Research on rubber inclusion defect detection based on terahertz time-domain spectroscopy technology and image fusion

Rubber, in the process of production and service, may contain foreign objects, such as fragments and films. These inclusions can lead to increased wear of rubber products during actual use, resulting in unpredictable safety risks. Therefore, it is essential to perform non-destructive testing to ensure the quality of rubber products. In this paper, a high-precision and non-destructive detection method for rubber inclusion defects based on terahertz technology is proposed. The terahertz time-domain spectroscopy was used to detect prefabricated Silicone Rubber samples with metallic inclusion defects. The obtained terahertz spectral data are optimized by using correction processing algorithms and feature extraction methods. And a highly efficient quantitative detection model for internal inclusion defects in Silicone Rubber is established using machine learning algorithms. Firstly, SNV, AirPLS, AsLS and BEADS are individually applied to preprocess the collected time-domain spectra. Secondly, the features of the spectra are extracted by using CARS, UVE and PCA, respectively. Lastly, Partial Least Squares Regression and Least Squares Support Vector Machines are employed to establish quantitative prediction model of the depth of rubber inclusion defects. The experimental results show that the BEADS and CARS algorithms can greatly reduces the computational load of the model while improving its accuracy. The LS-SVM model has the best prediction effect, and the RMSEP and R_P of the prediction set are 0.0717 and 0.9964, respectively. In addition, the physics-based model of time-of-flight is also employed to calculate the collected time-domain spectra and predict the defect depth, and the RMSEP and R_P of the prediction set are 0.1080 and 0.9986. In terms of terahertz imaging, this paper proposes a high-quality visualization processing scheme for internal inclusion defects in rubber. It utilizes various feature parameters for imaging and employs a combination of grayscale histogram equalization and wavelet transform image fusion methods to achieve high-quality imaging representation of internal inclusion defects in rubber. The THz-TDS techniques enables rapid and non-destructive detection of location, depth and shape of rubber inclusion defect, providing new technological methods for the quality inspection of other polymers.