Open surface tanks often are used in industrial practice. When harmful substances are involved, control of worker exposure requires the use of a local ventilation system. The push-pull system, among others, involves a jet of air that is blown from one side of the tank and collected by an exhaust hood on the opposite side; this system can save up to 50% of the ventilation air. Several guidelines are available for design of such a ventilation system, mainly based on experimental results. However, their validity is confined inside a narrow operating window. In this work a mathematical model developed based on computational fluid dynamics has been used to extend the validity of the existing guidelines outside the range in which they have been validated, with particular reference to tank width and to the velocity of the air drafts.

n order to achieve good comfort and safety conditions in industrial premises, a ventilation system, either general or local, is often I necessary. While general ventilation allows the contaminant to enter the indoor air and involves an overall exchange of air, the local one tries to capture the pollutant before mixing with the ambient air, thus providing a better control with the use of a lower air flow rate (ACCIH, 1995). However, in some industrial processes the use of local ventilation is not possible since it can interfere with the working practice. Crebenc and Coodfellow (1 995) have recently presented a general review of both local and general ventilation in industry.It is not always easy to design properly a general ventilation system since both geometry and airflow rate can influence the effectiveness of the system. Apart from rules of thumb (which are mainly based on previous experience, see for instance ACCIH, 1995), there are two approaches to the design of a ventilation system: performing experiments using a mock-up of the industrial premises, or, carrying out computer simulations through computational fluid dynamics (CFD) models (Heinsohn, 1991). The first method provides reliable results (at least when isothermal conditions are involved) but it is expensive and time consuming. Moreover, any other change in the ventilation system requires the construction of a new model. The second approach is less expensive, more flexible and faster. With respect to the pioneering works in the 1980s (Tanguy and Dupuis, 1986; 1988), nowadays, faster computers and more robust algorithms are available, thus allowing for the use of CFD models in daily design work (Nielsen, 1994). However, while for velocity field a close agreement between experimental results and model predictions have been already demonstrated, fewer validations are available for pollutant concentrations. This paper summarizes the results of a study for an industrial building where thermocouples are calibrated using methanol baths. As a first step a mock-up of the industrial building has been constructed and fitted with different general ventilation systems. For each configuration, the pollutant concentration values in the breathing zone have been measured. Then, the pollutant dispersal has been simulated using a commercial CFD code (CFX, 1997) to assess its reliability in these conditions. Before performing these simulations, the influence of several parameters (such as the turbulence model, boundary conditions, and grid size) have been assessed through some preliminary computations. Afterwards, other computer simulations have been carried out to investigate the effectiveness of different ventilation systems in realsize buildings. *Author to whom correspondence may be addressed. E-mail address: renato.rota~~olimi.itThis work investigates a general ventilation system for reducing the health hazard in industrial premises. Several tests carried out in a laboratory mock-up have been used to validate the reliability of a generalpurpose computational fluid ...