Sliding contact and rubbing and wear in the bow net system will affect the life of the carbon skateboard. To predict the changes in the friction coefficient and wear rate in the bow net system, this article uses its homemade sliding electric contact test machine to carry out the flow friction and wear experiment. Inquire about the characteristics of the friction coefficient and wear rate under different operating conditions, and then proposed to improve the pelican algorithm optimization nuclear limit learning machine network prediction model. Using Tent chaotic reverse learning strategy, Sine and cosine algorithm strategy and adaptive t-distribution strategy to improve the Pelicans optimization Algorithm (POA), and verify the superiority of the Pelicans algorithm. Then use the improvement of the Pelican optimization algorithm (IPOA) to optimize the kernel extreme learning machine (KELM) parameters to verify that the accuracy of the predictive model is as high as 98.9 % and 98.3%.