Aiming at the problems of large data volume, low correlation and poor reliability of the multisource signal obtained in locomotive gearbox detection, a new intelligent optimization algorithmmultivariate function particle swarm optimization algorithm is proposed. The influence of the variation ratio and fitness of the particle population on the inertia weight is studied. Based on traditional particle swarm optimization algorithm, the convergence speed and efficiency of the algorithm are improved. Taking the fitness function of the regularized modal difference as the evaluation index of the number of the measurement points, the multisensor detection optimization of the gearbox is realized according to the modal vibration type analysis of the gearbox. Taking the tooth break fault of the gearbox as the measurement object, through comparative analysis with traditional detection methods, the proposed method accurately obtain the results: the gearbox input shaft rotation frequency of 395 Hz, the thirdstage meshing frequency of 905 Hz and its 2~5 harmonics components, then the position of the faulty gear is identified quickly. The experiment results show that the proposed method can enhance the recognition rate of structural parameters, effectively improves the fault diagnosis accuracy and also provides a key technical foundation for locomotive fault warning and safe service.