Can Both the Economic Value and Energy Performance of Small- and Mid-Sized Buildings Be Satisfied? Development of a Design Expert System in the Context of Korea

Abstract

To design a High-Performance Building (HPB), a performance goal should be clearly set from very early design phases, and then a decision path of what performance measures have been chosen in the past stages and shall be chosen in a later stage should be visible. In particular, for small- and mid-sized HPBs that are constructed with a smaller budget, if applicable performance measures are subjective to change, supplementary design costs can increase due to intermittent performance evaluations. To help this situation, we are developing a design expert system for small- and mid-sized buildings that pursues a balance between economic value and energy performance. The economy rule base suggests the most economic building volumetry and form in view of the site context, while the energy rule base suggests a series of energy-sensitive design variables and their options. Based on these rule bases, the expert system presents multiple design decision paths. The design decision support model of the inference engine helps stakeholders choose a preferred design path out of multiple paths, compare the paths, trace back the paths, and effectively revoke past decisions. An actual small retail and office construction project was chosen as a test case to compare the utility and robustness of the pilot system against the conventional design practice. In case of a rather risky design change scenario, the decision-making using the pilot expert system outperforms the conventional practice in terms of selecting designs with a good balance between economic value and energy performance. In addition, it was easier for users of the pilot system to forecast risks upon critical design changes and, in turn, to identify reasonable alternatives.

1. Introduction

As the desire for High-Performance Building (HPB) in the construction sector has been increasing, Korean authorities have assigned minimum energy performance requirements for larger and public buildings [1,2], resulting in an increase in the number of new built HPBs. At the same time, the need to analyze the economic feasibility of whether small- and medium-sized buildings can be built as HPBs has increased gradually.

It is known that HPBs are capable of reducing lifecycle energy costs, which is one of the major reasons that we choose HPBs. Meanwhile, several empirical studies have reported that sale price and rent of HPBs are usually higher than conventional buildings without energy efficiency measures [3,4,5,6,7]. However, developers typically have little investment incentive as they do not have to pay buildings’ operating energy costs, whose burden falls upon property owners or tenants [8]. In addition, the design and construction cost of HPBs, which refers to the initial cost, has been known to be relatively high [9]. Thus, without the setup of performance measures from the planning phase and a clear decision path about what performance measures are selected as design progresses, the initial cost will be excessive due to contingency or the result will be a design draft that abandons the performance goals due to the high investment.

Until now, many Korean architects have made primary HPB design decisions with assistance from sustainability consultants at the actual decision-making moment of each design phase [10]. In general, consultants propose a design alternative in light of previous project experiences and professional knowledge in relation to building science and energy systems. They offer assistance in the HPB design decision-making by providing the quantified estimated performance of the design alternative through simulations.

However, many small- and medium-sized buildings in Korea are unfortunately placed in a blind spot for HPBs. The reason for this is due to insufficient project budgets, whereby developers are faced with many clients who only prefer satisfying the minimum performance compliance and thus consultants could not be hired. In actual practice, therefore, a combination of specific energy-efficiency measures is set as a template when planning small- and medium-sized HPBs; the template measures are then added or deleted to meet the budget of client without exploration of other alternatives, and, in the final stage of the design development, the energy performance of the final design is often conformed as post-approval.

In the long run, paradoxically, small- and medium-sized buildings should focus on economic value more than large-sized buildings. This is because clients who are passionately interested in HPBs at the early phase of the project may give up on HPBs as the design cost increases due to a number of formal performance evaluation tasks and complexity as the design progresses, where it is then recognized by clients that the increase in design is directly correlated to the increase in first cost.

To overcome the main energy efficiency challenges and leverage the private green finance, the Asian Development Bank Institute suggested that, in addition to strong policy frameworks with right economic and regulatory drivers, more resources should support technical assistance [11]. The technical assistance includes activities such as awareness raising and derisking that are essential to create sufficient demand and commitment of stakeholders. In addition, they indicated that a trustable technical supply chain connecting investors and suppliers should be generated because it does a crucial derisking function for unfamiliar energy efficiency investments [11].

This study aims to provide a technical advisor between investor and designers for small- and mid-sized buildings, by proposing a design expert system that can support transparent design decision-making. On behalf of consultants, the design expert system presents multiple decision paths based on quantitative performance evaluation. The proposed design expert system was developed by benchmarking the actual design decision-making process of the domestic design practice (Section 2.2). It is largely composed of rule bases where professional knowledge of HPB design is concentrated, and an inference engine by which transparent decision-making process can be supported from early design phases.

This study comparatively analyzed two design methods of actual mid-sized HPB projects: one follows the pilot design expert system and the other follows conventional design practice, thereby determining which advantages or disadvantages are found in HPB design decision-making when using the expert system and how well the expert system reflects the client’s requirements and designer’s intention compared to those using conventional practice. In particular, this study focused on the following study methodology.

(i)

The priority of selection varies according to client preference/abhorrence at the time of the design decision-making. In this study, decision-making based on the priority of economic value over the energy performance is called “Max. economy design discipline with affordable performance”, whereas the opposite, namely decision-making based on the priority of energy performance over the economic value, is called “Max. performance design discipline with affordable economy”. Although these two design philosophies can be somewhat polarized, many actual design-makings lie between the two disciplines. In this way, distinguished features of the decision-making process between the expert system and conventional practice can be clearly revealed.

(ii)

This study comparatively analyzed the utilities and robustness of the design expert system against conventional design practice using evaluation criteria that could measure how well the gain and loss of the designer and client—by their own choices—were accurately recognized and informed decisions were made during the design decision-making process of small- and mid-sized HPBs.

To do this, this study first reviewed literature in the field of design-decision support tools and compared strengths and drawbacks of existing decision-support-tools. After this, this study reviewed local HPB design process to analyze the roles of stakeholders at each design phase and to identify the current status of conventional practice, thereby analyzing the validity of the pilot design expert system as to whether it could solve the problems of existing tools and deliver desired value to stakeholders.

It is noted that all data and information collected in this study, as well as the comparative analysis, were based in the Korean HPB design context.

2. HPB and Related Design Decisions

2.1. Existing HPB Design Decision Support Tools

As quantitative rationales are very important in design decision-making with regard to building performance, either prescriptive specifications are applied or comprehensive performance of design alternatives are verified through simulations. However, because architects, who are primary design decision-makers, are nearly unable to exhaustively explore the most economically feasible design draft that also satisfies the energy performance criteria, many studies have been conducted on the development of design decision-making support tools that provided the logic and rationales to support decision-making by designers. Those studies are divided into four categories as follows:

(1) Design guideline

The design guideline such as ASHRAE Advanced Energy Design Guides [12] enumerates prescriptive design variables to improve building performance that must be set up in every design phase and provides rationales to recommend specific design variables and qualitative design advice about which expected effects can be obtained when specific variable values are selected. Thus, architects also need to have considerable knowledge of building physics and experiences in HPB design as well as analyze and digest the design guideline advices, thereby applying them to design-targeted buildings, and it is well known that the design guideline advices are no longer valid if the building context changes.

(2) Sensitivity analysis

In the sensitivity analysis, a base model of a target building is set up, and major design variables are selected. Then, options for each variable are configured and many combinations (hundreds to tens of thousands) are evaluated through simulations, thereby identifying variables that have the largest output change in response to input change—that is, ranking the most sensitive design variable to the most insensitive variable to the energy performance in order [13,14]. However, because sensitivity analysis does not designate what values should be used, even though it determines sensitive design variables in the context of target building, design-variable values are still determined by the subjective judgment of architects.

(3) Design optimization

Similar to sensitivity analysis, a problem space is created with several design value combinations, which are applied to the base model of the target building. Then, a mathematical optimization algorithm selects the most optimal solution(s) of the problem (e.g., minimum energy use) in the objective function [15,16,17]. Design optimization simultaneously provides design variables and values that are sensitive to the context of target building. This allows designers to worry less about the selection of specific design variables compared to sensitivity analysis. However, if architects do not select the combinations of the design-variable values suggested by the optimization algorithm, the optimization algorithm must provide another combination of design-variable values after re-computation, which slows down the decision-making.

(4) Metamodel

Optimization makes a problem space by simulating analytical models and evaluates a single instance model; if the model is found to be not optimal, it re-evaluates another instance model. In this way, if the computation time exponentially increases in a process of finding the optimal solution, the design decision-making must be delayed. Thus, it often happens that a vast problem space is constructed in a cloud computing environment in advance in preparation for the designer’s on-call inquiries. This online problem space in conjunction with the solution made by simulating a behavior of analytical model is called the metamodel [18,19,20].

As a result, exploring a virtual problem and solution space can reduce the turn-around time of the design decision-making more than evaluating actual analytical models using sensitivity analysis or an optimization algorithm. Another advantage of the metamodel is that it enables energy performance evaluation in real time without the need to develop alternatives made by designers via separated simulation models. Thus, if the design option space that can be adopted by designers is sufficiently reflected in the metamodel, simulations can be easily used as a quantitative basis of decision-making without special modeling expertise.

(5) Design expert system

Expert system is a knowledge-based system that emulates human experts’ decision making. Expert system aims to solve a complex problem by reasoning that evaluates the current state of the given problem, applies relevant rules, infers answers and then asserts new knowledge. Its knowledge base stores a structured information, and represents it mainly with if–then rules.

Compared to the 1980s and 1990s when many expert systems were defined and developed for AEC industry, only a few expert systems for building design decision support have been observed in the literature over the last decade. (daylighting [21,22], building design [23,24,25,26,27,28,29], building retrofit [30,31], and monitoring and fault detection and diagnostics [32,33,34]). Many expert systems in previous studies employed simulation, sensitivity analysis, optimization, metamodeling, and rule base as an inference engine logic, and proposed process models that are suitable to a specific application domain.

Commercial packages such as Autodesk Insight [35] display an energy-saving measure line-up that can be applied to a given BIM (Building Information Model) by running a large number of simulations in a cloud computing environment. The Home Quality Mark [36] provides home developers with sound rules which offer terms of costs savings against sustainable development measures. However, they may be considered as design platforms because they exhaustively list up what they have assorted, instead of informing users what to do based on analytics and intelligence.

The mostly known issue of the expert system is the knowledge acquisition [37]. However, the harder problems include: (i) to set up a knowledge structure that is enough flexible and has fine resolution to comprehend events occurring within the application domain; and (ii) to identify what kinds of tactical (small) problems that the expert system should tackle in first. This study seeks for the solutions by reviewing and investigating local HPB design process, tasks, and roles of relevant stakeholders, and finally what technical design decisions are typically made in each design phase.

2.2. Task and Responsibilities of Stakeholders for HPB Design

Generally, the building-design phase is divided into planning, schematic design, design development, and the construction document, as depicted in Figure 1. The task of each phase is summarized as follows:

(i)

In the conceptual design, project characteristics are specified by analyzing the purpose of site and building use, client requirements, and general conditions and constraints, and the building volume and economic feasibility are reviewed. The OPR are prepared by collecting project objectives and expected programs from clients.

(ii)

In the schematic design, the building’s base framework—that is, the building mass and geometry, purpose, structure, materials, construction methods, major systems, and facility program—are established.

(iii)

In the design development, the basic properties established in the schematic design phase become more concrete and are thereby reflected in the construction document, in which interim review on engineering design in relation to structure, civil, landscaping, mechanical, electrical, telecommunication, and fire protection engineering is conducted to confirm the design. In this phase, various impact assessments and preliminary certificates take place and, finally, the building permit is progressed.

(iv)

In the construction document, specifications determined at the design development phase are mapped in detailed drawings to enable actual construction. The construction period, cost, and all overall matters related to construction are written in documents.

Although the stakeholders at each design phase are client (i.e., building owner or project developer), architect, consultant, and engineer, the main design decision-makers are the client and architect. Once clients establish the basic goals and direction of the design and suggest requirements, including the economic feasibility and environmental performance of building, the architects select design variables that satisfy the basic design goals, including energy goals, while satisfying the OPR and statutory code requirement, overcoming the budget and schedule limitations, and determining the values of design variables. Consultants and engineers play a role in providing quantitative logics and rationales of the design decision-making of client and architects in each phase or major design changes.

Table 1. Design task, primary stakeholders, and milestones of each design phase of HPBs.

	Conceptual Design	Schematic Design	Design Development	Construction Document
Primary Stakeholders	Client, Architect	Client, Architect, Sustainability Consultant	Client, Architect, Sustainability Consultant, Engineers *	All Design Stakeholders
Main tasks	· Clarify client requirement · Clarify project scope and condition · Rough building mass, form, facility programming · Feasibility assessment · Project timeline and budget estimation	· Establish project objective and direction · Further develop building volumetry, form, layout based on facility programming · Review alternatives and determine final design · Estimated project schedule; estimated budget	· Space program, schedules of accommodation; horizontal and vertical circulation route · Architectural plans, elevations, sections, key assembles and key component drawings; block plans, road layouts/landscape, site plans, external works; outline specification, finishes, doors, use schedule · Structural plans, elevations, sections, specifications; building service plans; fire and safety strategy · Detailed cost plan; design review board, statutory approval, preliminary certificate	· Integrate architectural and engineering designs · Confirm all material, construction, and schedule · Refine architectural and engineering details · Construction details and specification · Quantity take-off · Confirm cost plan · Construction schedule
Requirement	OPR; Building regulation; planning permission	Code/standard requirement and target; Facility program requirement	Review and permit requirement; certification requirement	Technical constraints; Constructability
Budget/cost	Economic feasibility analysis	Defined and being estimated	Being developed and adjusted	Fixed at the end of the phase
Schedule	Lump sum schedule	Rough-in schedule	Being developed and adjusted	Fixed at the end of the phase
Energy goal and task	Prescriptive energy code **	Prioritized or preferred energy goal, and energy certification ratings defined	Measure energy performance index, rating, and code compliance	Design energy performance goal verified

* Concerning civil/landscape, structure, MEP/FP, ICT, and curtain walls. ** Such as floor area requirement for district heating and renewable energy ratio requirement for public facilities.

summarizes the design tasks at each HPB design phase; stakeholders during design decision-making; tasks and requirements that should be defined and achieved; milestones related to project budget, cost, and construction period and processes; and energy goals and tasks at each design phase. In addition,

Table 2. Primary design decisions and variables determined in conceptual design phase and roles of primary stakeholders.

	Conceptual Design Phase
Primary design decisions	Go or no go upon economic feasibility analysis Project timeline and budget estimation
Building design variables	Rough building mass to calculate building coverage ratio, floor area ratio, building height, number of floors, rough rentable area, and number of parking lots Rough building form to protect solar access right, and to secure unobstructed view of neighbor buildings Master plan to assess landscape ratio, open space & circulation, utility easement
HVAC & R design variables	N/A
Passive design variables	N/A
Architect	Feasibility assessment (volumetry, budget, planning approval, regulations and codes) Prepare contract if client approves Executes team formation, project timeline, cost estimate, procurement options, and buildability and construction logistics
Client	Clarify project scope including site Clarify OPR, in particular for project intends, budget, timeline, and conditions and constraints Prepare and plan project financing Provide additional checklists for functions, aesthetics, rent, etc.
Sustainability consultant	Consult the design per energy code requirement and standard Provide technical advices, if necessary
Engineers	Provide technical advices, if necessary

,

Table 3. Primary design decisions and variables determined in schematic design phase and roles of primary stakeholders.

	Schematic Design Phase
Primary design decisions	Establish project scope and objective Confirm building volumetry, form, layout, structure, construction Estimated project schedule Estimated budget
Building design variables	Mass and geometry, orientation, aperture, envelope Structure, construction Facility program, zones, core and shared area, entrance, and circulation
HVAC&R design variables	Prioritized or preferred specific system defined (e.g., district heating, ground heat exchangers)
Passive design variables	Natural ventilation Daylighting
Architect	Project team formulation Review regulations/code, renewable energy requirement, district heating, certification requirement Prepare alternatives Estimate budget and project timeline
Client	Confirm decision making protocol Set project timeline, budget, and contingency Confirm facility programming Trade-off between OPR and cost
Sustainability consultant	Use plan of passive measure, preliminary performance assessment (e.g., by simulation) Prescriptive and quantified certification requirement and constraints such as renewable energy ratio, maximum energy use Consult the design per energy goal and code requirement
Engineers	Reviews civil, structural/landscape, MEP/FP plans

,

Table 4. Primary design decisions and variables determined in design development phase and jobs of primary stakeholders.

Design Development Phase
Primary design decisions	Confirm legal and technical requirement Confirm space program, room functions, schedules of accommodation, horizontal and vertical circulation route, architectural plans/elevations/sections Confirm primary engineering plans/elevations/sections/specifications
Building design variables	Functional spaces and rooms described in plans and sections Transparent and air flow elements described in elevations Vertical circulation routes, e.g., stairs and elevators; EPS, PS, and TPS; dry area Properties for all envelope and constructions Exterior sun controls
HVAC&R design variables	Efficiency for major mechanical and electrical systems Location of plant (including renewable energy systems), power substations, water tanks, air handlers, terminals, and fans and pumps
Passive design variables	Insulated and air-tight envelope Solar and daylight controls Load shedding and shifting by thermal mass
Architect	Confirm layouts/plans/elevations/sections Apply primary engineering configurations Specification and key assemble/component drawing Confirm cost plan, procurement options, buildability, and construction logistics
Client	Confirm project plan Confirm spatial and functional requirement Confirm energy goal
Sustainability consultant	Quantified assessment of energy demand reduction by applying passive measures Coordinate between architect and MEP/FP and curtain wall engineers to satisfy certification conditions and environmental compliance Suggest alternatives to meet the energy goal, prepare reasoning and basis for HPB design decisions Measure energy goal and code compliance, and then ask for design revisions, if any discordance
Engineers	Confirm primary engineering configurations and specification

and

Table 5. Primary design decisions and variables determined in construction document phase and jobs of primary stakeholders.

Construction Document Phase
Primary design decisions	Construction integrity of architectural and engineering designs Confirm all material and construction Level of construction details and specification Construction schedule and cost
Building design variables	Thermal zoning All material specification Building operation schedules and details Internal heat gains
HVAC&R design variables	HVAC load calculation HVAC&R sizing Specification and performance curves for all primary and sub systems Supervisory and local controls Heat and energy losses through pipe, duct, etc.
Passive design variables	N/A
Architect	Perform design changes per client request and regulation authority’s request and guideline Examine architectural and all engineering design comprehensiveness and integrity Prepare construction document, quality assurance Prepare specification and quote
Client	Set operation and management strategy and policies Approval of construction document Project financing plan
Sustainability consultant	Assess the performance of final design Assess the compliance of final design against performance certification, performance incentives and penalties Apply for performance certification
Engineers	Prepare all engineering drawings, specification, and details

describe the building and system design variables in relation to the general primary design-decision and energy performance at each design phase during the actual HPB design practice and specifies the roles and duties of stakeholders who make primary design-decisions and decision-making supporters. When carefully investigating the work and primary design-decision in each phase of the local HPB design practice, as described in Table 2, Table 3, Table 4 and Table 5, some precautions that should be taken during the HPB design are as follows:

• Conceptual-design phase: A project brief and design concept is set up at this phase, and most important decisions are related to increase the economic feasibility of the design concept. Building volume, mass, orientation, and structure thus tend to be determined depending the economic value rather than energy performance. Once the conceptual design draft turns out be economically feasible, performance related decisions can be made next.
• Schematic design phase: As properties concerning building geometry, layout, and structure are further developed at this stage, opportunities for passive measures including natural ventilation, daylighting, and load shifting or load shedding in terms of structural mass should be assessed. If geothermal heat pumps are employed, the location of ground heat exchangers (open-air ground vs. subterranean) must be determined at this stage because a wrong placement not only lowers energy efficiency but also causes irreversible first cost upturns.
• Design development phase: All plans, sections, elevations, the overall building form, its detailed layouts, space programs, fabric and facades, the overall appearance, and construction materials and finishes are determined at this stage. In addition, all primary engineering configurations and specifications are set up. Materials for permit and approvals are produced at the end, which are binding legal documents to which the client, the designer, and other stakeholders must adhere. Values of all energy-sensitive design variables are determined and system configurations are fixed at this stage. It is very difficult to modify the energy goal in later phases. Accordingly, the client, designer, and engineers should have a consensus over primary design decisions and design changes made at this stage by evaluating the energy performance upon fix and change events.
• Construction document phase: Because all the engineering details are added to the frame developed during the design development phase, the capacity, specification, and performance curve of plant and storage systems, air conditioning and ventilation system, hydronic and sanitary system, and air loop system are determined. Operation and control strategies of those systems depending on the purpose of the space and thermal zoning are confirmed as well. At the end of the construction document phase, which is just prior to construction, performance evaluations of the entire building and system should be carried out to confirm that the final design-level energy goal is met. If not, it is unavoidable to either incur critical design changes or else lower the energy goal.

Prior to the construction document phase, the energy performances of the design drafts should be frequently evaluated to make timely and proper design decisions. However, since the details and specifications at a level of the construction document phase are needed to perform the energy simulation, despite that the design draft is prepared in earlier phases, only two measures can be adopted: (1) experts who have experiences and professional knowledge in HPB designs and performance simulations with the construction document level details; or (2) simulations are performed depending on the default setting of simulation tools if simulation experts cannot be acquired or hired.

Realistically, simulation experts cannot evaluate the energy performance of design drafts at each design phase of small- and medium-sized buildings. Thus, designs tend to be processed per the architect’s experiences and intuition without quantitative evaluation, and, eventually, a final single round simulation may be once performed to calculate the energy rating grade, which can be considered as a practical norm in small- and mid-sized building designs in Korea.

3. Design Expert System for Small and Mid-Sized HPB

When the client chooses preferred options, he/she does not necessarily prioritize the energy performance or even choose an uneconomical option. Thus, a design expert system should also reflect this kind of human-oriented uncertainty in actual design decision-making.

That is, design expert systems should have the mechanism to support design process-oriented decision-making, which provides designers and clients with a variety of decision options rather than simply showing optimal or best conditions fragmentarily. Moreover, the proper structure and mechanism also reveal which impacts are exerted on the follow-up design variable selection when users select specific design variables. In addition, they should show transparently which direction is driven by this decision-making in terms of comprehensive performance.

The features and significances of the design expert system for HPBs that this study pursues are summarized as follows:

(1)

Stakeholders can make informed decisions after fully recognizing and understanding the influence between the chosen design and the alternatives at each design phase.

(2)

Enough energy-sensitive design options that are suitable to the context of the project should be presented. Thus, stakeholders can make significant decisions in terms of energy performance, economic value, and other performance aspects by selecting relevant options.

(3)

When a specific design is chosen in the decision support model, the energy use of that design should be predictable with reasonable accuracy at each design phase.

Figure 2. depicts the main flow of the HPB design expert system. Its submodules include: (i) the inference engine that interprets users’ questions, formulates a machine-readable problem based on the question, prepares solution spaces, and explores/analyzes/presents solutions [38]; and (ii) the rule base that is a logical format of the extraction of HPB design knowledge and expertise. Specifically, the pilot system has two rule bases: (i) the economy rule that suggests building volume, form, and layout, which gives the increased economic value given context; and (ii) the energy rule that suggests energy-sensitive design variables and context-dependent design options.

In particular, the inference engine executes a massive sequence of simulations, each instance of which matches an energy model and a design variable combination of conditions and constraints by the economy and energy rules. Then, it produces a surface response model (i.e., metamodel), and then extracts a transparent tree-algorithm-based decision support model out of the response model.

3.1. The economy Rule

The most influencing design variable in the economic value of small- and mid-sized buildings is rentable area. Thus, although economic feasibility improves with an increase in the overall building volume, building coverage, floor area ratio, number of stories, and building height are restricted depending on building use and site district plan (e.g., commercial or residential district). In addition, building shape is also restricted so as to not infringe on the solar access right and views of neighboring buildings. In some cases, there is a restriction on minimum open-space ratio or landscape ratio, and installation conditions of building access and vehicle access openings may affect the building mass and floor shape depending on the location of main roads.

Eventually, the economic value of buildings is secured by satisfying the restrictions and conditions due to site and building purposes, while maximizing the rentable area. The restriction and constraints of building massing can be formatted and formulated in terms of if–then rules. To build a pilot expert system, this study sets the small- and mid-sized commercial facilities constructed in urban commercial zones as its system scope, and then derives the economy rules regarding the restrictions and constraints of building massing, as some examples are described in

Table 6. An example of the economy rule.

Rule #1: IF site is located in the commercial zone THEN building coverage ratio ≤ 40% of the site area and floor area ratio ≤ 600% of the site area Rule #2: Spare a parking lot for every 134 m2 of the floor area

.

Although a Graphical User Interface has not yet been developed, according to the economy rule, the inference engine draws the building footprint that is set-backed from the site boundary while maximizing the footprint ratio. For now, only a rectangular footprint is applicable. Then, the engine calculates a possible number of floors that maximizes the floor area ratio. Finally, the XYZ coordinates of all the eight points of the building mass are calculated by applying unit floor heights. The number of underground floors is calculated per the minimum required number of parking lots. Further geometry allocations depend on specific use case.

3.2. The Energy Rule

Design variables that influence the energy performance of buildings the most are determined from the conceptual to the design development phase. In the conceptual and schematic design phases, the levels of building mass and form, simplified envelope geometry, orientation, structure type, and primary energy system are determined; in the design development phase, the zoning, space program, and construction and engineering details and specifications are determined.

The energy rule provides appropriate design options for each variable by selecting energy-sensitive design variables suitable to the building and site context, while the economy rule provides a building size that can maximize the rentable area.

In particular, since the option range per design variable can significantly influence the energy sensitivity, design variable options that are selected frequently in actual practice according to building purpose were set in energy rules (e.g.,

Table 7. An example of the energy rule.

Rule #1: IF site is located in the mid-northern area AND it is NOT for residential use THEN U-value of the exterior wall ≤ 0.24 W/m2K;

Rule #2: IF number of floors ≤ 7 AND it is for commercial uses THEN exterior wall type SHALL BE ONE OF { ExtWall1: Granite + Reinforced concrete + EPS 110 mm, ExtWall2: Granite + Reinforced concrete + Phenolic foam 70 mm, ExtWall3: Aluminum panel + Reinforced concrete + Phenolic foam 70 mm};

), which were developed through interviews with architects, consultants, and MEP engineers. Because the simulation model should include not only the design variables that were mentioned in each design phase but also the specification values (e.g., fan efficiency or water loop pump efficiency during fan coil unit selection) that were determined in the later construction document phase or dependent variables according to major design variables (e.g., SHGC and U-value should be lower than the certain maximum when double-glaze windows is selected), dependent and default variables according to options for each design variable were also defined through interviews with experienced designers of construction documents and MEP design engineers.

3.3. Design Decision Support Model for Inference Engine

The logic that selects HPB design variables and options that are suitable to the needs of the designer and client—that is, a design-decision support model—was implemented with a metamodel. A metamodel substitutes a simulation model. Thus, as a simulation model must be prepared with a different level of details (LOD) for each design phase, a metamodel for each design phase shall also include the design variables whose LOD corresponds to the design phase. Therefore, a metamodel can be divided into a schematic design metamodel and design development metamodel according to a resolution of design values and their variables.

Hundreds or thousands of simulation model instances should be prepared for a metamodel. Instances are made of combinations of options for each variable and design variables determined at each design phase. However, for the usability test of the pilot system in this study, simulation instance models, whose properties were determined in both the schematic and design development phases, were developed using EnergyPlus.

A metamodel should be transparent enough to intuitively identify the difference in energy performance between the chosen design options. The previous study reviewed some of the machine learning algorithms that can make a transparent metamodel, including clustering and decision tree algorithms [39]. Then, it was confirmed that the Confidence Inference Tree (CIT) algorithm had the best fit as a transparent decision-support model [39].

4. Comparative Experiments for a Mid-Sized HPB

It was assumed that a client makes a request to build a mid-sized commercial building in Hanam, South Korea. A typical client would ask for the maximum economic value and an energy performance that complies with legal requirement. However, some clients may place more emphasis on enhanced performance as long as the economic value is satisfactory. To obtain a range of clients’ varying requirements and expectations, this study experimented on how the pilot design expert system helps designers and clients balance between economic value and energy performance at a level with which any given client would be satisfied. Moreover, the current design practice was set as the control group, which the utilities and robustness in design decision-makings using the pilot system were then compared against.

4.1. Site Situation and Owner’s Requirement

With an area of 1100 m², the test site is located in a commercial zone of a new town. Therefore, the construction project should be controlled under legislation and standards such as those listed in

Table 8. Summary of local building codes applied to the test site.

Statutory Requirement	Condition	Statutory Requirement	Condition
Max. building coverage ratio	60% of site area	Min. number of parking lots	One for every 134 m2 of floor area
Max. floor area ratio	400% of site area	Min. open space ratio	Not applicable
Max. number of floors	No limit	District heating	Can be applicable for some spaces
Min. landscape ratio	15% of site area	Min. solar access and view of neighbor buildings	Not applicable

. The test site is one of many, and the entire district faces an eight-lane circulation road in the west and a pedestrian road in the east, as depicted in Figure 3. Fortunately, the building is unlikely to infringe on solar access rights and the view of neighboring buildings given that there is a vacant lot to the south.

A huge floating population is often observed to the west of the site because there are large apartment complexes and public bus stops. Because the site will be built only for commercial rent, which primarily accommodates retails and offices, the client wants as much rentable space as possible. However, obtaining the G-SEED certification [40] and energy rating are necessary to take advantage of acquisition tax reduction and lower loan interest. Usually, higher-grade performance certifications are preferred for tax and financial benefits.

4.2. Evaluation Criteria for HPB Design Decision Makings

Although designers and clients recognize the importance of building-energy performance, performance measures could often incur conflict with the economic value of the building. On the other hand, if only the economic value of the building is pursued, not only the energy but also the overall environmental performance can be degraded, resulting in a long-term decrease in asset value.

Whether a building’s economic value or its environmental performance is prioritized is eventually determined by client; the designer should technically uphold the client’s preference in the design, but it can make reasonable adjustments in certain cases if necessary. Furthermore, the initial requirements may be diluted or strengthened as the design becomes more concrete.

Thus, the design-decision support should provide scientific and objective logic and reasonable rationales to help designers and clients designate or change design variables when the preferences of decision-makers either change or must be changed. It should also provide information for them to make an informed decision with clear knowledge of the gain and loss due to their choice.

Evaluation criteria for the HPB design decision-making process are thus defined as follows:

(1) Practical and exhaustive design options and alternatives.

Decision-makers should be able to select an option out of multiple alternatives that are procurable and implementable within the scope of the project.

(2) Quantified assessment of performance and reasonable evaluation cost.

Comparative performance evaluations between alternatives should result in a quantified and visual artifact; thus, decision-makers should be able to intuitively observe the difference and identify the reasons that cause the difference. In addition, evaluation costs should be reasonable, such that decision-makers should not be reluctant to avoid the performance evaluation due to added evaluation costs.

(3) Transparent decision-paths and revocability

From the earliest decision-point up to the current decision-point, decision-paths should be clearly described, such that decision-makers should be able to discern the (potentially) wrong decision-split whenever underperformance is observed. In addition, previously made decisions can be easily revoked.

(4) Minimized iterations of performance evaluation to meet the goal

Iterations of performance evaluation to confirm whether the intermediate or final design meets the energy goal should be minimized, such that the decision-making process can be more efficient and less expensive.

4.3. HPB Design by Conventional Practice

Conventional domestic design practice follows the work flow described in Figure 1. Initially, the architect prepares viable rough drafts that would satisfy the client’s requirements including budget, site appraisals, and accommodations, while meeting building regulations, such as those shown in Table 8. This initial draft was assumed to be a rectangular box with six above-ground floors and two underground floors, having about 3900 m² of total floor area and 400 m² of rentable area on the ground floor. Once the client reviews the conceptual draft and feasibility assessment, he/she can decide whether or not to proceed to the next stage.

4.3.1. If Client Wants Max. Economy and Affordable Performance

Designers are likely to adopt the design option selected in similar previous projects (in particular, when a similar-purpose building was permitted in nearby regions). The rectangular mass of the building was maintained while the core was pushed over to one side, thereby minimizing the shared space (core, circulation, and service area) in the ground floor particularly. In addition, a void was arranged in the upper floors to maximize the total floor area while maintaining the sixth floor, and curtain walls were selected as an envelope to give a modern look. According to the client’s requirements, retail shops were placed on the ground and second floors, and offices were assigned on other floors. Reinforced concrete was used as the structure and an EHP was installed to provide air conditioning. Other detailed matters followed those used in previous similar projects, thereby completing the schematic design.

The detailed design was reviewed by the client and modifications were reflected, thereby completing the floor plan, elevation, cross-section, and major finish materials, and developing all major specifications of HVAC&R engineering to complete the detailed design draft. In this stage, details selected in similar projects are also likely to be borrowed without deep modifications. Normally, the estimated energy use in the detailed design and compliance with legal requirement are examined just before applying for statutory approvals, and, if the pursued certificate grade is not achieved, options that can reduce energy use with minimum initial investment without overhauling the design draft through changes in mechanical and electric systems, such as luminaire changes, are first attempted. For example, this case had to change the design to extend the district heating over all spaces according to the advice of the official design review board because the test site is located in the district heating supply region. Although the estimated energy use increased with the change to district heating (EUI of 150 kWh/m²), the client did not want to consider photovoltaic (PV) installation. Thus, radiators and package air conditioners were replaced with FCUs and energy simulations were executed again, thereby obtaining an EUI of 140 kWh/m², as depicted in Figure 4. However, because this change still could not meet the EUI requirement, insulation and glazing were upgraded resulting in an EUI of 130 kWh/m². Since then, the detailed design draft has been completed, thereby applying for a building permit and preparing the construction documents.

4.3.2. If Client Wants Max. Performance and Affordable Economy

From the schematic design phase, designers tried to reduce energy demand first while complying with the building volumetry that secured at least minimum economic value. Because the building size was relatively small, they determined that inducing cross ventilation by placing the openings to face each other, and expanding a daylight lit area by placing the occupied area in the perimeter, instead of inducing natural ventilation or day lighting by making a wind way or courtyard garden which would fragment the mass. Thus, the core was arranged in the center, and the hallway in the ground floor was arranged in the east–west axis while the hallway in the reference floor was arranged in the south–north axis. A green roof of 400 m² was applied rather than employing landscaping on the ground floor to maximize the rentable area in the ground floor. In addition, to minimize the aperture area, 60% of the window area ratio was applied to the east–west side facing the main road and 40% for the south, and no windows were placed in the north façade (Figure 5). The U-value of the glazing was based on the energy code, but the SHGC was set to 0.3 or lower to minimize the solar heat gain. How much energy demand could be reduced if a fin was installed in the west-facing window was estimated using simulations. However, the estimated reduction in EUI was minimal so that the fin was not installed.

According to the requirements from client, retail shops were placed in the ground and second floors, and offices were placed in the rest of the floors. Reinforced concretes were chosen for the structure. To determine whether FCU would be chosen as the primary HVAC system against CAV, energy simulations were conducted to calculate the estimated EUI. However, the client worried about the increasing initial investment and maintenance cost, and reduction in parking spaces due to increase in mechanical room area if absorption chiller and cooling tower were installed for district heating plant system, so that an EHP was set to a primary HVAC system. After this, energy simulations were conducted again and estimated 90 kWh/m² of EUI could be obtained. The effect of an installation of 300 m² of PV, which was to offset the electricity use increased due to EHP, was also estimated using simulations, showing –80 kWh/m² of EUI, as depicted in Figure 6. However, the PV was removed and returned to district heating again according to the advice of official design review board that district heating should be employed for all spaces.

Since then, design development was completed by confirming the floor plan, elevation, cross-section, and major finish materials and developing all major engineering specifications including MEP. The resulting design was simulated once again, thereby verifying the estimated energy usage was less than the requirement (estimated EUI was 120 kWh/m²). If the estimated energy use should be reduced further, changes in mechanical and electrical systems should be attempted first such as replacing chiller, and if energy performance at the level of zero energy building was needed, renewable systems should be re-examined again. However, PV was not considered by client’s request, and geothermal heat pump system was thought as uneconomical in the district heating supply region. In the construction document phase, energy saving controls of various lighting and electrical appliances such as variable frequency drive controls, daylight controls, occupancy sensors, and plug load controls were applied to further reduce the estimated energy use. However, because the resulting EUI did not decrease significantly, those controls were not adopted to the final draft.

4.4. HPB Design by the Pilot Design Expert System

Once user inputs the main purpose of the test building (i.e., retails and offices), and site geometry and location, the expert system calculates the building size and volume that can secure the maximum economic value according to the economy rule, as presented in

Table 9. Geometry and configuration of the test building suggested by the economy rule.

Building Description	Value	Reference
Building footprint	650 m2	Less than 60% of building coverage ratio
Total floor area	3900 m2	Including shared area; less than 400% of floor area ratio
Number of floors	6 above grade floors and 2 under grade floors	40 of under grade parking lots
Landscape area	Replaced by green roof	More than 15% of site area
Green roof	340 m2 at the rooftop	More than 30% of site area
Main facade	West	Main road in the west
Footprint shape	Rectangle	Default
Building core	Central	Building foot print ≥ 300 m2

. The expert system proposed the west side facing the main road as the main façade, and a rectangular box building by placing the shared core in the center.

Once the building size, main façade, and footprint shape that secured the economic value of the retail and office were confirmed, the economy rule suggested design variables that can reflect the preferences of the client or designers, as presented in

Table 10. Optional configurations suggested by the economy rule.

Building Description	Value	Remark
Aspect ratio	{1:1, 1:1.4}	Keeping 650 m2 of the building foot print
Shared area ratio	{20%, 30%}	Including core, service area, and corridor; only for offices
Retail location	{Only ground floor,	Offices at the remaining floors
	Ground and 2nd floors,
	All floors}

. Because the preference significantly varies depending on the individual client, these variables are difficult to be determined by the expert system.

In addition, the energy rule suggested design options for an energy-sensitive design-variable lineup that is suitable to the context of the building and site from the pool of the energy-sensitive variables, as presented in

Table 11. Optional building properties suggested by the energy rule.

Design Variables	Options	Remark
South WWR	{30%, 60%, 90%}
North WWR	{30%, 60%, 90%}
East WWR	{60%, 90%}	Next to the road
West WWR	{60%, 90%}	Main façade
Ext. wall construction	{Granite + Reinforced concrete + EPS 110 mm (0.231 W/m2K),	U-value ≤ 0.24 W/m2K
	Granite + Reinforced concrete + Phenolic foam 70mm (0.237 W/m2K),
	Aluminum panel + Reinforced concrete + Phenolic foam 70mm (0.232 W/m2K)}
Roof construction	{Plain cement + Urethane 160 mm + Reinforced concrete (0.144 W/m2K),	Insulation over the roof; U-value ≤ 0.15 W/m2K
	Plain cement + Reinforced concrete + EPS 180mm (0.144 W/m2K)}
Below grade wall construction	Cement Concrete+ EPS 110 mm (0.252 W/m2K)	Default
Ground floor construction	Subslab Concrete + EPS 90 mm + RC (0.267 W/m2K)	Default; U-value ≤ 0.29 W/m2K
Int. wall construction	Cement Concrete (2.541 W/m2K)	Default
Int. floor construction	Cement Concrete (6.278 W/m2K)	Default
Glazing	{Low-E glass + Argon + Regular glass (1.481 W/m2K, SHGC =0.568),	Double glazing; U-value ≤ 1.50 W/m2K
	Regular glass + Argon + Regular glass + Argon + low-E glass (1.486 W/m2K, SHGC = 0.579),	Triple glazing; U-value ≤ 1.50 W/m2K
	Regular glass + Air + Regular glass + Air + low-E glass (1.436 W/m2K, SHGC = 0.563)}
Exterior sun controls	{No shade,
	Overhang,	Only for west and south
	Exterior Venetian Blinds}
Infiltration	1.5 ACH	Local design convention
HVAC	{CAV,	In case of all retail floors; district heating required
	FCU,
	EHP (retails) + CAV (offices)	District heating required
	EHP (retails) + FCU (offices)
	FCU (retails) + CAV (offices)
	FCU (retails) + FCU (offices)}
Lighting	LED (7W/m2 on average)	Default
Lighting controls	{Daylight controls,	Natural daylighting
	Occupant sensors,	Dimming controls
	Both}

. For example, since the commercial building façade should have some opening, the WWR on the east–west side was set to 60% or larger, as the main entry was located on the west facade and the secondary entry on the east facade. Frequently adopted configurations in actual design practice were proposed for the construction of an exterior wall, window, and roof at a level that the limited U-value of the energy code was not exceeded, and the construction configurations that were applied widely on mid-size commercial building were proposed. On the other hand, multiple options were suggested for important variables whose preference of designer or client is more important, whereas default options were suggested for the variables that were insensitive to the preference. In addition, a set of the variables that almost always used similar values in the actual design practice was designated as the local design convention, although they were energy-sensitive variables. For instance, for infiltration, although air change rate may significantly vary according to the weather-stripping material and construction precision, the local design convention is 1.5 ACH, assuming normal construction.

The energy rule considered the practical design options and building context as much as possible in the options of HVAC&R system and lighting system. Thus, multiple options were suggested to suit for a large range of selections for designers, as presented in Table 11. For the six-story retail and office building, although the FCU or CAV with district heating was the HVAC&R option that is widely used in the market, EHP was the preferable option for building owners due to its easy maintenance, thereby suggesting as one of the system options.

The inference engine built the building base volumetry suggested by the economy rule and applied a combination of design variables and their values suggested by both economy and energy rules over the base volumetry. They were then converted to EnergyPlus models, which were simulated on the parallel computing environment. Sets of the design-variable values and corresponding EUIs were made as a tree-form design support model, as shown in Figure 7. Conditions of the design variables determined at the schematic design and design development phases of the building were specified in each leaf, and designers and clients select the variable values while running down from the root node to the bottom leaf sequentially. Since the EUI range according to the selected design path is denoted in the bottom leaf, the EUI ranges of different design paths can be estimated and compared.

The design variables presented in the HPB decision support model while running down the tree from the root node to the bottom leaf, may not match the same sequence of design variables determined in the actual design process. The reason for this is that the conditions in nodes closer to the root leaf node are more definite for earlier splitting in the tree data structure. Thus, it is possible that design variables determined later in the practical design process can be present in the nodes closer to the root node. This means that the variance of the estimated EUI range is decreased for nodes closer to the bottom leaf, thereby increasing the predictability and lowering the selection risk.

4.4.1. If Client Wants Max. Economy and Affordable Performance

If the client wants the ground and second floors as retail and other floors as office, the estimated EUI range may vary depending on whether EHP was selected as the HVAC of the retail space. The client chose EHP for retail, the initial investment cost of which was inexpensive. However, for the office space, the client chose an FCU that employed district heating because the initial cost of FCU is lower than that of CAV. Although the estimated EUI may vary depending on whether lighting control was used, only a ±5 difference in the EUI range around 105 kWh/m² was estimated; thus, it was decided to not select additional lighting control. Since official design review board strongly recommended the use of district heating for entire building as a plant system, the retail HVAC system was changed to FCU, in which the EUI range was estimated to be 130 ± 5 kWh/m², as depicted in Figure 8.

In the case that client wants to use the entire building for retail purposes, the estimated mean EUI can be up to around 270 kWh/m² if CAV is chosen for the HVAC system. However, this may decrease to 190 kWh/m² if the FCU is selected. When the shared area increased to 30% of the reference plan, the mean EUI decreased to 180 kWh/m². When the shared area was set to 20% (i.e., the area for exclusive use increased), the mean EUI became around 200 kWh/m², which was still affordable energy-performance. Thus, once 20% of the shared area and FCU were selected in the all-retail building, the mean EUI was expected to be 200 kWh/m² even if the client and designers selected other options autonomously.

If only the ground floor was planned as retail and other floors as offices, the mean EUI was expected to be around 110 kWh/m². If the shared area is planned to be 30%, the mean EUI could decrease to 100 kWh/m². However, if the shared area was planned to be 20% to maximize an area for exclusive use and if daylight control and occupancy sensors are installed, the mean EUI can be 100 kWh/m² as well.

4.4.2. If Client Wants Max. Performance and Affordable Economy

Designers and clients who are more focused on energy performance searched the lowermost leaves first in the decision support model (Figure 7), thereby searching for the design path that leads to the leaf whose estimated EUI is the smallest. That is, retail spaces on the first and second floors, and offices on the other floors are arranged, the shared area was planned to be 30%, and EHP was used in retail while FCU was used in offices for HVAC. The estimated EUI was around 100 kWh/m², which is the smallest. In addition, the estimated EUI range of the design draft—where retail is only arranged on the first floor, shared area was 30%, and daylight control and occupancy sensor are installed—is also around 100 kWh/m^2, as depicted in Figure 9.

Clients who select a variety of design options such as the number of retail floors, systems, etc., focusing only on energy performance without considering economic value at all, realistically do not exist. Thus, several design variables that might be the most important to this client are first determined to calculate their estimated EUI ranges, and the client can be persuaded to change his/her preference of design variables that may sacrifice economic value for energy performance to finalize the design draft, or vice versa. For example, if the client insists on putting retails on every floor, choosing a design that reduces an exclusive area slightly (i.e., 30% of the shared area) and mandatory selection of FCU can be a method to reduce the estimated EUI in a relative sense. As another example, if the official design review board strongly recommends district heating for the entire building, the decision support model gives a good insight that selecting FCU instead of CAV, regardless of the number of retail floors, is an easier way to lower the estimated EUI range.

4.5. Comparison of Decision Makings in Case of Serious Design Changes

To compare decision-making processes between using conventional practice and using a pilot expert system (Figure 10), and to analyze their robustness in the case of critical design changes, a serious design-change episode is assumed: the design team submitted building regulation approval documents and applied for an energy performance certification. However, the official design review board recommended to apply district heating to all spaces.

(1) Practical and exhaustive design options and alternatives

Since the test site was located in a district heating region, designers had to consider an HVAC system that at least partially uses district heating such as FCU or CAV. However, as per the request from clients who saw a previous case of using an EHP in a similar building, the client insisted on EHP for the entire building. However, the EHP had to change to FCU eventually for the statutory approval.

In the conventional practice, design options and alternatives in HPBs are likely to be borrowed from previous similar projects or designers’ experiences, which are seen to have considerably high practicality. However, since the conventional practice is likely to employ specific design solutions in repetition, it would be difficult to apply an exhaustive solution that approaches the design problem from various perspectives.

On the other hand, the design expert system may allow the use of EHPs instead of district heating, according to the circumstances of the energy supplier and similar sized buildings built in other district heating regions. The expert system also includes district heating system option for the entire building. Thus, such a risky design change was prepared, and it was flexible for making a decision change to install an FCU for the entire building.

(2) Transparent decision paths and revocability

Conventional practice tends to imitate so-called proven existing design paths rather than using an optimal draft through trial and error. Thus, it would be very difficult to reverse the fixed decision and go back to a previous state of the current decision-point even if fatal errors occur. Even with successful reversal, its rationales must be strong. For example, when the official design review board recommended the use of district heating, the design was changed to supply heating with radiators and cooling with package air conditioner first, instead of FCUs. This is because of the convention of designers who were accustomed to central heating and individual cooling as usually done in apartments. Only after the designer witnessed and verified the quantitative performance reported in the simulation, a system change was finally made to adopt FCUs that further reduced the estimated EUI.

The base case made by the design expert system in the same design-change case was to use individual heating by EHP in retail only, while using central cooling and heating by FCU in offices. Thus, when the official design review board recommended extending district heating further, a design draft that selected FCU even in retail was naturally adopted, which also naturally led to the smallest estimated EUI range. This was because, when the design option of the expert system was planned, opinions of not only specific engineers but also designers and engineers who experienced designs of apartments as well as small- and medium-sized offices were reflected to configure the options.

(3) Quantified assessment of performance and its evaluation cost

In the conventional practice, expected performance was measured using the following two tracks: prescriptive scoring and performance simulation. However, performing simulations is the only option to quantitatively evaluate the comprehensive performance of buildings. In this circumstance, designers have no choice but to run “expensive” simulations (in terms of knowledge and modeling effort) as few times as possible. Simulation therefore became a “measurement ruler” to test only major design change events just before submitting applications for building permits and preliminary certificates. Thus, the design cost will eventually increase in the long term because performance simulations need to be frequently performed, in particular in the Max. performance design discipline, where design changes often occur to optimize the performance.

However, the design expert system does not need to carry out performance evaluations since it already employs a metamodel derived from comprehensive performance simulations.

(4) Minimized iterations of performance evaluation to meet the goal

In conventional practice, performance evaluation had to be carried out whenever design change events occurred, in addition to base case performance evaluation. Even in the Max. economy design discipline, where performance simulations are rarely conducted except for in primary design changes, performance simulations were conducted four times—including the base case evaluation—and in the Max. performance design discipline, performance simulations were conducted as many as six times.

On the other hand, in the design decision-making by the expert system, only two performance evaluations were conducted: when the design development had been completed and when the permit office made a request for all district heating. This was because the transparent multiple-decision path and visual expected EUI per design path enabled the comparison of small or large design-change alternatives prior to the final selection.

5. Discussion

Although many designers and building owners prefer HPBs, building owners are concerned about high initial costs. For designers, the process of designing more cost-effective HPBs with an appropriate level of performance is not straightforward. For buildings involved with high public interests or large capital investment, HPBs should thus be considered from the conceptual-design phase, and experts such as consultants or simulationists should participate from the early design phase to reduce ambiguity. However, it is told that clients of small- and medium-sized buildings constructed by individual investors do not feel the need to construct HPBs aside from preferring the opportunity to charge higher rents. This may be the result of losing trust that HPBs have only high initial investment costs but do not return the utility or satisfaction as much as the investment paid by clients.

To promote small- and medium-sized HPBs, economic feasibility must be thoroughly examined from the conceptual-design phase, and energy performance should be pursued within a range that does not significantly damage the economic value. In addition, since clients’ requirements are changing, differing from original OPRs, and the requirements of the permit office or official design review board are constantly updated throughout the design process, it may be impossible to proceed with the design process in view of the possibility of all the design changes from early design phases. However, the responsibility of designers becomes emphasized in the HPB because designers must predict future changes (even with great uncertainty) related to performance and economic feasibility from early design phases, and then should reflect the prediction as early as possible.

Most of the economic feasibilities of buildings are assessed when the building volume and rentable area are calculated in the conceptual phase, and the initial investment cost depends on which structures, materials, and options—such as systems—are added or removed. Thus, this study started from the initiative that affordable HPBs could be established—if the design-decision support model covers the design variables and practical options whose energy sensitivity is higher out of the design variables determined from the conceptual phase up until the construction document phase. Then, the designer is able to select the variables and options that strongly influence the next selection instantly. In addition, both the initial investment-cost calculation and performance evaluation are possible in real time whenever the designer makes a decision.

As explained in Section 4 of the comparative case study, clients or designers do not necessarily make a reasonable selection; unexpected design changes caused by factors that are external to the design team may impose a big risk potentially to final economic feasibility and performance. While existing design-decision support tools evaluate the determined specific design alternatives by designers and assign the “the fittest” selection to stakeholders (e.g., the design whose EUI was the lowest), the design expert system proposed in this study focused on “selection” itself and supports the decision-making that respects the stakeholders’ selection; even if some selections may be unreasonable at the time of selection in some respect, the second-best decision that was in line with the goal and direction of the project can be made subsequently after the unreasonable selection. The conventional HPB design practice mechanically chooses the configuration of energy-saving options within the budget range because there was a pool of (customarily chosen) options that improved energy performance. However, the proposed design expert system displays an estimated EUI range according to the selected option, and helps the decision-maker to promptly abandon extra energy efficiency options if there is no motivation to further improve the performance, i.e., when the variance of the estimated EUI range is significantly small.

As presented in the design scenario using the expert system in Section 4, when a client prioritized the max. economy, he/she saw the following options as the primary factors that would guarantee economic feasibility: retail shops on the ground and second floors of a six-story building whose total floor area was 3900 m², all WWRs of the facade were maximized to 90%, the shared area was minimized to 20%, and EHPs, and chose to normally select options for the rest of the design variables. Since then, the HVAC system was changed to FCU so that the estimated EUI was calculated at 130 kWh/m².

When a client prioritized the max. performance, retail shops were still on the ground and second floors of a six-story building with total floor area of 3900 m². However, the WWRs in the west/east façade and north/south façade were 60% and 40%, respectively, which minimized the WWR. In addition, lighting control, exterior sun control, and EHPs were chosen as the performance option. Since then, the design change was made to use FCUs such that the estimated EUI was calculated to be 120 KWh/m².

Surprisingly, the difference between two mean EUIs of the above two clients was insignificant, namely 10 KWh/m², despite the different preferable choices they made. The reason for the insignificant difference in estimated EUI was because: (i) the window size did not significantly affect the energy performance because of relatively smaller envelope area of a mid-sized building; (ii) the compliant U-values of the opaque wall and glazing were considerably low; and (iii) the same HVAC system type, namely FCU, was selected. That is, when the expert system is used in the HPB design decision-making, since all the results according to both design disciplines are visible, even if a decision-maker prioritizes energy performance, he/she will select economic options until the utility of the chosen options is large enough compared to the investment cost. Contrarily, even if a decision-maker prioritizes economic value, he/she will select performance options if he/she knows that utility of the performance options is greater than its investment. Conclusively, if the design expert system is employed in HPB decision-making, the final draft would be similar regardless of whether the economic value or performance is prioritized.

However, in the conventional practice, if the max. economy is prioritized, design variables that can improve economic value are set to a constant first, and then performance measures (e.g., changes in windows or insulation from default products to performance products) are likely to be inserted to raise the performance at the final determination phase. On the other hand, if the max. performance is prioritized, a set of measures that maximize the energy performance is applied first and then fixed, and then other measures will be added or removed at the final determination phase. In both cases, the design decision-making can be delayed and thus their design cost will increase, because not only the economic feasibility assessment but also the simulations must be performed again whenever measures are added or removed.

The merits of the proposed expert system lie in the HPB design-decision support model based on the transparent metamodel, and the rule bases based on exhaustive rule sets on which expert knowledge and experiences are concentrated. Since the HPB design-decision support model is made by the base building volumetry that is suggested by the economy “rules” according to site and building context, and a combination of design variables and options that are suggested by the energy “rules”, the performance and usability of the expert system depend on the well-formedness of the rules, such as: (i) how much the economy rule and energy rule are matched with the conventions and customs of the actual HPB project design practice; (ii) how much the actual decision-making conditions and restrictions are reflected; (iii) how much conflict between rules and uncovered regions are minimized when applying the rules; and (iv) whether unexpected results are produced when applying complex rules.

While a majority of the economy rule sets are based on building codes, a majority of the energy rule sets are based on the de facto standard in the field. It means the energy rule sets can be customized via trial and errors if the design expert system becomes a decision support groupware of an architectural design firm. Then, it is anticipated that such well-formedness issue of the rule base can be practically resolved, e.g., frequently employed measures and customized set-ups can be appended to the rule set.

However, the rule base of the pilot expert system still has the following technical vulnerabilities, which need to be improved in the future.

(1)

The current pilot system considers only a square or rectangular footprint. In addition, since the applicable building type now includes only small- and mid-sized commercial buildings, rules about retail and residential buildings, which is more general type for Korean small- and mid-sized buildings, should be developed in the future.

(2)

Only the inference engine and rule bases that lack a graphical user interface are implemented in the current pilot expert system. The rules also follow a very simple formats, such as if–then rules, rather than using semantic formats. Thus, the pilot expert system can express the results according to conventional calculations such as based on building coverage ratio or floor area ratio, but, under more complex conditions, it is difficult for the results to be represented in terms of detailed building geometry and form. For example, the pilot system can calculate the number of parking lots by calculating the total floor area from the number of stories and building footprint and by computing the number of underground stories approximately. However, if there are complex vehicle circulations and functional rooms in the underground floors (mainly in the first basement level), the number of underground stories would increase. Moreover, there might be some ambiguity in applying large WWR to the main façade because there are elements that humans can but machines cannot clearly judge (e.g., a designer can intuitively select the building’s main façade by simply seeing the main road and a floating population). Thus, it is necessary to develop a rule structure that can apply the rules by finding site signature and to generalize the special context and conditions of the site and building, such as the main vehicle entrance and parking circulation.

(3)

The design variables that are dealt with by both rules are divided into continuous and discrete variables. For example, the number of stories of the retail and aspect ratio are continuous, whereas the HVAC system or wall construction are discrete. If the design option preferred by a decision-maker is different from the value proposed by the expert system, EUI prediction may be possible using interpolation if the design option is a continuous design variable (e.g., if a decision-maker prefers to place retail on the ground, second, and third floors). On the other hand, it the decision option is a discrete design variable (e.g., DOAS system vs. FCU or CAV), there is no particular way to predict EUI in the pertinent option of the decision support model that is already made. Nonetheless, this may be solvable by re-structuring a design-decision support model with the addition of options preferred by stakeholders.

(4)

A runtime platform that implements the developed rule bases, simple rule interpreters, simulation populator, and primary process and logic for the inference engine are still under development.

6. Conclusions

To design an HPB, a performance goal should be clearly set from very early in the design stages, and then a decision path of what performance measures have been chosen in the past stage or will be chosen in a later stage should be visible. Otherwise, either the initial cost could be extraordinarily high due to contingency or performance measures may not be selected due to their higher threshold.

For small- and mid-sized HPBs with a smaller budget, more emphasis is placed on economic value, because the client who was interested in the earlier phase of the construction project may give up on performance enhancement; as design and supplementary costs are increasing, due to many intermittent performance evaluations and their complexity, initial costs can be way off the budget even prior to the onset of construction.

We are developing a design expert system for small- and mid-sized buildings that pursues a balance between economic value and energy performance. The design expert system advises HPB stakeholders by presenting multiple decision paths based on a massive volume of quantified performance evaluations that ran before the decision process began. The economy rule base of the pilot expert system was composed of rationales and premises in which developers, realtors, and architects assess the economic feasibility of the construction project. The energy rule base was composed of line-ups of energy-sensitive design variables collected through surveys and investigations among architects and engineers. Those variables and their values are carefully selected measures that are frequently used in small- and mid-sized retail and office’s local design practice. Once a user inputs the site geometry (including address) and building type, then the economy rule base suggests the most economic building volumetry and form in view of the site context, while the energy rule base suggests a series of energy-sensitive design variables and their options. The HPB design support model of the inference engine is presented as a form of transparent tree architecture, such that stakeholders are able to choose preferred design paths, compare the paths, trace back the paths, and effectively revoke past decisions.

An actual small retail and office construction project was chosen as a test case to compare the utility of the pilot system against the conventional design practice. Most of all, the robustness of decision-makings was tested assuming a rather risky design change scenario in which the official design review board of the permit authority asks to expand the district heating to all spaces. As expected, the decision-making using the pilot expert system outperforms the conventional practice in terms of selecting designs with a good balance between economic value and energy performance. In addition, it was easier for users of the pilot system to forecast risks upon critical design changes and, in turn, to identify reasonable alternatives. Moreover, many performance simulations have to run whenever primary design decisions are made and/or in the event of design change in current design practice. However, running simulations is least likely if the expert system is used for the same situations, and thus there is a lesser chance of unforeseen increases in design and supplemental costs.

The pilot design expert system should be further developed to troubleshoot observed limitations and drawbacks. In addition, quality assurance by means of applying more robust and exhaustive test cases must be done. However, this expert system would meet the needs of HPB stakeholders who want to run building simulations without extensive knowledge and experience in building physics and/or HPB design practice.