Corrosion: Simulating and Rendering Stephane Merillou, Jean-Michel Dischler, Djamchid Ghazanfarpour Corrosion: Simulating and Rendering Stephane Merillou, Jean-Michel Dischler, Djamchid Ghazanfarpour 이승주 1/15

-Introduction -Physics of Corrosion -Modeling and Rendering corrosion -Result Contents 2/15

corrosion reactions can be divided into two main categories: metallic patinas and rusty layers. we propose a model to simulate and render new forms of corrosion that correspond to the second category. we describe our random-walkbased spreading process, resulting in a “corrosion map” Introduction 3/15

Corrosion can be defined as the deterioration of a material against its environment. On a physical point of view, corrosion is an electrochemical process requiring the presence of an electrolyte. The chemical reactions in the case of iron for example imply the creation of many different oxides composing rust. These oxides all have their own colors, explaining the usually noisy and non-uniform aspect of rust color. Physics of Corrosion Schematic atmospheric corrosion principle (a) metallic patinas, (b) destructive corrosion. Some rust constituents and their colors Introduction 4/15

Uniform corrosion: corrosive attack over the entire surface area (at least a large proportion). Galvanic corrosion appears when two different materials are coupled in a corrosive electrolyte Pitting corrosion is a process which produces localized cavities in the material Crevice corrosion is often due to a differential aeration: that is, a sudden difference of the oxygen accessibility over two parts of the same material Physics of Corrosion Corrosion: qualitative models 5/15

sufficient estimations of the steel corrosion rate are given in [14] according to a simple specific atmosphere corrosivity classification. These values permit us to estimate the amount of wasted matter according to a certain time the following equation (its coefficients permits the experimental results of table 2 to be well matched [17] –w(t) = k ⅹ tⁿ Quantification of corrosion Physics of Corrosion steel corrosion mean rate (g=m2=year) vs atmosphere corrosivity as a function of time. Analytic functions to evaluate corrosion rate by evaluating the lost weight in (g=year) 6/15

For each metallic object, the user selects the amount n o of starting points. Assume that the object is given in the form of a mesh. Initially each face F i has a probability coefficient p i. The probability coefficients are modified by the two conditions (internal and external). Concerning internal conditions –There is no simple way to physically quantify micro-structural imperfections. –We empirically propose to use two new coefficients. A microstructure imperfections factor k str;i A greasy factor k gr;i Modeling and Rendering corrosion Starting points of corrosion 7/15

Concerning external conditions –the pi coefficients are modified according to the global scene and object geometry –Two objects are in contact (collision) with each other. –The object is isolated –Isolated metallic objects (no collision and no significant differential aeration) are affected by uniform corrosion Modeling and Rendering corrosion Starting points of corrosion Galvanic corrosion and differential aeration corrosion principles 8/15

Once all faces Fi of all objects have their own probability coefficient p i given by p i = p i;initial ⅹ k srt;i ⅹ k gr;i ⅹ k ext starting points simply can be selected randomly according to these coefficients The coefficients k ext, k str;i, k gr;i are selected by the user Modeling and Rendering corrosion Starting points of corrosion 9/15

Our approach consists in simulating the corrosion process on a “thick plate” (corrosion map), then mapping this plate onto the object. With these starting points, the spreading process can be run onto the map. Each corrosion map pixel contains following information: a color, a porosity coefficient, a roughness coefficient Modeling and Rendering corrosion Computing the corrosion-map 10/15

The first step consists of computing a list l cp of “already” corroded pixels. the corrosion spreading process is applied to p s by randomly choosing one of its non-corroded neighboring pixels. Concerning p v, the spreading process goes directly in depth, that is, we decrease the depth value by. Modeling and Rendering corrosion Computing the corrosion-map corrosion spread 11/15

Corrosion does not affect all the pixels of the map at the same time as induced by the use of a time dependent spreading technique with preferential starting points. we need for each pixel to evaluate a specific δ This is achieved by calculating for each pixel a “local corrosion time” tlocal –w eq = pixel heigth ⅹ S p ⅹ ρ iron –t local = e^(1/n) ⅹ ln(w eq /k) –Δw = w(t local + 1) - w(t local ) –hl = Δ w/( ρ iron ⅹ S p ) Computing the corrosion-map Calculating δ 12/15

On non-corroded parts :use the Cook and Torrance BRDF model [3] for iron. On corroded parts :use the model which accounts for both porosity and roughness. –Porosity is empirically chosen to be 80%, since rust is a very porous matter –Roughness factor : measured using a HommelWerke HommelTester T2000 Affecting roughness and porosity Modeling and Rendering corrosion 13/15

Result 14/15

Result 15/15