The Application of Machine Learning to Detect Rust on Steel Structures in Warehouse Buildings

Abstract

Nowadays, the logistics industry has a demand for warehouse floors in each key point in Thailand. Since reaching and demanding of product from service users is higher in the electronic commerce business group resulting in this type of building in causing more warehouse buildings to be built. The majority of which use steel as the main structural material, it is one of the most important structural materials in the building industry even though it is an unstable object. When used, corrosion and wear occur. The corrosion that usually occurs with steel is rust. Rust is a form of corrosion caused by the oxidation of iron when exposed to air and moisture. This study aims to detect images of rusty parts to enable efficient planning of building repair management that collecting data from 133 pictures of rusty and non-rusty parts in a warehouse construction project by smartphones. Then, the pixel counts of red, green, and blue colors were classified and creating a model to evaluate the performance with the K-Nearest Neighbors algorithm. The results showed that the model was more than 70 percent accurate in predicting rusty images. The model can estimate preliminarily rust in steel structures in order to plan for maintenance.