No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

An expert system has been constructed for the purpose of assisting in the design and analysis of direct gain passive solar environments. This system has been constructed for the use of senior undergraduate architecture students in a computer-based design studio. The primary use of the system is in the role of an educational tool which generates design recommendations from user input and predicts some physical characteristics of the environment.The system is applicable to passive solar environments with vertical, south-facing glazing. The system incorporates three models. The first model represents an attached sunspace with no thermal mass storage. The second model represents a direct gain living space. The third model represents a direct gain living space integrated with thermal mass storage. The third model allows the use of floors, ceilings, and walls as mass for thermal storage. Four representative mass materials (concrete, adobe, common brick, and dense concrete masonry) have been included for comparison purposes. Four representative sub-climates are also incorporated into the system: cold / arid, hot / arid, hot / humid, and cool / humid. For educational purposes, the system makes separate calculations for identical structures based on models for inhabited and uninhabited cases.The system incorporates scientific and mathematical relationships as well as rulesof thumb which have demonstrated their applicability to passive solar design. The system performs calculations based on work by Balcomb, et al. [5, 9], and Duffle and Beckman [1], to estimate environmental temperature swings, total solar energy input, and thermal absorption by mass storage elements. The system also utilizes models based upon work by Mazria [4] to recommend glazing areas. Recommended glazing areas are calculated from user input variables such as structure type, site latitude, and floor area.The system's ease of use allows it to be adapted for various classroom goals, and its generalized nature permits the instructor to adapt it easily into different areas of architectural design curricula. The system is written for use with the CLIPS expert system shell. / Department of Physics and Astronomy

No description available.

Knowledge compilation improves search-intensive problem-solvers that are easily specified but inefficient. One promising approach improves efficiency by constructing a database of problem-instance/best-action pairs that replace problem-solving search with efficient lookup. The database is constructed by reverse enumeration- expanding the complete search space backwards, from the terminal problem instances. This approach has been used successfully in counter-planning to construct perfect problem-solvers for sub domains of chess and checkers. However, the approach is limited to small problems because both the space needed to store the database and the time needed to generate the database grow exponentially with problem size. This thesis addresses these problems through two mechanisms. First, the space needed is reduced through an abstraction mechanism that is especially suited to counter-planning domains. The search space is abstracted by representing problem states as equivalence classes with respect to the goal achieved and the operators as equivalence classes with respect to how they influence the goals. Second, the time needed is reduced through a hueristic best-first control of the reverse enumeration. Since with larger problems it may be impractical to run the compiler to completion, the search is organized to optimize the tradeoff between the time spent compiling a domain and the coverage achieved over that domain. These two mechanisms are implemented in a system that has been applied to problems in chess and checkers. Empirical results demonstrate both the strengths and weaknesses of the approach. In most problems and 80/20 rule was demonstrated, where a small number of patterns were identified early that covered most of the domain, justifying the use of best-first search. In addition, the method was able to automatically generate a set of abstract rules that had previously required two person-months to hand engineer. / Graduation date: 1992