Abstract: In order to quickly identify the type and content of additives in the lubricating oil, the lubricating oil additive Vulcanized isobutylene (T321), Alkylated diphenylamines (T534), and Ammonium thiophsphonate (T307) were mixed in the base oil at different proportions. Using an extreme learning machine trains and tests the infrared spectral data construction model of the mixed additive, the greedy algorithm and genetic algorithm were used to optimize the input band, while the optimization results were compared. The optimal band interval combination was selected to eliminate the high excessive correlation band and improve the computational efficiency. The test results showed that the ELM model can effectively identify the type and predict content of lubricant additives. Compared with the traditional physical and chemical detection methods, it was an economical and rapid new lubricant additive detection method. And after the genetic algorithm band filtering optimization, the model output was better. The accuracy of identification of three additives reached 100% after band screening, and the content prediction determinant coefficient (R2) increased by 43.8%, 39.0% and 24.4%, respectively.