Abstract: According to the case in air spray process, the spray shape of robot spray gun is not a cone and an elliptical paint area adhered to a planar surface is formed, the paint-film thickness of ellipse area is obtained by a painting experiment with unmovable spray gun. Neural network(NN) method is employed to fit the paint-film surface function that denotes the thickness distribution of paint-film in ellipse area. A new analytical paint-film thickness distribution function named as ellipse dual-β model is proposed by comparing the fitted paint-film profile on some different sections in x and y directions. Genetic algorithm (GA) and least square are used to fit parameters of the new model according with experiment datum. By the ellipse dual-b model, simulation result of a single painting stroke is made out to compare with a practical single painting stroke experiment result to prove that the model and its parameters are right to the experiment case. The formula for computing paint-film thickness of any painted point on a curved surface and a motion equation for spray robot painting the curved sur-face are also given. This model is a new practical spray-painting gun model for off-line robot painting programming.

Key words: Ellipse dual-b distribution model, Genetic algorithm, Robotic spray-painting, Spray gun model