Identificação do desalinhamento de transportadores de correia por meio de sinais de vibração em um circuito de manuseio

Abstract

Misalignment in conveyor belts is a significant factor contributing to premature wear, operational failures, and increased maintenance costs in industrial material handling systems. This study presents a methodology based on vibration signal analysis for the automatic detection of belt misalignment, using an experimental test bench developed to simulate this condition under controlled parameters. Geometrical modifications were introduced in transfer chutes, designed through simulations using the Discrete Element Method (DEM), to generate gradual misalignment. Data were collected with triaxial accelerometers, and time-domain statistical features were extracted from the signals. These features were analyzed using machine learning algorithms, including Support Vector Machines (SVM) and Decision Trees, evaluated through cross-validation techniques. The models achieved high accuracy in classifying alignment conditions, with features such as kurtosis, peak value, and K-factor showing strong influence. The proposed approach proved effective for misalignment diagnosis and supports its integration into predictive maintenance strategies.