Abstarct: Automating fault detection in a manufacturing process is an important step to improve quality, reduce manpower and promote customer satisfaction. In Tile factories, an essential fault, which sometimes cannot be detected by human, is the lamination fault, which is due to the remained air within the tile. More quickly finding this fault, the cost will be less. In this thesis an intelligent method baxsed on ultrasonic waves is proposed which can detect the fault before firing up the tile. At first, we prepared a dataset of 546 ultrasound waves of healthy and damaged tiles using an ultrasound transmitter and receiver. Then we used Fast Fourier Transform and Wavelet for feature extraction. Several classifiers including MLP, KNN, SVM, LVQ and Bayesian are tested to classify healthy and damaged tiles. MLP achieved 97% accuracy which is the best results among the selected classifiers. At the end, Sugeno Integral is used for classifier combination, which improved the classification accuracy up to 100%.