REDUCING INDOOR AIR POLLUTION
MIT researchers are studying an approach to ventilation that could improve indoor air quality and save energy.
While conventional ventilation systems mix large quantities of newly conditioned air into a room's air, "displacement ventilation" systems prevent mixing by injecting limited amounts of air, slowly and near the floor. Pollutants and heat produced by people and equipment rise naturally to ceiling exhaust vents, and the fresh air rises into the breathing space. The difficulty is that the precise system specifications--air temperature, velocity and so on--must be tailored to the space being ventilated or occupants may be uncomfortable.
Researchers led by Professors Qingyan Chen and Leon Glicksman of the Department of Architecture, the Building Technology Program and the Energy Laboratory are developing tools that can help. A computer model calculates how different system specifications affect airflows and heat and pollutant dispersion in a well-defined room.
They have also built a full-size experimental room to test the effects of different ventilation strategies in various situations. For example, the experimental "office" has models of people, equipment, light and a window which emit heat and contaminants and affect airflows as they might in a real office. The researchers inject air with controlled properties (temperature, velocity, humidity) through the diffuser. They can then both measure and observe directly how changes in those properties affect airflows and heat and pollutant dispersion throughout the room.
The model predictions and experimental results agree quite well. Both confirm that a displacement ventilation system can provide clean air and comfort and also reduce the amount of ventilation air that must be heated or cooled--a major consumer of energy in commercial buildings. Sponsors are the American Society of Heating, Refrigerating, and Air-Conditioning Engineers; Halton, Inc.; Trox GmbH; and the NSF.
OPTIMIZING INVENTORY LEVELS VIA DATA MINING
Knowledge management and "data mining" can benefit companies in a variety of ways. For retail companies, data mining can be used to optimize inventory levels, detect fraud, identify target customers and forecast demand.
Dr. Amar Gupta, co-director of the PROFIT (Productivity from Information Technology) Program and a senior research scientist at the Sloan School of Management, studied the inventory levels of a retail distribution organization. He used a neural network, an approach to computer architecture and programming that emulates human brain functions, to study customer demand patterns for different items at different stores.
His research led to the creation of a new approach that enables a retail distribution organization to decrease inventory levels by 50 percent, yet still maintain a 90 percent customer satisfaction level. The work was funded in part by the Defense Logistics Agency and the Defense Advanced Research Projects Agency.
A version of this article appeared in MIT Tech Talk on January 28, 1998.