Fire Simulation Prototype Tool Based On Diffused Limited Aggregation

The fire is an essential part of nature and it provides heat energy, which helps people to perform various tasks such as cooking, steel manufacturing and constructions industry however fire is also very dangerous if misused because it might cause a loss of property and life. Researching on fire simulation tools is more important because of the following reasons; firstly, the fire simulation tools are used for training the firefighters and society to understand the behaviour of fire and its spread prediction, secondly, the fire simulation can be used in the evaluation of rescue plans, tools required for rescue and conducting various experiments for different fire fighting strategies and tactics that will be required during the rescue and control operations, thirdly, the fire simulation plays an important role in the film special effects, computer games, military emulation, and virtual reality, fourthly, the fire simulation also has been a very important topic in the computer graphics due to the varying shape and the complexity of its physical mechanism.The existing fire simulation tools uses computational fluid dynamics (CFD), which use many particles at a time and utilizes the equations and physical properties that requires complex algorithms. Hence, many computation discrete steps are required to obtain more accurate simulation results of the fire behaviour and its spread prediction as a result a huge computation power and cost is required to perform complex computations and rendering of many particles. In this thesis we developed fire simulation prototype tool based on Diffused Limited Aggregation (DLA) as a means of reducing the computation complexity and cost of resources. Our idea of simulating behaviour of fire by using DLA is due to the fact that the DLA uses simple steps to achieve very complex simulations and it can be generated by one particle at a time instead of using many particles at a time as used by Computational Fluid Dynamics (CFD). For this reason the cost required for rendering will be reduced because only one particle will be rendered at a time. The implementation of our fire simulation prototype tool based on Diffused Limited Aggregation (DLA) is achieved as follows:-The Diffused Limited Aggregation (DLA) that used by our tool to simulate the behaviour of fire is generated by the arbitrary point known, as a seed, which will be fixed and should be allocated on the scene as the fire ignition point, this point will save as DLA start point. The next DLA particle will be launched far from the seed point and perform a random motion, our tool uses a random walk to model the random walk that will be performed by a newly launched particle with the interval of time. The launched particle, which performs the random motion, will stop whenever it collides with the seed and becomes a part of the formed structure and another particle will be launched. This process will continue till the maximum number of particles that are required to create the DLA have been reached as a result the path of fire will be formed. The position of launching new particle should be computed randomly every time whenever the new particle is going to be launched, to achieve this we modified the Mel Siegel algorithm which generates the random points and distribute these points in a sphere by considering the point P(x,y,z) to be a position where the particle will be launched in a sphere with Radius R, this radius should be projected to the XY plane to form projected radius Rxy which will be used to compute the new value of x and z since the particle’s launching position should be computed in XZ plane then the value of y will be set to zero. To avoid the particle to perform a random walk forever walk which will reduce the efficiency of our tool; we enhanced the efficiency of the DLA formation by limiting the maximum number of steps that are required by the newly launched DLA’s particle to perform a random walk; if this number has been reached and the DLA’s particle not yet attached itself to the growing DLA cluster then the DLA’s particle will be killed and a new DLA’s particle will be launched; to enhance the chance of attachment for the DLA particle to the growing DLA cluster; the radius of the sphere will be altered every time whenever the DLA’s particle is generated. This will ensure that the DLA’s particle will be launched within the length of the radius of the sphere by doing so the DLA’s particle will not be launched far away from the growing DLA’s cluster as a result the number of killed DLA’s particles will be reduced. The DLA’s branches can be formed by DLA’s particles that attached freely to the growing. However, in our tool the formation of each branch should take into consideration of all factors, such as those that affect the growth of fire because the formation of the branches is required during the computation of the growth rate of fire. We achieved this by ensuring that every new launched particle is assigned the creation number which refers to the already generated DLA’s branch; by using this number the DLA’s particle will identify itself which position of the currently formed DLA to attach itself and form a branch.To simulate fire the computer generated scene whereby the DLA, which simulate the behaviour of fire, will be created on it, all the factors that affect the spread of fire and other objects such as fuel, burnable and barrier will be generated and positioned on the scene. In order to achieve this the simple3D scene have been implemented. A3D scene will be generated first by creating a3D wire mesh, and then applying the texture, which is2D bitmap image to the created3D mesh. We then created the slope by implementing a function to generate a topography by choosing circle with its centre randomly and the radius R of this circle will be calculated, then the circle inside points will be upward displaced and the other points will be downward displaced. Then the highest point will be chosen as the centre of the newly created topography. We have implemented the function which will position and assign the size of the burnable and burrier objects randomly on the scene and each object which placed on the scene should be identified whether it is a burnable or barrier object so that can be distinguished during simulation. Then the tool, will initialise and assign the initial value of moisture to the light fuel and burnable object on the scene. We have added several collision detection functions in our tool to ensure that the simulation run as smoothly as possible by checking the collision that might occur at the scene. Our computer generated3D scene can have more than one DLAs; so the collision detection function have been implemented to ensure that these DLAs will not overlap to each other and prevent another path of fire to be created on the already formed path and avoid the fire to start burning the area, which has already burnt out. The function for collision detection between the scene’s borders and generated DLAs have been implemented so that the simulation will not be done outside the scene; to do so the scene’s area should be known first and then the area of the scene where the DLA will not grow, will be computed. The collision detection of the growing DLA, and other objects in the scene such burnable and barriers have been added to our tool, whenever the collision detected by this function first it will identify whether the object on the scene is burnable or burrier. After the identification the DLA, then it will ignite the nearby burnable objects and continue creating the path of fire and pass through it; otherwise, if it is a barrier object then the DLA will not pass through it instead it will find another available position and continue creating the path of fire.The wind, moisture and temperature play the important roles on the spread of fire. The function has been implemented to assign the strengths of the wind and its direction to the DLA. Hence, the branches of DLA that are formed in the same direction with wind will grow faster compared with the DLA branches that are formed in opposite direction of wind. The temperature of the scene has been related to the moisture contents of the scene by implementing the Rule of Thumb.To compute the growth rate of fire the factors that affect the behaviour of fire and the time needed to burnout all fuel points on the ground have been taken into consideration and will be assigned to the DLA, which simulate the path of fire. Thus, the rate of DLA growth will depend on the computed fire growth rate, which can also be used to determine the rate of changing colour of the formed DLA, for this case the faster the colour change the higher the growth rate of fire. The burning time progress function which controls the simulation has also been implemented ensure that the simulation is as accurate as possible.We have performed the simulation output analysis of our prototype tool by altering the settings of variables that affect the behaviour of fire growth through the user interface and we found that the tool has simulated the behaviour of fire by considering all factors that affect its growth and meet the goal of our thesis. The issues that need to be improved have been discussed and left for the future improvement.