Industries such as oil and gas, refining and power companies face the challenge of replacing obsolete safety and environment protection systems, with computer-savvy, enhanced methods and tools, in order to minimize risks and establish scientifically sound emergency safety processes.
A problematic aspect of exceptional events like large fires is that they cannot be experimented in real industrial environments. This work presents a system that implements, inside Blender, dynamic simulation of fires in complex industrial plants, that includes the assessment of heat transferred by the evolving flame on all surrounding objects.
The work is part of the Italian Government funded project SAFEMOD (Safety Modeling for High Risk Industrial Application). The system presented here is the core of a decision-support system, in which the safety management of an chemical/petrochemical plant can verify which are the most critical sectors of an industrial site, under which atmospheric conditions a fire might become more dangerous, what is the effectiveness of emergency procedures in containing the fire and in limiting damages.
The flame is simulated using Blender particle system, embedded in wind and vorticity fields. Parameters of the particle system are tuned to reproduce types of fire sources in oil plants, and the overall volume of particles is dynamically wrapped by a bounding box. One of the challenging components of our system is the computation of proximity between polyhedra representing the objects in the plant, and the deforming bounding box wrapping the flame. We adopted a strategy based on Voronoi regions, adapted to the specific application.The system has been fully implemented on a real oil and gas refining plant (Raffinerie Milazzo), located in southern Italy, on a section that spans an area of 1300 square meters. This section has been entirely scanned with 3D laser, and post-processed to obtain an OBJ format file,which has been imported into Blender.
Fire outbreak scenario are defined within a decision-support system, which supplies initial conditions,such as location of fire event, type of fire and fuel, atmospheric conditions, duration of the required simulation. The DSS feeds the simulation initial condition to a blender instance running on a multicore workstation, which runs the simulation and produces output files. The simulations results comprise radiation heath at all plant components affected by the fire, at the requested sampling time. A video of the running simulation is also provided at the end of the process to be used in conjunction with numeric data.