Online Fault Detection Configuration on Equipment Side of a Variable-Air-Volume Air Handling Unit

YANG Xuebin(楊學(xué)賓), LI Xinhai(李鑫海), YANG Siyu(楊思鈺), WANG Ji(王 吉), LUO Wenjun(羅雯軍)

1 College of Environmental Science and Engineering, Donghua University, Shanghai 201620, China

2 Shanghai Division, China Ship Development and Design Center, Shanghai 201108, China

Abstract：With the development of the technology of the Internet of Things, more and more operational data can be collected from air conditioning systems. Unfortunately, the most of existing air conditioning controllers mainly provide controlling functions more than storing, processing or computing the measured data. This study develops an online fault detection configuration on the equipment side of air conditioning systems to realize these functions. Modbus communication is served to collect real-time operational data. The calculating programs are embedded to identify whether the measured signals exceed their limits or not, and to detect if sensor reading is frozen and other faults in relation to the operational performance are generated or not. The online fault detection configuration is tested on an actual variable-air-volume (VAV) air handling unit (AHU). The results show that the time ratio of fault detection exceeds 95.00%, which means that the configuration exhibits an acceptable fault detection effect.

Key words：fault detection; software configuration; online monitoring; equipment side; variable-air-volume (VAV); air handling unit (AHU)

Introduction

With the continuous development of Internet of Things and data science technologies, the measured data in air conditioning systems increase significantly. These are such new challenges that a traditional air conditioning system has no choice but to face seriously, especially for fault detection and diagnosis (FDD). The existing controllers in air conditioning systems, however, usually pay much more attention on the realization of controlling function rather than the roles of data storing, processing or calculating, that is, data analysis or computing capability seems to be weak as well as the function of FDD[1]. It is much essential to expand the equipment-side functions of storing the operational data, computing or mining this data, detecting the operational parameters, diagnosing the potential faults,etc.

For the existing controllers in real air conditioning systems, many scholars have developed many models of system operation monitoring and online FDD. Yangetal.[2-3]developed a fault detection strategy combing sliding windows with principle component analysis for air conditioning systems. Lietal.[4]utilized wavelet energy entropy for fault detection and diagnosis of sensors in air handling units.Wangetal.[5]integrated the control panel into a building management system to collect the operational data online for detecting the sensor fault of the centrifugal chiller. Xiaoetal.[6]developed an online fault diagnosis tool, which collected the operational data from the building automation system for online sensor health monitoring and fault diagnosis of the air handling unit (AHU). Wangetal.[7]developed an online fault diagnosis tool for pressure-independent air supply terminals of variable-air-volume (VAV) air conditioning systems. This tool was installed in building management controllers to read the operational data online through the local area network. Based on the limited data processing capacity of the building automation system, the above studies have expanded their functions to implement remote supervision and online fault diagnosis. In addition, many scholars have devoted themselves to acquiring the operational data and expanding the fault detection functions of air conditioning systems with the software LabVIEW. Yanetal.[8]used LabVIEW to collect the operational data of VAV air conditioning systems in the laboratory, to verify the effectiveness of the proposed sensor FDD method. Based on the stress wave theory, Wangetal.[9]developed an air conditioning pipeline leakage detection system which can locate the leakage source quickly and accurately with LabVIEW. Based on the fuzzy theory and the characteristics of automotive air conditioners, Wang[10]developed a new type of air conditioning automatic fault diagnosis system by using LabVIEW. Combining neural and fuzzy technologies, Zouarietal.[11]used LabVIEW and MATLAB to develop a real-time fault detection tool for centrifugal pumps.

Many researchers have conducted much work on extending the data processing functions of existing air conditioning controllers. They have laid a certain research foundation for the development of an online fault detection configuration. Many roles more than controlling functions, however, should be further explored or validated in real systems, especially on their equipment sides.

This study intends to explore the online fault detection functions on the equipment side of an existing air conditioning system. A fault detection methodology covers identifying whether the measured signals are within their certain limits, detecting whether the sensor reading is frozen and whether some faults possibly in relation to the operational performance occur. An online fault detection configuration developed on the equipment side comprises online data acquisition, measured signals exceeding limits identification, sensor reading frozen detection, and operational performance-related fault detection programs. And then, the configuration is tested on a real system and evaluated by the index named fault detected time radio.

1 Methodology

1.1 VAV air conditioning system

This section elaborates on the studied air conditioning system and measured variables. The VAV air conditioning system includes an AHU with four pressure-dependent air supply terminals. Outdoor fresh air enters the AHU and is mixed with indoor return air. The mixed air is sent to each air supply terminal by the fan through the cooling coil to control the temperature or humidity of the room. The total air volume is accumulated according to the air volume at each terminal and is applied as the set-point entitled the total required air supply volume[12]. The air supply volume is regulated by the motor input frequency of an air supply fan according to the proportional-integral-derivative control. The measured variables are shown in Table 1.

Table 1 Variables measured for online fault detection

1.2 Fault detection method

This section introduces the methodology of identification of the measured signals exceeding their limits, detection of sensor reading frozen and abnormal operational performance.

The measured data depends on whether they exceed the normal range. The upper and lower limits are assigned according to the system operational performance characteristics, the sensor range, the actuator feedback and other physical characteristics. If the measured value is within this range, the operating status is deemed normal conditions.

The measured values for the current minute and the previousn-1 time points are recorded to detect the fault of sensor reading frozen. The fault of sensor reading frozen is reported if these values do not show any change in a period.

Air handling unit performance assessment rules[13]have been employed as a fault detection tool based on mass and energy conservation. The method is used to identify the abnormal operational performance of air conditioning systems. Table 2 lists the expressions of expert rules and their explanations.

Table 2 Expressions of expert rules [13]

2 Calculation Procedures for Online Fault Detection

LabVIEW has wide applicability in data acquisition, process control, and data visualization. MATLAB has strong numerical value calculation and analytical abilities. By the means of combining these two development tools, an online fault detection configuration has been developed on the equipment side of a VAV air conditioning system. With this configuration, LabVIEW acquires the real-time operational data via modbus transmission control protocol (TCP) communication. Numerical computation is performed in MATLAB to determine whether the measured signals exceed normal limits, to detect frozen sensor readings, and to identify abnormal operational performance. The fault detection results are then displayed on the user interface by LabVIEW. The calculation procedures for online fault detection are divided into three steps, as can be seen in Fig. 1.

Fig. 1 Calculation procedures for online fault detection with the configuration

(1) The first step is data pre-processing. The missing data should be screened out and deleted before the fault detection computations to prevent or reduce false alarms. The order of measured variables is inconsistent with that of input variables in MATLAB calculation programs. And thus, the order of measured variables must be adjusted before the fault detection calculation.

(2) In the second step, the program includes measured signals exceeding limits detection and sensor reading frozen detection. The upper and lower limits are set based on the physical characteristics of air conditioning systems to identify whether measured signals are within normal limits. Ifε1≤Xy≤ε2, then the measured valueXyis normal,ε1is the lower limit, andε2is the upper limit. The measured variables for the current time and the previous 19 min are constantly monitored to determine if the sensor reading is frozen. IfX1=X2=…=Xn, thenXis frozen, elseXis normal.Xdenotes a sensor measurement.X1,X2,…,Xnare the measured values for the current time point and the previousn-1 time points.

(3) The third step is fault detection in relation to operational performance. The expert rules are mainly employed for the detection of potential failures related to operational performance in air conditioning systems. The basis for this method is a set of expert rules derived from mass and energy balances. If the expert rule is satisfied, then the fault related to the operational performance will be successfully detected.

3 Configuration Program Design

This section includes the design of online data acquisition program, measured signals exceeding limits identification program, sensor reading frozen detection program and operational performance-related fault detection program. There are two ways for the MATLAB program to be invoked for mathematical computation in LabVIEW[14]. The first way employs the ActiveX function, and the second one directly invokes MATLAB script nodes. The MATLAB script node is adopted because of its programming flexibility.

3.1 Online data acquisition program

The air temperature, humidity, pressure, fan frequency, and other operational parameters are collected in real time for mathematical computations related to fault detection. Modbus TCP[15]is employed to communicate the data between the software configuration and the existing air conditioning systems.

The existing controller implements the functionality for collecting the real-time operational data measured by sensors. In order to realize network communications between the existing controller and the software running in a computer, the "Network Connection" function on the computer control panel is used for setting the connection properties of the Ethernet interface, where the "Internet protocol (IP)" is selected and the IP address is set.

The LabVIEW program is designed for online data acquisition, as shown in Fig. 2. The reading holding register function is chosen from the NI Modbus library. LabVIEW TCP functions include TCP Open Connection, TCP Read, TCP Close Connection, and so on. These functions are applied for network communication. The data transmission port must be specified in TCP communication. The parameters of LabVIEW external interface are as follows. The IP address is set as 192.168.1.1, the TCP port number is set as 502, and the Modbus ID is assigned the value of 1. The sampling interval is defined as 3 s.

Fig. 2 Structure for online data acquisition and transmission designed in LabVIEW

3.2 Fault detection program

As shown in Fig. 3, the fault detection program consists of the measured signals exceeding limits and sensor reading frozen detection. The variableRinputis a two-dimensional array for the current minute and the previous 19 min, and is stored in the shift register of LabVIEW. This variable transferred to the MATLAB script identifies whether the measured signals exceed their limits or not, and detects sensor reading frozen. The variableRlimtis the output of the measurement within or without the normal range. The variableRfrzonis the output of frozen sensor reading. The variableRsigndenotes the frozen signals. The output 1 indicates that the sensor reading is frozen, while 0 indicates normal. The output format in the script nodes cannot be a string array. The variables,RlimtandRfrzon, are the output stored as strings, and then are post-processed to transform into string arrays.

The operational performance fault detection program is shown in Fig. 4. The variableRinputis a one-dimensional array at the current minute, and is transmitted to the MATLAB script nodes for fault detection related to operational performance. The variableRruleis a one-dimensional array in relation to the abnormal operational performance.

Fig. 3 Structure for measured signals exceeding limits and sensor reading frozen detection designed in LabVIEW

Fig. 4 Structure for online fault detection related to operational performance designed in LabVIEW

4 Field Tests

On the equipment side of a VAV air conditioning system, the configuration is tested at a sampling interval of 3 s. The fault detected time ratio (FDTR) is defined as the percentage of the detected fault time points accounting for the total time (sixty data sampling time points).

4.1 Measured signals exceeding limits

Three parameters include supply fan frequency,supply fan power, and valve control signal. The testing faults are set in an actual VAV air conditioning system, which are the motor frequency of a supply fan lower than the minimum value, the current of a supply fan exceeding the maximum value, and the valve opening position exceeding the maximum value. The online fault detection configuration is employed to test these faults. As illustrated in Table 3, all FDTRs exceed 95.00%. The configuration is able to determine whether the measured signals exceed normal limits or not.

Table 3 FDTRs of measured signals exceeding limit

4.2 Detection of sensor reading frozen

The detection results of sensor reading frozen are listed in Table 4. The four testing faults are conducted in the real VAV air conditioning system. The supply water temperature and return water temperature are frozen at 6.6 ℃. The supply air temperature and return air temperature are also frozen at 6.6 ℃. The measured data for the current time and the previous 19 min is stored in the form of an array by LabVIEW. The FDTRs of the four faults exceed 97.00%, and the configuration exhibits desirable detection accuracy for sensor reading frozen.

Table 4 FDTR of sensor reading frozen detection

4.3 Fault detection in relation to operational performance

The testing faults in relation to operational performance are the air supply volume lower than the total required air supply volume (fault A), the cooling coil water valve stuck at a small position (fault B), and the cooling coil water valve stuck at a large position (fault C). The results of them are shown in Table 5. The FDTRs of faults A, B, and C are 99.76%, 99.33%, and 97.24%, respectively, all exceeding 97.00%. For example, if the air supply volume and required volume differ by more than a threshold value, a fault related to operational performance will be reported. The interference from the changes in external cooling and heating loads, the measurement error of the sensor itself, and the instability of the air conditioning system may lead to the intermittent beating of the operational data, which leads to the FDTR being lower than 100%.

Table 5 Results of operational performance fault detection

5 Conclusions

An online fault detection configuration has been developed on the equipment side of a VAV air conditioning system. The proposed configuration includes four functions: operating data acquisition, identifying whether measured signals exceed limits or not, sensor reading frozen detection, and operational performance fault detection. The main conclusions are as follows. (1) The FDTRs of the measured signals exceeding normal limits, frozen sensor reading, and abnormal operational performance all exceed 95.00%, and the configuration exhibits acceptable fault detection effects. (2) More computational programs will be embedded in the configuration to further expand the fault diagnosis functions. In addition, this configuration can be introduced to other actual VAV air conditioning systems for more tests.