Bringing Automated Fault Detection and Diagnostics Tools for HVAC&R Into the Mainstream

Heating, ventilation, and air-conditioning (HVAC) systems consume over 5 quads of energy annually, representing 30% of energy consumption in the U.S. commercial buildings [1]. Packaged rooftop units (RTUs) provide heating and cooling for over 60 percent of the commercial building space (about 90 billion ft²) in the U.S., and they are a significant source of energy consumption and peak demand. It is estimated that 40,000 10-ton RTUs are sold each year in the U.S. There are over 486,000 RTUs in the Northeast region and about 2700 units sold in 2014 in Connecticut. Another important market segment is commercial refrigeration. It contributes about 1 quad to commercial building energy consumption [2].

Most RTUs have one or more fault types such as low/high refrigerant charge, valve leakage, condenser/evaporator fouling, filter/dryer restriction, economizer, and controls. These types of faults can increase RTU energy consumption. If these faults are detected, diagnosed, and addressed, then significant energy could be saved. Automated Fault Detection and Diagnostics (AFDD) tools have the potential to save considerable energy for existing commercial RTUs. These devices, when installed on RTUs and other HVAC systems, can be used for retro commissioning. In addition, when faults are addressed, AFDD tools can be used for continuous commissioning. According to Navigant Consulting [2], national energy savings of 111 TBtu can be saved by employing fault detection and diagnostics (FDD) tools for commercial RTUs. Faults in commercial refrigeration systems not only result in significant energy use, but also can lead to shutdown of equipment that can cause damage to refrigerated products. Goetzler et al. [1] indicated that a popular supermarket employed a leakage detection and energy usage monitoring system that reduced electricity use by 23 million kWh per year.

To address this need, countless efforts to develop FDD tools have been ongoing for at least three decades [3–8], and now, there are several commercial FDD products marketed by companies. These products include field-portable FDD tools, factory-installed on-board FDD tools, hardware-based retrofit on-board FDD tools, and software-as-a-service (SaaS) FDD tools. However, FDD systems have not yet achieved a high HVAC market penetration. In an online article, “FDD Going Mainstream? Whose Fault is it?,” and several issues with FDD implementation are listed, such as lack of data, rules specific to systems, how to handle the FDD information, using the diagnostic data, prognostics data, and alternative ways to deploy FDD [9]. A key market barrier for this technology appears to be the lack of awareness of FDD products among building owners, who are the majority of potential customers. Most HVAC contractors are not familiar with the latest FDD technologies, and HVAC technicians lack knowledge and skills regarding these technologies. Quantifying potential benefits to building owners is difficult. Moreover, many building owners look for short-term return on investment (ROI) for their investments. Many of the hard to quantify FDD benefits, such as reduced HVAC downtime due to early warning and repairs, avoidance of catastrophic failures, and predictable maintenance, will not be so evident to the building owners. FDD technology appears to be like an “unknown saint.” It has promise but is not fully understood, evidenced, or exploited. To make matters more complicated, many companies package FDD tools with energy-efficiency retrofit kits, and it is not clear to customers which ones to use. Even though there are efforts to develop standards, currently there are no standards and methods to define functions, capabilities, accuracy, and reliability of FDD tools. Many potential customers do not have the tools to evaluate the cost of FDD tools or to assess real value and understand the value proposition. In addition, it is not clear how the process of FDD installation and the communication of faults, severity of faults, and actions occur during use. It is quite clear that addressing these issues and overcoming these barriers are challenging, but necessary. In that regard, there is a continued push for bringing FDD products to the commercial market. Western HVAC Performance Alliance has been working on an FDD Road Map and released a Master List of existing FDD products (over 100) [10]. Some efforts by utilities, federal agencies, and state agencies are underway to bring attention to FDD tools. For instance, California Energy Commission's Title 24-Part 6 requires that economizer FDD tools be installed on air-cooled unitary air-conditioning systems over 4.5 tons cooling capacity, with the ability to detect faults.

The uniqueness of this study approach is its comprehensive nature. It includes the identification of diverse FDD products, selection of different commercial building types for the field demonstrations, process evaluation, performance verification, determination of technical, and economic viability, supporting the development of utility incentives, education, and outreach. Identification of technical and market barriers, and development of strategies to address them by bringing together all stakeholders are critical aspects of the study. Regarding establishing economic viability of AFDD tools, one of the aims of the study is to determine the energy savings due to AFDD implementation in the field under naturally occurring fault conditions rather than introducing artificial faults. For achieving this, AFDD tools are installed on RTUs in the field and the energy use is monitored during the baseline period. At the end of the baseline period, faults identified by the AFDD tools are verified and any repairs and adjustments needed are implemented as a process of retrofitting. The energy use following the baseline testing is monitored (considered a part of continuous commissioning), and the energy savings due to AFDD tools is calculated considering weather normalization. Since there might be a significant variability in naturally occurring faults for different sites, and there are differences in features of AFDD tools, there is no direct comparison made among different AFDD tools regarding their performance. The AFDD tool capabilities and results are presented independently as individual case studies. This paper presents the first phase of the project from AFDD tool selection through initial baseline testing results. The results of phase II (post-retrofit testing), which will conclude in August 2021, will be presented in a future paper.

The goal was to select ten different commercially available FDD tools for the field demonstration study; therefore, an FDD tool selection matrix was developed. The matrix included various performance parameters of FDD such as the tool type (whether hardware-based or SaaS-based), type and range of faults detected, capability to detect heating and cooling faults, feature to diagnose and verify faults, limitation on the size of HVAC system, fault communication and frequency, skill level needed to use the tool, and FDD tool pricing. FDD vendors received a Request for Information (RFI) that included the questions from the selection matrix. The study received responses from 12 vendors. Based upon the responses, each tool received a normalized score for each performance metric. Table 1 shows the scoring results. Of these twelve FDD tools, one tool is not market ready and one FDD tool vendor dropped out from the study. Therefore, the study considered the remaining ten FDD tools for field-testing and evaluation.

The project's target was ten sites for the field demonstration of the FDD tools, with the goal of matching one tool per site. The project team developed the following FDD installation criteria: (1) commercial or industrial building type, (2) ease of access to rooftop units and corresponding ductwork, (3) RTU age, (4) a diverse portfolio, or building set, to demonstrate a variety of cases to potential customers, and (5) the ability to install a varied set of software versus hardware FDD tools across the ten sites. Table 2 shows all criteria details. Connecticut utility partners, United Illuminating, and Eversource sent a memorandum of understanding (MOU) along with the project scope to several of their customers for the purposes of developing candidate sites. The research team conducted several site visits after receiving MOUs. During these visits, the team-validated responses submitted as part of the MOU process and collected additional information for site criteria, such as ease of access to equipment. A set of ten sites were selected from 15 candidate sites, and each site was paired with one of the ten FDD tools, as shown in Table 3.

The purpose of the field demonstration is to study and verify the technical feasibility, economic feasibility, ease of installation, operational impacts, and environmental impacts of FDD through the field installation of FDD tools and independent monitoring. Section 3.1 describes the background and basis for the evaluation methodology. Sections 3.2 through 3.5 describe the field-testing methodology.

Braun and Yuill [11,12] developed methods to evaluate and assess the performance of FDD protocols for air-conditioning devices (RTUs and split systems) using a fault intensity (FI) and fault impact ratio (FIR). The study demonstrated that their developed FDD evaluator can be used to evaluate the strength and weakness of the FDD protocol by implementing the Refrigerant Charge and Airflow (RCA) diagnostics protocol and normal model approach (developed by a non-faulted set of a data). Five scenarios of test outcomes are presented in the FDD protocol (no response, correct, false alarm, misdiagnosis, and missed detection) for six types of faults (refrigerant charge, low-side and high-side heat transfer faults, liquid line restriction, non-condensable gas in the refrigerant, and compressor valve leakage). The authors used the approach of feeding the FDD protocol with a different set of experimental input data under different conditions and then observing the responses. They used the normal model set of data and four versions of the RCA protocol as a case study to demonstrate the use of their developed evaluation methodology. The RCA protocol was found to perform poorly using their evaluation methodology, resulting in up to 51% of faults where no faults were presented, 26% misdiagnosing faults, and 32% not detecting faults where the faults were presented.

Wall and Guo [13] studied the next-generation automated fault detection and diagnostics (AFDD) tools for commercial building energy efficiency. The results of the study reported benefits and outcomes of AFDD and performance evaluation of AFDD. The authors studied six different AFDD tools at six different buildings located in Australia for different space types (office, airport, museum, hospital, and laboratory). They connected the AFDD tools to a building management system (BMS) to obtain real-time data, and the results for each building were reported in terms of energy reduction in annual energy savings, National Australian Built Environment Rating System (NABERS) ratings, improved occupant comfort, avoidable energy, maintenance issues, and site energy intensity. The study also showed the top issues identified in each building.

Dexter and Pakanen [14] studied demonstrating AFDD methods in operating buildings in twelve countries (Belgium, China, Canada, Finland, France, Germany, Japan, Netherlands, Sweden, Switzerland, United Kingdom, and the United States of America) to develop different computer-based demonstration systems, which can interface with the HVAC system. The authors developed 23 prototype performance-monitoring tools and three performance validation tools based on artificial intelligence techniques (neural networks, fuzzy logic, expert and rule-based system, case-based reasoning, bond graphs, and qualitative methods). The authors studied 26 different FDD tools and approaches for different HVAC systems in different buildings for different space types (factory, hotel, lab, office, research and development center, college, school, and indoor swimming pool). The report defines three main groups of faults (natural fault, artificial fault, and simulated fault) that are considered in the study.

Jagpal [15] conducted a new study based on the results obtained in a previous study [14] for demonstrating AFDD methods in operating buildings to illustrate the benefit of a computer-aided FDD system by working with FDD vendors, contractors, customers, and end users. The report considered three main objectives: (1) assessing the cost-effectiveness of FDD methods, (2) building a prototype computer-aided performance evaluation system that has the ability to detect the performance deficiency and diagnose faults, and (3) investigating the feasibility of developed performance evaluation systems by testing them in operating buildings. The author identified that the availability of measured data and faults that have similar symptoms are the main barriers to identifying the HVAC faults.

Most of the approaches used in the previous study cases and literature are based upon laboratory experiments where authors introduced faults to the system to study fault behavior and to see if the FDD protocol is responding to the fault. However, it is difficult in this study to convince building managers to agree to introduce failures and faults into real-time operating systems currently servicing operating buildings. Therefore, the study team developed the monitoring and validation system and implemented it in parallel with the FDD tool to obtain real data and to evaluate the faults indicated by the FDD products based on FI and FIR of each fault. The study team developed a measurement and verification (M&V) plan based on the rules and techniques illustrated in previous research [11,12,16–18] to show the feasibility of market-available FDD tools by testing them in a set of real-time operating commercial and industrial buildings.

One key objective of the study is to undertake a comprehensive evaluation of FDD tools in the field, which is a complex task that requires examination of the FDD tool performance, cost, ease of implementation, ease of use, data requirements, training requirements, and applicability to the needs of a particular site or customer [19]. A framework proposed by Frank, et al. [20] for FDD performance evaluation was reviewed carefully. The framework is found to be useful for evaluating FDD protocols, but not for evaluating FDD tools in the field under naturally occurring faults, which can vary from no-fault to single faults of varying intensity and frequency, to multiple faults. Therefore, there is no control over some aspects of input samples for the FDD tools. However, the framework principles are used in the FDD tool performance verification. For instance, three categories of faults will be considered—(1) condition-based, (2) outcome-based, and (3) behavior-based. Also, ground-truth is employed in the verification of faults by FDD. Ground-truth is defined by a combination of independent monitoring of the RTU with essential sensors and instrumentation, and in-the-field checking and observation of the system, its operation, and controls.

The study team developed a measurement and validation plan based on methodologies developed by Braun and Yuill [11,12] and Mehrabi and Yuill [16–18]. The plan is also based upon the framework developed by Frank et al. [20]. The plan considers faults of an RTU and uses FI and FIR measures to identify them and to calculate their energy impact. The identification and impact of faults associated with the following subsystems and flows (energy and refrigerant) are considered: (1) refrigerant charge, (2) refrigerant quality, (3) economizer, (4) controls, (5) condenser, (6) evaporator, (7) expansion device, (8) compressor, (9) refrigerant-liquid lines, and (10) sensors as well as the RTU system level effectiveness of (11) energy performance degradation. Based upon possible faults occurring in the subsystems listed earlier and the fault impact calculations employed by the methodology in Sec. 3.4, the team determined the required components and sensors needed for collecting input data. The Independent Monitoring System (IMS) evaluator tool uses sensors and meters to collect real-time data as shown in Table 4. The table presents the specifications and planned locations of monitoring system sensors and instruments. These sensors are capable of measuring both refrigeration-side and air side performance. In addition, the monitoring system is capable of calculating airside and refrigerant-side performance indicators such as energy-efficiency ratio (EER) and cooling capacity. Complete specification of all sensors in the FDD evaluator sensor network is shown in the  Appendix.

A comprehensive IMS evaluator tool comprising a remotely accessible data acquisition system along with essential sensors and instrumentation has been designed and used in the verification process. IMS is literally a laboratory installed in the field on an RTU with over 20 sensors. One-minute interval data is stored and can then be downloaded remotely through a cellular modem. This prevents the project team from having to return to the site to collect data. A schematic of this Intellilogger data acquisition system is shown in Fig. 1.

The location of each sensor was determined based on the required input data collection needed. Based on the initial site visit, the team was able to customize the required sensors and instrumentation according to the RTU characteristics and the location of supply and return ducts for the building. A schematic of an RTU with locations of all sensors is shown in Fig. 2.

A majority of the FDD tools require a BMS, a building automation system (BAS), or energy management system (EMS) to be installed to read the data points within an RTU. A BMS is an intelligent system that controls and monitors a building's main technical components, such as the HVAC and lighting. Since BMS-based FDD tools can monitor all the RTUs on a building without additional hardware expense, the FDD tool will be setup to monitor all RTUs on a building. However, the independent monitoring system (IMS) is installed only on one RTU for validation. For non-BMS, hardware-based FDD tools, one FDD tool will be installed along with the IMS on one RTU per building. In general, all FDD tools share the same four general installation and use processes: (1) receive and install the tool, (2) verify functionality of the tool, (3) allow the tool to run over baseline period, (4) review the faults after the baseline period, and (5) implement repairs or controls changes to address faults. Many FDD tools require additional tuning processes for more accurately reading and detecting faults.

The team developed an M&V plan using an independent monitoring tool. The plan outlines the steps for the tool pre-installation, installation, data collection, analysis, and post-processing. For pre-installation, the team visited each study site to collect information and field measurements. These included RTU make/model, duct dimensions, determining pressure transducer/temperature sensor locations, and determining RTU accessibility. The team then used the field information to customize each monitoring system to the site. After ordering and receiving the sensors and instrumentations, the team pre-wired and calibrated the monitoring system and verified operation before field installation. A licensed HVAC technician installed all monitoring tools in the field with support from the project team. The HVAC technicians spent 1–2 days installing each monitoring tool in the field, which varied depending upon unit accessibility and a technician's skill and experience level. Once an FDD tool and the IMS were installed, the RTU was monitored for six to eight weeks and the data were collected based upon naturally occurring faults. The research team reviewed and analyzed the real-time data from the IMS weekly.

A complete set of fault algorithms has been implemented as part of the verification of FDD tools, based upon a previous work by Yuill and Braun [12] and the Pacific Northwest National Laboratory (PNNL) [7]. Additional calculations from the ASHRAE Handbook of Fundamentals [21] have been used during this development, as well as an online thermophysical property tool from National Institute of Standards and Technology (NIST) [22]. This section provides the methodology for identifying RTU faults using the monitoring system by calculating the FI, as well as examining the impact of these faults on the energy efficiency and energy consumption through the FIR. The methods that will be described are selected based on practicality as well as time constraints of the field study. The FIR can be defined as the ratio between the faulted and non-faulted values of the metric of interest. If the FIR passes a certain threshold determined by the user, then a system is considered “faulted.” The FI can be defined in terms of measurable numeric quantities related to the condition of the fault [12]. This approach can also be used to sort the prevalence of a fault. By setting a higher threshold, fewer alarms are detected. All calculations for this study are for constant-speed equipment rather than variable speed. The coefficient of performance (COP) is calculated based upon real-time input data from the monitoring system for purposes of determining FDD tool performance and calculating energy savings.

Control Faults: Control faults are generally difficult to directly measure and consist of incorrect set points, scheduling errors, or unusual control board malfunctions. Instead of causing wear and degradation to the equipment, these faults increase the energy expenditure through a multitude of faulty or suboptimal operations. To determine inefficient and faulty operations, a power analysis is conducted by monitoring the energy use of the RTU through a main meter and submeters for the compressor and evaporator fan. Using these energy data, an operator is able to determine whether the unit is active during unoccupied building times, running for either the entire day or a long duration time when it should not be or short cycling periodically. This power analysis method is demonstrated for Tool 1 in Sec. 6.1.

The annual energy cost savings will be calculated using Eq. (20) by multiplying the energy usage by the cost of electricity (e.g., $0.12)

To calculate energy savings, the energy use data, which was obtained by the IMS for faulted and un-faulted (retrofit) scenarios, were correlated with the outdoor ambient temperature or cooling degree-days (CDD) or heating degree-days (HDD) to calculate weather-normalized annual energy savings by FDD tool by site.

The purpose of the market evaluation is to determine the current market conditions for streamlining adoption of FDD technologies and to analyze stakeholder feedback for the purposes of designing a utility marketing and incentive program that increases the adoption of FDD technologies. The goal is to propose a marketing and incentive program that the study partners or other utility companies can replicate in other states. To identify energy-efficiency technology adoption attitudes and FDD market barriers, two preliminary surveys were conducted at a utility customer outreach event in Connecticut in 2018. Thirty people responded to different aspects of the survey, and their roles are characterized in Fig. 3.

Table 5 shows survey questions developed to understand the perspectives on the value proposition for energy efficiency and participant responses. A five-point Likert scale was applied to understand relative value propositions. Each respondent was asked, “How important are the following value propositions to your organization, customers, or stakeholders when investing in building energy efficiency technologies?” Table 5 depicts the responses from 28 people and calculates the mean and rank order. The results indicate that (1) energy consumption, (2) energy cost, (3) energy demand, and (4) occupant comfort ranked with the highest value priorities followed by (5) operation cost and (6) maintenance cost.

Table 6 shows survey questions developed to understand the severity of market barriers to the adoption of FDD technologies. A five-point Likert scale was applied to understand relative market barriers. Table 6 also tabulates the 30 participant responses and calculates the mean and barrier rank order.

Figure 4 graphically depicts the survey results for information barriers, organizational barriers, and technical barriers.

Based on the preliminary survey at the outreach event and previous work [14,26–32], detailed market surveys targeted for different stakeholders (building owners/facility managers; HVAC contractors/energy consultants; utilities; FDD vendors) have been developed and executed with stakeholders participating in the current study. The questions are organized into sections as shown in the  Appendix: (i) knowledge for FDD products, (ii) awareness for FDD products, (iii) attitudes for FDD products, and (iv) respondent characteristics. The results of the market study will (1) help vendors in developing more effective FDD products, (2) support state Conservation and Load Management (CLM) programs in determining proper incentive and rebates for FDD products, and (3) assist FDD vendors and CLM programs in the development of effective FDD marketing campaigns.

The purpose of the process evaluation is to analyze the current processes of purchasing, installing, tuning, and using current FDD tools for retro commissioning (RCM). Common process steps across all installations are shown in the  Appendix. The process for BMS-based FDD tools is different than equivalent processes for non-BMS, hardware-based FDD tools. For example, Tool 1, which is a hardware-based tool, was installed at Site 1. A trained technician was able to install it in less than 4 h. Another hardware-based FDD tool, Tool 2, was installed at Site 2, and this installation proved to be a difficult installation due to limited access to ductwork in the attic space and limited space within the RTU to place sensors, especially the refrigerant pressure sensors. The installation was completed over 1.5 days, including troubleshooting time. The detailed process elevation of FDD products of the ongoing will be completed during fall 2020. The full process evaluation results and analysis will show how each step differs by FDD tool, will analyze variability by process step across tools, and will offer recommendations to improve these processes for future FDD product installations.

The IMS was installed at ten sites in Connecticut to analyze the performance of each FDD product in the field. Data were collected to verify all of the sensors were connected and working properly. Additional visits to the sites are made to rectify monitoring sensor errors when irregularities in sensor data are identified. Preliminary faults detected by the FDD tools are categorized into sensor, control, hardware, and other fault groups and summarized in Table 7 for each tool's baseline period. The IMS data were analyzed, and the faults were identified based on the FI and FIR-based methods described in Sec. 3.4.

A summary of findings based upon analysis of IMS data is presented in Table 8. The EER is calculated using three different methods: two by using air side performance and one-way by using refrigerant-side performance. This method is applied for redundancy since airflow measurement in the field may not be accurate in some instances. Also, sometimes inaccurate temperature measurements may result in inaccurate refrigerant-side EER. The main EER is calculated based upon return air and mixed temperatures and represents the performance of cooling, based on the ability to cool the building. The refrigerant-side EER represents the performance of the vapor compression system and uses the refrigerant line temperatures and refrigerant pressures for calculation.

This paper demonstrates the methods employed in the study and discusses only preliminary results for the FDD Tool 1 at Site 1 in detail in Secs. 6.1 and 6.2. Analysis and results for the other nine tool-site pairs are planned as part of future work. Figure 5 displays photos of the installation of the data logger and pressure/temperature sensors for the IMS, while Fig. 6 presents the photos of the FDD Tool 1, which is a hybrid, hardware-based tool.

In this paper, Site 1 was selected for presenting detailed results from the independent monitoring tool. Figure 7 presents the calculated overall system EER (air side) for a hot summer day (July 21, 2019). The weather data corresponding to this day for this site are shown in Fig. 8. The rated EER for this RTU is 11 as denoted in Fig. 7.

As shown in Fig. 7, the overall EER is slightly over 11.0, indicating no reduction in its performance related to the rated EER.

Figure 9 shows percentage of outdoor air entering the building as well as return, outdoor, and mixed air temperatures. As seen in Fig. 9, when the outdoor temperature is low, the unit is bringing in 100% outdoor air and when the outdoor temperature begins to increase, the outdoor air is reduced to about 20% indicating minimum ventilation air. This result shows that the economizer seems to be working as desired. By examining the outdoor, return, and mixed air temperatures, one can glean how much outdoor air draws through the economizer.

Figure 10 represents the fault intensity (FI) of the refrigerant charge for both refrigerant circuits. The FI value indicates a slight overcharge for Circuit A (about +6%) and a slight overcharge for Circuit B (about +15%). It is known that refrigerant overcharge has little effect on EER and the only concern, generally, is that it may flood the compressor and damage it. Since there is no significant reduction in EER based upon the calculations, overcharge is not a large concern. Since the RTU at this site has TXV, subcooling is calculated and used to verify the refrigerant charge impact. Subcooling temperatures are illustrated in Fig. 11. For Circuit A, the subcooling is in the range of about 12–14 °C and for Circuit B, it is in the range of 16–18 °C. Typical subcooling for TXV-based systems is in the range of about 8–10 °C, indicating both circuits are slightly overcharged.

The fault intensity (FI) for evaporator and condenser fouling faults were also calculated and found to be acceptable, indicating no fouling faults.

Finally, the RTU runtime was examined to see if there is excessive cycling of the unit. Figure 12 presents the power for the main and its components. As shown in Fig. 12, there is cycling of the unit during nighttime, but the system is on continuously (at least one compressor stage) during the day. However, this July day was a hot summer day. During nighttime, there were three cycles per hour, indicating the system has been sized properly and that the cycling losses are minimal.

As presented previously, FDD Tool 1 is a hybrid tool. A sample screenshot of the FDD Tool 1 dashboard is shown in Fig. 13. It shows various monitored variables as well as whether there are any issues related to energy performance, health of the unit, and comfort. The color green was observed for Energy and Health indicating there were no issues. For the initial baseline monitoring period, the FDD Tool did not indicate any performance, health, or comfort faults for the RTU at Site 1. The Tool indicated only a few transient drive fault alarms, related to its variable frequency drive (VFD), and they resolved themselves as shown in Fig. 14. These alarms did not have any impact on the RTU performance and the FDD tool indicated that the RTU at Site 1 was performing well during the initial baseline period. The IMS tool did not identify any faults for this site as discussed in Sec. 6.1. However, this limited verification of FDD Tool 1 does not indicate how effective it is in identifying faults. Unfortunately, it is the reality when demonstrating such tools in the field operating conditions.

Also, the capability of IMS in identifying faults can be verified by introducing controlled faults artificially. However, this approach would interfere with study that is based on naturally occurring faults. Also, getting the approval from the building owners to introduce faults into RTUs is a barrier for recruiting sites for study participation.

The importance of FDD tools and the need for a comprehensive approach to bring the FDD tools into the mainstream are highlighted in this work. A methodology for evaluating the technical performance was proposed for FDD tools installed on RTUs operating under real conditions. The methodology was demonstrated for one installed FDD tool. The baseline testing will continue through summer of 2021, and recommendations for repairs and retro commissioning of RTUs will be identified by the data analysis using the IMS and FDD tools. Once the retro commissioning is complete, the RTUs will be monitored through continuous commissioning, and energy savings potential will be calculated. Also, the capability of IMS in identifying faults can be verified by introducing controlled faults artificially. This approach would interfere with study that is based on naturally occurring faults. However, this task could be attempted at the end of the project. This analysis, results, and recommendations for all sites will be provided in future work. This research also develops and defines an approach for conducting market and process evaluations for FDD products in the field. Detailed results and analysis of the market and process evaluations will also be provided in future work.

The authors would like to thank the United Illuminating Company and Eversource employees for helping the research team to recruit sites for the study and for review of proposed methods and results. We would also like to thank building owners, energy managers, mechanical contractors, and FDD vendors for participating in this study.

• This study has been funded by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy under contract no. EE0008189. Cost sharing is provided by the United Illuminating Company and Eversource through funds from Energize CT, the University of New Haven, and the University of Connecticut.

The authors attest that all data for this study are included in the paper.

h =

enthalpy

m =

mass

P =

pressure

Q =

system capacity

T =

temperature

W =

power

$m˙$ =

mass flowrate

$V˙$ =

air flowrate

X_hs,rated =

ratio of high-side charge to the total refrigerant charge at the rated condition

α₀ =

ratio of refrigerant charge necessary to have saturated liquid at the exit of the condenser to the rated refrigerant charge

$υ˙$ =

specific volume of air

act =

actual

CA =

condenser fouling

CAI =

condenser air inlet

CAO =

condenser air outlet

ch =

refrigerant charge

comp =

compressor

cond =

condenser

dis =

discharge

EV =

evaporator fouling

evap =

evaporator

liq =

liquid line

MA =

mixed air

nom =

nominal

OA =

outdoor air

ref =

refrigerant

RET =

return air

sc =

subcooling

sh =

superheat

suc =

suction line

sup =

supply

Table 9 provides more details on the specification of the FDD evaluator tool and the sensors used as part of its sensor network. The first column describes the sensor, the second column describes the model number, and the third column provides further details and specifications of the sensor. All sensors were pretested in the laboratory before installing in the field evaluator tool.

The goal of the market study is to determine the levels of knowledge, awareness, and attitudes that exist for study participants that may influence the adoption of the FDD technology. The market study will be conducted with all study participants including customers, HVAC and controls technicians, FDD vendors, and utility program managers and engineers. Four market studies were reviewed to determine best practices for conducting this market study and survey: (1) 2012 Connecticut Efficient Lighting Saturation and Market Assessment [32], (2) 2015 Barriers to Commercial and Industrial Program Participation with a Focus on Financing and Cancellations [26], (3) 2014 Connecticut Ground Source Heat Pump Impact Evaluation & Market Assessment [28], and (4) 2015 Connecticut Commercial & Industrial (C&I) Market Research [31]. Table 10 depicts the survey questions grouped into four sections: (1) customer knowledge for FDD products, (2) customer awareness for FDD Products, (3) customer attitudes for FDD products, and (4) respondent characteristics. Descriptive statistics will be calculated and reported for each level of knowledge, awareness, and attitude for FDD adoption. Advanced analysis will be conducted to determine if correlations exist between knowledge, awareness, and attitudes and the respondent's intention to adopt FDD technology. Results of this study will be used to design marketing, outreach, and education programs that educate customers on the use of FDD technologies.

The goal of the process study is to record the current process for installation of FDD products in the field. The research team monitored major installation steps and recorded what went well and where there were issues or problems with an FDD installation. Table 11 indicates 23 major steps that occurred in some or all of the installations. Typical steps, times, and issues will be reported in the process study as well as variance between installations.