts the interdependency of the three sub-systems and how heat is being transferred around the ACMV system.

Fig. 1
A diagram of a central air conditioning system. The labeled components are the cooling tower, C W P, a chiller plant with condenser, evaporator, compressor, and expansion valve, C H W P, A H U 1 to n. An air-handling unit exits, returns air from rooms, takes outdoor air, cools it, and supplies it to rooms.

Main components of central air conditioning system

Full size image

The cooling tower removes the heat from the chiller plant’s condenser through heat exchangers filled with water. This water loop is called the condenser water loop, the water is then pumped from the condenser to the tower. The water is cooled by pumping it up to the top of the cooling tower where the ambient air and wind can remove the heat from the water. Heat is removed from the water as it is sprayed on to thin film materials. As the water flows from the top to the bottom of the cooling tower, it is exposed to the ambient air and wind which cools it down. The cooled condenser water is then circulated back to the chiller plant’s condenser to absorb heat again.

The chiller plant provides the cooling capacity for the building as the centralized cooling system. It provides cooling based on vapor compression which is the fundamental process in the refrigeration cycle. In the chiller plant, the compressor compresses the refrigerant which allows heat absorption at the evaporator and releases that heat at the condenser. The refrigerant circulates through an evaporator, compressor, condenser, and an expansion valve. A thermodynamic process at each component to ensure a continuous flow of cold refrigerant to absorb the heat from the chilled water loop. The cooled chilled water is then sent to the cooling coil for the AHU to cool the air.

The AHU is a system of ducts, filters, dampers, and fans to circulate cold fresh air and exhaust stale air. Fresh air is collected through an intake and cooled at the cooling coil as cold chilled water circulates the coil to absorb the heat from the air. The cooling coil dehumidifies the air as well by removing moisture in the air. The cold fresh air is then supplied to the various cooling zones in the building through the fans and ducts. Fans are used to aid in the circulation of conditioned air around the building as well as exhausting stale air through an outlet. Filters are used for cleaning and dampers regulate the amount of cold air entering each zone.

Cooling is provided through the removal of heat in spaces, the heat can be removed through radiation, convection, or conduction. The heat is transferred through a refrigeration system which uses a medium to absorb the heat and disperse it outdoors. The medium is known as the refrigerant, which can be water, air, or chemicals. This is possible by making use of a thermodynamic cycle called the refrigeration cycle. There is a compressor in the system that changes the pressure and state of the refrigerant to absorb or release heat.

Dehumidification of air drying is provided by the cooling coil as the temperature at the coil is below the dew point, thus moisture in the air will condense around the cooling coil. The moisture is then collected at the bottom of the coil and removed via piping.

Cleaning is done by air filtration at the AHU to remove particles, contaminants, vapors, and gases. This helps to remove air pollutants such as pollen and dust from the air, improving and regulating the air quality within the building. This is done by implementing filters of various materials in the air ducts, especially the outdoor air inlets to prevent pollutants from entering the building. This also prevents ducts or fan blades from dust accumulation.

Ventilation is normally done mechanically through the AHU via the multiple fans within the duct system. Conditioned outside air is delivered into the building by a supply air fan and existing air is being exhausted to prevent carbon dioxide build-up. Ventilation is necessary as it is one of the most important factors for maintaining indoor air quality within the building. Without ventilation, stale air can happen which has an unpleasant smell and can become a health hazard for the occupants at higher levels.

ACMV has become an essential system and integrated into our society, both in residential and industrial. The ACMV system contributes to more than 60% of the total building energy consumption. Within the ACMV system, the chillers and fans are the two components that contribute significantly to energy consumption. Chiller consumes up to 55% and fans up to 35% of the total energy consumption [2,3,4,5].

Therefore, early FDD is very important for ACMV’s health monitoring as well as optimal utilization of electrical energy, which is directly dependent on the optimal selection of the dataset and its severity level w.r.t. the fault condition. The main objective of this study is to analyze the severity and reliability of the dataset, which is used to detect and diagnose faults at low severity. By detecting faults at low severity, fixes, and maintenance can be done at a lower cost when compared with a more severe fault. This ensures the system operates optimally and reduces costs related to operation and maintenance. This in turn helps building owners to reduce electricity consumption and their carbon footprint.

The organization of this paper is comprised of seven sections. Section 1 represents the introduction. Brief information of AHU is represented in Sect. 2, which includes main components of the ACMV/HVAC and its related components. Section 3 represents brief information of AHU faults and its severity. The simulation of the model is shown in Sect. 4. Sections 5 and 6 represent the dataset collection and sensitivity analysis, respectively. Finally conclusion is represented in Sect. 7.

2 Brief Information of Air Handling Unit (AHU)

Figure 1 represents the main components of the central air conditioning (CAC) system, which includes mainly three sub-systems, i.e., cooling tower, chiller plant, and air handling unit (AHU).

The cooling tower is the point where the heat absorbed from the cooling zones are being dissipated into the atmosphere through the evaporative process. Warm condenser water is circulated to the top of the cooling tower, where the water is exposed and evaporated into the air passing through the cooling tower. As the water evaporates, the air absorbs heat and cools the remaining condenser water. The remaining condenser water will flow into a collection basin, where it can be pumped back to the condenser to absorb heat at the chiller plant. There are five main components to a cooling tower, exhaust fan, drift eliminators, water distribution, fill, and basin. The hot condenser water is sprayed from the water distribution using a nozzle, it then flows on to the fill section which aids in the evaporation rate. The cooled water is then collected at the cooling tower basin which is then circulated back to the condenser. The hot air is exhausted by the fan and drift eliminators prevent droplets of water from being blown out of the tower [6].

The chiller operates on the vapor compression or vapor absorption concept. The most frequently used air conditioning cycle is the vapor compression. The low-pressure refrigerant evaporates in the evaporator to produce a cooling effect. A compressor, which often requires electrical energy, compresses the generated refrigerant vapor to condenser pressure. The high-pressure refrigerant condenses in the condenser, and the generated heat is discharged into the atmosphere/cooling tower. The condensed refrigerant is then expanded to evaporator pressure by an expansion valve to continue the cycle. There are two types of chillers, namely air-cooled and water-cooled chillers. Water-cooled chillers can be further categorized as absorption, centrifugal, helical rotary, and scroll [7]. The chiller plant consists of the compressor, condenser, expansion valve, power unit, control unit, and water box. The compressor increases the pressure of refrigerant vapor from the evaporator to the condenser. All the components are located centrally in a chiller plant room.

AHU is a main component of the central air conditioning system of commercial buildings to maintain zone temperature in a multi-zone application [1]. There are many different types of AHU, such as (1) as per location, (2) as per fan position, (3) as per mixing capability, (4) as per energy recovery, (5) as per service zone, (6) as per ducting, and (7) as per air volume. In our air conditioning and mechanical ventilation system (ACMV) context, the AHU is a draw-through, multiple zones, single duct variable air volume (VAV) AHU with mixing capability for energy recovery. The AHU is located indoors.

The variable air volume (VAV) system of the draw-through AHU modulates airflow according to the building’s thermal load condition. Connected to the supply air ducts, the VAV box in each zone controls the temperature and amount of supply air necessary to maintain zone temperature in a multi-zone application [1]. A typical representation of VAV and CAV type systems is represented in Figs. 2 and 3, respectively. The main properties of the CAV system are: (1) chilled water flow to AHU coils is modulated based on return air temperature, (2) if the percentage change of cooling load of different rooms is not same, few rooms could be over-cooled or under-cooled. Similarly, the main properties of the VAV system are: (1) the opening of the VAV box damper is based on room temperature, (2) the speed of the fan is modulated based on duct pressure, (3) the chilled water valve is modulated based on the pre-set off-coil temperature of supply air, (4) suitable if the cooling load of different rooms is different.

Fig. 2
A diagram of a C A V. Return air from rooms goes to the return fan, and the air exits by exhaust. Some exhaust air and the outdoor air pass through a damper, followed by a filter, cooling coils, and supply fan, and go to the rooms. and the typical room and service area.

Constant air volume (CAV) system’s type AHU

Full size image
Fig. 3
A diagram of a V A V. Return air from rooms goes to the return fan and exits through the exhaust. Some exhaust air and the outdoor air pass through a damper, a filter, cooling coils, and a supply fan, and go to rooms and the typical room and service area via a V A V box. It also has a controller and V S D.

Variable air volume (VAV) system’s type AHU

Full size image

Measurements are recorded from various points of interest within the AHU through the sensors installed. They include pressure, flow rate, temperature sensors, etc. The AHU’s air handling controller (AHC) receives these readings and sends corresponding signals to the cooling coil, fans, and dampers for regulation to achieve the designated supply air temperature and operating conditions—such as the indoor air quality (IAQ) standards. In Table 1, the functionality of each component is described.

Table 1 Summary of different functionality and location of AHU components
Full size table

Controlled automatically by the AHC, AHU dampers operate to conduct appropriate mixing of the air (outdoor and recirculated return air) to conserve energy by limiting heating or cooling and meeting the AIQ standards [8].

$$ \theta_{{{\text{outdoor}}}} = 1 - \theta_{{{\text{return}}}} = \theta_{{{\text{exhaust}}}} $$
(1)

where \(\theta_{{{\text{outdoor}}}}\) = angle of outdoor air damper, \(\theta_{{{\text{return}}}}\) = angle of return air damper, and \(\theta_{{{\text{exhaust}}}}\) = angle of exhaust air damper.

In the AHU, the control of mechanical cooling and heating is sequentially programmed to synchronize with the damper control [9]. It's worth noting that when the amount of outdoor air mixed is kept to a minimum, the outdoor air damper position is at its lowest, while the return air damper position is at its highest, as shown in Eq. (1).

During the heating mode, the heating coil valve is regulated for hot water supply from a boiler to maintain the designated supply air temperature, while the cooling coil valve is closed and the dampers are modulated to keep the amount of outdoor air being mixed to a minimum while meeting minimal ventilation requirements [9].

As the outdoor air temperature increases, the sequential operation switches from heating to free cooling, cooling the building with outdoor air. The air dampers modulate in this mode to maintain the appropriate supply of air temperature [9].

As the outdoor air temperature further increases, the AHU will be unable to cool the supply air to the desired temperature using only the outside air, triggering the cooling coil valve to carry out mechanical cooling [9].

The fans located at the supply and return air ducts are equipped with a variable frequency drive (VFD) which modulates its speed, influencing the airflow within the AHU as load condition varies. The purpose of each VFD however, differs. The supply fan is modulated with the purpose of maintaining duct static pressure, while the return fan is modulated with the purpose of maintaining the airflow difference between the supply and return fans. The control signal sent to the VFD of the supply fan is based upon the duct static pressure read from the pressure sensor at the supply air duct (P1), while the control signal sent to the VFD of the return fan is based upon the airflow reading from flow sensors (F1 and F2) located at the supply and return air duct, respectively [8].

It is good to note that all AHU components complement each other to achieve the operating conditions dictated by the AHC [1]. In the context of this study, the heating coil is excluded due to its inapplicability in tropical climates like Malaysia.

3 Brief Information of AHU Faults

Faults that occur in AHU can lead to inefficient energy consumption, occupant discomfort, and safety issues. Thus, it is necessary to utilize FDD tools to locate faults automatically. According to Bruton et al., implementing FDD technologies resulted in a total savings of €104,000 per year in 18 AHUs across four testing sites [10]. The faults that occur in AHU are also listed in the NREL project “An Open-Based Platform for Whole-Building Fault Detection and Diagnostics.” According to the NREL, the faults are ranked by the amount of energy consumed, and duct leakage was identified as a major fault because it has the second-highest energy impact [11].

The IEA Annex 25 provides a list of faults for the air mixing section, filter-coil section, fan section, and VAV boxes that are listed without regard to severity or frequency of occurrence. A preliminary survey among designers, construction engineers, and commissioning engineers was conducted in Japan in 1992 to determine the importance of various faults in AHU. Before assessing whether or not a fault is critical, there are seven factors to consider which are environmental degradation and occupant complaints, increased energy consumption, serious secondary damage, frequent occurrence, difficult detection, lengthy repair time, and costly repair [12].

A list of AHU faults from the NREL, ASHRAE RP 1312, and IEA is shown in Fig. 4. The most impactful faults resulting in increased energy consumption are included in Table 2, which was compiled from the ASHRAE and NREL reports.

Fig. 4
A chart classifies A H U faults into 4 categories. The categories are components with 6 types of faults, sensors with 6 types of faults, actuators with 4 types of faults, and controllers with 3 types of faults.

List of all possible faults in AHU

Full size image
Table 2 Summary of AHU faults based on NREL, IEA, and ASHRAE RP-1312 [11,12,13]
Full size table

Since these ten faults impact energy consumption and indoor thermal comfort, it is critical to conduct FDD so that the ACMV system will work efficiently with minimal operating costs and a high level of comfort for the occupants.

4 Simulation of ACMV Model

OpenStudio (OS) is an open source software development kit (SDK) for building energy modeling (BEM) that intends to significantly reduce the time and effort necessary to develop and maintain BEM-based applications, as well as to extend and support the ecosystem of end-user BEM tools and services [14]. Building fault models in OS can be used to quantify the impact of faults on building performance as well as create and test FDD techniques [15].

EnergyPlus (EP) inputs and outputs are presented in OS as a dynamic, object-oriented data model with an application programming interface (API), which is a much easier construct for programs to deal with [14]. The OS Platform includes a range of GUI programs that aid in the construction of mechanical systems, simulation job, and workflow management on local workstations and high-performance computing (HPC) clusters and result visualization [16].

The US Department of Energy (DOE) small office together with the tropical weather file is used as the reference base model. The small office reference building is a single-story building with five zones plus an attic. Certain parameters such as the scheduling of occupants will be modified to represent the typical office setting in Singapore more accurately. The US DOE contributes to the development of commercial and residential building energy codes and standards by taking part in industry review and update processes and providing technical assessments to support both published model codes and proposed modifications. DOE discloses its findings in order to guarantee that its support is transparent and that its analysis is available for public examination and use [17].

OS application facilitates the design, simulation, and review of specific building energy models by building the model. Although, not all tabs are required for modeling a building, they are generally placed in the sequence in which they are utilized along the left-hand side of the window. The tabs are listed in Table 3 [60], along with a description of their functions. The maximum outdoor air flow rate was calculated as shown in Eq. (2).

$$ {\text{Flow Rate}}_{{{\text{OutdoorAir}}}} = {\text{Ventilation Rate}} \times {\text{Total Floor Area}} $$
(2)
Table 3 Representation of OS tabs and description
Full size table
Fig. 5
A diagram presents the loop for the A H U. It has 2 input points, 2 resistors, and a cooling coil in the supply equipment section. The demand equipment section has 6 zones and a library.

Simulated AHU loop

Full size image
Fig. 6
A list and 3 pie charts. The list has 20 options. An annual overview is selected. Pie chart 1 presents the end use with the highest value for cooling. Pie chart 2 presents energy use as the whole share of electricity. Pie chart 3 presents E U I with the highest value for cooling.

Overview representation of the annual report of electrical consumption

Full size image

Figure 7 depicts the simulated operation on the base model with three parameters of interest, namely the outdoor air flow fraction, outdoor air temperature (system node 27) and the various zone temperatures. It is good to note Eq. (3). An increase in the amount (Q) of return air being mixed is equivalent to a decrease in the amount of outdoor air being mixed within the mixing chamber.

$$ Q_{{\text{Supply Air}}} = Q_{{\text{Mix Air}}} = Q_{{\text{Outdoor Air}}} + Q_{{\text{Return Air}}} $$
(3)
Fig. 7
2 line graphs. 1. C versus days has a curve that falls, rises, and falls. It also has 6 curves that rise, descend, grow horizontally, rise, and rise slightly. 2. A H U run period timestep versus days has a curve that ascends, falls vertically, fluctuates horizontally, rises vertically, and ascends.

Representation of outdoor air flow trends

Full size image

The designated zone temperatures are set at 23 ℃ from 18:00 h to 06:00 h and 25 ℃ otherwise. At 06:00 h, the ACMV strives to cool the zones to the designated 23 ℃, mixing more return air to save energy while conducting mechanical cooling, resulting in a tremendous drop in the outdoor air flow fraction. From 06:00 h to 18:00 h, the various fluctuations are a result of the increase or decrease in infiltration levels (Fig. 8) and occupancy (Fig. 9). It is good to note that time is required for the zone to reach the designated temperature due to the thermal time constant, resulting in small modulations between the said time frame. For instance, from 13:00 h to 17:00 h, it can be seen that when the infiltration and occupancy levels remain constant, there were no modulations seen in the outdoor airflow fraction. After 18:00 h, the designated zone temperature changes from 23 to 25 ℃. With the zones previously cooled to 23 ℃, more outdoor air is introduced into the system for mixing, resulting in a spike in outdoor air flow fraction.

Fig. 8
A line graph of infiltration level versus time. It has a curve that grows horizontally, rises vertically to a small distance, grows horizontally, rises vertically, grows horizontally, falls vertically, grows horizontally, followed by a repeating pattern, a vertical fall, and a horizontal growth.

Representation of door opening (infiltration) schedule

Full size image
Fig. 9
A line graph of occupancy versus time. It has a curve that grows horizontally, ascends in a step up manner, rises vertically, grows horizontally, falls vertically, grows horizontally, rises vertically, grows horizontally, falls vertically, and descends in a step-down manner.

Representation of occupancy schedule of building

Full size image

It is observed that from 20:00 h to 22:00 h, there was a linear increase in outdoor air fraction, suggesting more outdoor air being mixed. This is attributed to a drop in light scheduling seen in Fig. 10 at 20:00 h. With lesser lights present, the zones become relatively cooler compared with when more lights were on. In an attempt to maintain zone temperatures at 25 ℃, more outdoor air is used for mixing resulting in an increase in outdoor air flow fraction.

Fig. 10
A line graph of building lights versus time. It has a curve that grows horizontally, ascends in a step up manner, rises vertically, grows horizontally, falls vertically, grows horizontally, rises vertically, grows horizontally, falls vertically, and descends in a step-down manner.

Representation of building lights schedule

Full size image

5 Dataset Collection for Further Study

For the purpose of sensitivity analysis and result demonstration, three different types of datasets have been collected such as (1) simulation datasets from our OS fault simulation model, (2) real datasets from the ASHRAE RP-1043 research project by Comstock et al. [18], and (3) real datasets from the Inventory of Data Sets for AFDD Evaluation by Jessica Granderson and Guanjing Lin [19].

The dataset from ASHRAE RP-1043 contains data measured and calculated based on a 90-ton centrifugal chiller. The experiment was designed with the specifications of air conditioning and refrigeration institute (ARI) Standard 550 for centrifugal and rotary screw water-chilling packages.

The dataset from the inventory contains performance dataset generated through simulation of physical experimentation. Both simulation and physical experimentation implement a VAV AHU for generating condition-based datasets. There are other differences between the dataset as well such as the total number of data points available for the parameters. The table below summarizes the difference between the two datasets, as given in Table 4.

Table 4 Representation of differences between simulation and real datasets
Full size table

It is good to note that the simulation and real data is vastly different. The real data was extracted from a heating, ventilation, and air conditioning (HVAC) system, which would suggest that both heating and cooling operations are involved. The simulated data, however, only operates in the latter operation.

6 Sensitivity Analysis and Result Demonstration

Data generation is done across the four faults once the base model is made. The data is used for data correlation analysis to determine any possible connection between the parameters and the fault. All the parameters related to the system, namely air system were enabled for the analysis (in this study), followed by the model simulation. Fault propagation is captured through the enabling of parameters that are not explicitly within the system. For example, the data is generated with the consideration of the fault severity and the range of load. The range of load used is 0, 50, 100, 150, and 200% of the full load and the fault severity ranges from 0% (used as base value), 10, 20, 30, and 50% of the fault level. The collection of data is done for all the load variations and the fault severity. The collected dataset points are: (1) for cooling coil (time, cooling coil total cooling energy, cooling coil source side heat transfer energy, cooling coil sensible cooling energy, cooling coil total cooling rate, cooling coil sensible cooling rate, cooling coil wetted area fraction), (2) for the fan (fan electricity rate, fan rise in air temperature, fan heat gain to air, fan air mass flow rate), and (3) air system (air system outdoor air heat recovery bypass minimum outdoor air mixed air temperature, air system outdoor air flow fraction, air system mixed air mass flow rate, air system outdoor air mechanical ventilation requested mass flow rate).

Once the data is extracted, parameters with changes according to the severity of fault and load variations are documented and arranged by its sensitivity to the varying fault levels. Sensitivity analysis is used to prioritize parameters for data dimension reduction based on the percentage difference between the average values of the parameters from the base value. The relative change formula is applied to calculate the percentage difference between the severed parameter and the base parameter, where the final value is the average fault severed value and the initial value is the average base value for the parameter. For the case study of fan fault conditions, Table 5 gives how Eq. (4) is applied to the data, and Table 6 gives the compiled parameters with its respective sensitivity analysis.

$$ {\text{Relative Changes}} = \left( {\frac{{{\text{Final value}} - {\text{Initial value}}}}{{\text{Initial value}}}} \right) \times 100 $$
(4)
Table 5 Representation of relative change formula applied for fan measurement
Full size table
Table 6 Representation of ranked parameters based on the percentage difference for fan measurement
Full size table

Similarly, for the case study of sir system fault conditions, Table 7 gives how Eq. (4) is applied to the data, and Table 8 gives the compiled parameters with its respective sensitivity analysis.

Table 7 Representation of relative change formula applied for air system measurement
Full size table
Table 8 Representation of ranked parameters based on the percentage difference for air system measurement
Full size table

This is done to rank and identify the most sensitive parameters which will be the input for artificial intelligence (AI) and machine learning (ML)-based diagnostic models [20,21,22,23,24,25,26,27,28,29,30]. This is important as it simplifies and improves the accuracy of data processing. For the real data, a similar method is done for its data correlation and sensitivity. From the real data, sensitivity analysis is done by collating the average values of the parameters, and Eq. 2 is applied to calculate the percentage difference between the fault severities and the base. The ranking is then done to identify the most sensitive parameters to input for AI/ML. The selected sensitive parameters (top 10) from the simulation data and real data are represented in Table 9.

Table 9 Representation of sensitive parameters (selected)
Full size table

Based on Table 9, the chosen sensitive parameters are different from the simulation parameters due to the availability of the data as not all parameters are being monitored for the real system as well as the difference in system, HVAC for real data, and ACMV for simulation. However, there are overlapping parameters that are directly or indirectly related to each other. For example, in the AHU system, real data contains CE which is related to CCSCR in the simulation. Thus, the sensitivity analysis for simulation can be applied to a real system if the data or related data are available.

It can be noted that the real data did not contain parameter data from other sub-systems but solely on their respective sub-system, such as cooling tower data for chiller faults and pump data for AHU faults. In the simulation, Chiller and AHU faults occurring have a significant impact on the cooling tower and pump, respectively, as given in Table 9. The overall summary of effect of parameters due to fault conditions in AHU is represented in Table 10.

Table 10 Representation of effect of parameters due to fault condition in AHU
Full size table

7 Conclusion

In this chapter, data reliability analysis for early fault diagnosis of AHU is performed by using two categories of datasets (category-1 dataset: emulated data, and category-2 dataset: real-data measurement). EnergyPlus+-based OpenStudio software platform is used to simulate the building envelope to generate the emulated dataset under different operating scenarios. In the simulation, building operation is fixed as per the working hours of the building’s user, and a dataset is generated for different fault scenarios starting from healthy condition to the most severe fault condition. The fault severity is considered from 0 to 30% w.r.t. healthy situation, and different electrical and non-electrical signals were recorded for further analysis. Based on sensitivity analysis of different signals w.r.t. the fault severity levels, the most sensitive signals have been selected, which may be used for the development of the classifier/identifier of the faults at an early stage. For the real-recorded dataset, the chosen sensitive parameters are different from the simulation parameters due to the availability of the data as not all parameters are being monitored for the real system, as well as the difference in system, HVAC for real data, and ACMV for simulation. However, there are overlapping parameters that are directly or indirectly related to each other. In the AHU system, real data contains CE which is related to CCSCR in the simulation. Thus, the sensitivity analysis for simulation can be applied to a real system if the data or related data are available.