Establishment of a classification model for different grades of ribbed smoked sheets using near infrared spectroscopy and convolutional neural networks

Natural rubber is a vital industrial raw material and strategic resource. Ribbed smoked sheet is a traditional solid natural rubber that accounts for a significant proportion of the manufacturing of high-performance tyres. Presently, the grading of ribbed smoked sheet is primarily accomplished through manual labor, with considerable influence stemming from subjective factors. Accurate and objective differentiation of ribbed smoked sheet grades is imperative for subsequent production and processing. This study used a handheld near infrared (NIR) spectrometer to collect spectral data on natural and ribbed smoked sheet across two main categories and four grades. A classification model for ribbed smoked sheet was then established using a convolutional neural network (CNN) algorithm and compared with traditional methods, such as the multilayer perceptron and random forest algorithms. The results showed that the CNN-based classification model for ribbed smoked sheets had the highest accuracy on the test set at 98.33%, and the Area Under the Receiver Operating Characteristic (ROC) curve (AUC) for each category exceeds 0.99, demonstrating the best overall performance. The model was further validated using an independent external sample set, achieving an accuracy of 80%. The classification model developed using a handheld near infrared spectrometer and convolutional neural network algorithms offers a new technical approach to grading ribbed smoked sheet.