Application of Artificial Neural Networks in the Prediction of Tire Manufacturing Defects

The article presents actual challenges faced by tire manufacturers and contemporary industry. Directions of development of methods for detecting and eliminating defects generated in the tire production process were discussed, with particular emphasis on methods using artificial intelligence. An exemplary classification of tire defects is presented. It was noted that a solution to reduce the amount of tire waste due to exceeding the uniformity limits is needed. Quantities describing tire uniformity were characterized. In the frame of the main purpose of the research, it was checked whether the model based on a traditional artificial neural network (with one hidden layer) can predict the value of conicity (output variable) based on five input variables. To solve this problem, the authors used the Multi-Layer Perceptron (MLP) - machine learning method, due to its ability to train non-linear models in “almost real time”. The parameters of the network structure were determined to guarantee the achievement of root-mean-square error (RMSE) for the training set data at a very low, satisfactory level. The authors see the high potential of using the built model in the mass production of tires. Application of mentioned model will minimize the waste of time and tire components scraps, and also will actually improve the quality of the final product.