Fatigue life prediction of natural rubber components using an artificial neural network

Uniaxial tension fatigue tests conducted on dumbbell-shaped specimens made of vulcanized natural rubber are carried out. Using the fatigue test data, a back-propagation neural network (BPNN) model for estimating the fatigue life of natural rubber specimens is established. An improved sine-cosine algorithm (ISCA) is proposed to optimize the parameters of the BPNN model. The peak engineering strain, ambient temperature, and Shore hardness of natural rubber specimens are used as the input variables while the rubber fatigue life as the output variable of the BPNN model. The regression results and predicted life distribution of the established BPNN model are encouraging. For comparison, a genetic algorithm, a particle swarm algorithm, and a standard sine-cosine algorithm are also used to obtain the BPNN model parameters, respectively. The results showed that the prediction accuracy and efficiency of ISCA are better than other three algorithms. In addition, the sensitivity analysis is introduced to quantify the relative influence of the model inputs on the output.

Automated summary

Uniaxial tension fatigue tests are conducted on vulcanized natural rubber specimens to establish a BPNN model for estimating fatigue life. An improved sine-cosine algorithm (ISCA) is proposed to optimize model parameters, showing better prediction accuracy and efficiency compared to other algorithms.