Development and implementation of HVAC-KBCD: A knowledge-based expert system for conceptual design of HVAC&R systems - Part 1 : General framework

The typical HVAC&R design process involves generating a relatively small number of alternative solutions (configurations of systems, subsystems, and components) and then utilizing analysis and estimation tools to select a “best” solution based on predefined selection criteria, for example, lowest lifecycle cost. This approach is limited by the engineer’s experience, capabilities, and time constraints, often resulting in missed opportunities for optimum system selection. Further, the latest equipment, design approaches, and concepts are often not even considered. As a result, the overall design process is being gradually compromised. A knowledge-based expert system (KBES) can augment the capability of current advanced building simulation tools by creating the necessary means for automatically generating, analyzing, and sorting numerous design solutions subject to both site and user constraints. This paper describes the general framework of such a KBES called HVAC-KBCD (knowledge-based conceptual design), capable of automatically synthesizing the complete set of possible HVAC&R systems, which can then be analyzed using available building energy simulation programs such as DOE-2.1 E. This synthesis is done by first pruning or shrinking the solution space of the design alternatives by applying static expert knowledge (Level 1) and then by a guided search dominated by dynamic application knowledge (Level 2). Level 1 starts with a database of all feasible secondary systems (and another for all feasible primary systems) and uses assembly rules and application rules based on heuristics expert knowledge, design practice, and standards to generate the permissible design alternatives of secondary systems (and primary systems). Level 2 uses initiation rules along with matching rules to combine secondary systems among themselves (since the building has different zones) and with primary systems. These capabilities have been programmed into a commercial hybrid KBES shell along with procedural programming capabilities that have the ability to perform simple heuristic cooling and heating design load, as well as first and maintenance cost calculations. ANSI/ASHRAE/IESNA Standard 90.1-2001 has been used to set the minimum allowed energy performance for equipment such as fans, chillers, boilers, etc. This paper describes the general framework, structure, capabilities, and inner working of the HVAC-KBCD, while a companion paper will illustrate and demonstrate the entire methodology described here to the conceptual design of office buildings. CURRENT CONCEPTUAL DESIGN PROCESS HVAC&R engineers and designers face an increasingly complex process while designing cost-effective and energyefficient HVAC&R systems. This complexity is a result of a relatively large number of variables: increasingly stringent design criteria, availability of different forms of energy and energy prices, introduction of new technologies, and emergence of new methodologies for analysis. AIA (1997) divides the design services conducted by the architects and other building professionals, such as structural, mechanical, and electrical, into three activities. 1. Schematic or conceptual design (elaborated below). 2. Design development phase when, upon approval by the architect/owner, the HVAC&R engineer produces detailed design documents, including equipment schedules, layouts, and typical construction details. 2004 ASHRAE. THIS PREPRINT MAY NOT BE DISTRIBUTED IN PAPER OR DIGITAL FORM IN WHOLE OR IN PART. IT IS FOR DISCUSSION PURPOSES ONLY AT THE 2004 ASHRAE WINTER MEETING. The archival version of this paper along with comments and author responses will be published in ASHRAE Transactions, Volume 110, Part 1. ASHRAE must receive written questions or comments regarding this paper by February 6, 2004, if they are to be included in Transactions. Itzhak Maor is the engineering director of PWI-Energy, Philadelphia, Pa. Agami Reddy is a professor in the Civil, Architectural and Environmental Engineering Department, Drexel University, Philadelphia, Pa. 3. Construction document generation, when, upon further approval, detailed drawings and specifications are produced. The focus of this paper is upon the HVAC&R schematic design phase—also called the concept development stage— where the HVAC&R engineer generates the design concepts by synthesizing1 systems from a given number of subsystems and components. Examples would include all-air variable volume, packaged rooftop DX, and water-source heat pump systems. For large systems, the number and type of chillers, boilers, cooling towers, pumps, etc., need also to be selected. The configuration of systems and components is normally based on the engineer’s own experience from previous similar projects. It is important to note that the number of the design alternatives that can be evaluated is usually limited by time constraints. After establishing several design concepts, the engineer uses tools, such as hour-by-hour energy simulation programs (such as DOE-2.1, Winklemann et al. [1993]), for predicting energy consumption and energy cost. Such programs are being used to assist the HVAC&R engineer in selecting the optimal HVAC&R systems during the conceptual design phase. Once the structure of the HVAC&R system is selected and configured, these simulations execute up to 8,760 runs (representing a year of hour-by-hour simulation) of the predefined HVAC&R systems, resulting in a relatively accurate estimate of the year-long performance, energy consumption, and/or operating cost. However, this method relies heavily on the designer’s knowledge and experience in selecting and synthesizing near-optimal configurations for the software to simulate. Building energy simulation programs normally have a library of systems and components (for example, chillers, boilers, pumps, fans, coils, etc.), which can be recomposed (or synthesized) to simulate entire systems using predefined equipment models. For new equipment, the engineer can use curve-fit techniques to fit the data provided by the equipment vendors and thereby generate new models that reflect the performance of this specific type of equipment. These libraries are extensive and normally include most of the systems available in the industry, and they even include systems that do not comply with current standards and practices, such as the energy standards (ANSI/ASHRAE/IESNA Standard 90.12001). The energy estimates can then be incorporated into more accurate life-cycle cost (LCC) analysis, where usually the system selected has the lowest LCC; in many instances, the selection is based strictly on the lowest owning (or first) cost system. Pricing information can be obtained from sources such as R.S. Means Mechanical Cost Data (Means 2002a), R.S. Means Maintenance and Repair Cost Data (Means 2002b), equipment representatives, and construction and service contractors. In many cases, the engineer and the owner use a “selection matrix” (as described in ASHRAE [2000]) for decision making instead of the more rigorous procedure explained previously. The final product is a set of documents that include a full description of the design criteria and the design constraints, a description of the selected systems, sizes and capacity, a preliminary sequence of control for the proposed systems, and conceptual drawings and schematics. This allows the owner (or owner’s representative) to identify the most appropriate design to satisfy both the needs and stipulated budget. PROBLEM STATEMENT AND SOLUTION APPROACH Currently there are no mechanisms to automatically synthesize feasible secondary and primary systems that can then be exported and linked to the corresponding models in an hour-by-hour building energy simulation program. As explained earlier, the configurations have to be defined apriori, resulting in a limited number of alternatives (as shown in Figure 1) and limited system configurations. The proposed solution methodology is to automate the process of generating a set of feasible HVAC&R secondary and primary systems that can be evaluated using available, detailed simulation programs. Leveraging the computing speed available today, hundreds of combinations can be run in a reasonable time. After the simulation and the preliminary evaluation of such a large number of alternatives, the engineer selects the most promising alternatives (say, the top ten) that meet the owner/ engineer’s criteria on which to perform more precise evaluations. This proposed solution methodology will benefit both experienced as well as less mature professionals during the conceptual design stage, resulting in a more scientific and comprehensive evaluation and, hence, a better design solution. The practical culmination of the research described in this paper will be the development of an HVAC&R synthesizer (or configurator) utilizing AI methods for search and knowledge representation. A comprehensive literature review of artificial intelligence and KBES in buildings and HVAC&R system design has been published in a previous paper (Maor and Reddy 2003). The design process can be presented as a problem-solving process or, more accurately, as a search process (Simon 1999). For example, design synthesis can be described as the search for one or more design solutions through the selection of systems and subsystems. Similarly, artificial intelligence (AI) focuses on two areas: search and knowledge representation (Ginsberg 1993). Search is concerned with the examination of a large number of possibilities (which can be extremely large) and finding an optimal solution. Knowledge representation, on the other hand, is intended to introduce intelligence (like matching the right secondary systems to the appropriate primary system) into the search stage, thereby reducing the number of possibilities to be studied. The conceptual similar1. The terms “synthesis” and “configurations” are used interchangeably in this paper.