These studies show that computerized approaches have the potential to improve energy efficiency of HVAC systems by addressing issues associated with managing complex systems through the elimination of human involvement in maintaining these systems。 They can enable the systems to automatically detect abnormal conditions, diagnose the causes, and mitigate the faults, thus eliminating their impacts on the performance of the systems。 However, many of the studies and developments are still conducted in academic fields。 Very few commercial products were deployed in real-world projects (Liang and Du 2007)。 One primary reason identified by the researchers is that because these approaches were developed by researchers, their deployment requires thorough knowledge of HVAC systems so that the correct information can be provided to them and the systems can be adjusted according to their outputs。 This requirement is beyond the average skill level of most system operators (Katipamula and Brambley 2005a)。

We envision that an integrated framework, which can automatically manage the computerized approaches by providing the needed information and reconfiguring the HVAC systems according to their outputs, can solve this problem。 In this paper, we discuss the identification of functional requirements for developing such an integrated framework。 We will introduce the synthesis of existing self-configuring approaches, and an analysis of the requirements for developing the integrated framework using an implemented prototype system。

804 COMPUTING IN CIVIL ENGINEERING

PROBLEM STATEMENT

Previous studies showed that computerized approaches are able to improve the energy efficiency of HVAC systems by automating two processes: (1) detecting, diagnosing and mitigating faults; and (2) evaluating the performance of the HVAC systems and improving their control strategy。 We identified three challenges which contribute as possible impediments for the deployment of these approaches in the real-world。

First, it is very difficult for system operators to prepare the needed inputs and process outputs for the approaches (Kumar et al。 2001; Venkatasubramanian et al。 2003; Katipamula and Brambley 2005b)。 As shown in Figure 1, every approach requires some inputs, such as the condition measures of the building environment, the configuration of the HVAC systems, or the properties of the building elements。 For different buildings and different HVAC systems, the inputs are very different in terms of data type, communication protocol, file format and the stakeholders who create them。 Outputs of these approaches also need to be interpreted by the system operators so that they can use the information to re-configure the systems。 As a result, it is very challenging for the system operators to collect and process all the required information manually。

Figure 1。 Illustration of the requirements for preparing inputs and processing outputs for the computerized approaches

Second, there is no single solution that fully automates the process of mitigating faults or improving control strategies。 Each of these approaches focuses on different portion of the processes。 For example, the rule-based approach proposed by Schein et al。 (2006) aims to identify the faults and does not address the diagnosis and mitigation of the faults。 Similarly, the approach developed by Fernandez et al。 (2009) focuses on automatically mitigating the sensor drift fault and assumes that the fault can be correctly identified。 There is no single approach that is able to achieve the overall objective of improving energy efficiency by itself。 As a result, there is a need for integrating various approaches developed in this domain to address the overarching goal of increasing energy efficiency through better performing HVAC systems。

To summarize, it is challenging to apply the computerized approaches in real-world systems because it is very difficult for system operators to prepare inputs and process outputs of these approaches manually and many different approaches need to be combined to achieve the improvement to energy efficiency。 To utilize the energy saving  potential  of  these  computerized  approaches,  a  framework  is  needed     to