Ultrasonic inspection of foundry pieces applying wavelet transform analysis

Object identification techniques are finding increasing use in many industrial applications. A defect recognition method for foundry pieces in this field is proposed. The system classifies the pieces and selects the apt ones, which will later be machined within the automobile industry. The inspection of the pieces is carried out applying ultrasonic sensing. Due to the ultrasound properties, this type of vision is very appropriate for industrial environments. Starting from the signal reflected from the pieces, the treatment of the data is approached in two significant steps. First, the discrete wavelet transform, DWT, is applied to the analysis of ultrasonic waves for feature extraction. Second, a neural network is used to carry out the discrimination of the foundry pieces. This automated signal classification system obtains great results and the use of the tandem DWT analysis-neural network is shown to be a powerful technique for this type of application.