The Application Of The Optimize Manifold Learning In Vibration Signal Processing

The multimodal decomposition algorithm of vibration signal is the core essential algorithm ofvibration detection technology. The study of more accurate and reliable multi modaldecomposition algorithm is very important and has a good application value in the field of publicsafety monitoring. The future research in the field of Modal decomposition is to obtain a morerobustness and adaptability and better precision of modal parameter extraction method, so theeffectively reducing the modal aliasing and separating those modes as well as reducing thecorrelation between those modes is the core of the research in vibration signal detection algorithm.Manifold learning, the new dimension reduction theory which developed recently, is the coretheory of the various clustering recognition algorithm. Clustering algorithm(such as spectralclustering) based on manifold learning can effectively cluster the more complex distribution datapoints, and get good clustering results.This paper mainly studies new multimodal parameter identification method from theperspective of particle swarm optimization search, striving to avoid and improve the defects of thetraditional algorithms.First, modal vibration signal function as the kernel function is transformed to obtain higherdimensional space, it builds the high-dimensional space and low dimensional modal subspace forthe manifold learning algorithm. Next, by introduction of particle swarm optimization searchalgorithm, it builds appropriate learning evaluation function to obtain the optimal matching of lowdimensional modal subspace, then search the effective local low dimensional subspace embeddingthe higher dimensional manifold space, all of these are aimed at solving subspace search,subspace dimension selection, as well as the shortage of learning algorithm itself, it ensure the highefficiency, robustness strong ability of generalization of the algorithm. Last, it was used to extractthe modal parameters of vibration signal, mainly in the zhaobaoshan bridge cable, the test turns outthat the algorithm has high accuracy, stability and reliability.In conclusion, this paper proposes an effective vibration modal separation theory andalgorithm by introducing the particle swarm optimization for optimal manifold learning algorithm.This algorithm projects the multi-modal vibration signal onto a higher dimensional space and thenuses particle swarm optimization algorithm to improve manifold learning, so the intensive modal parameter identification problem is solved effectively. Using manifold learning and optimizationsearch algorithm search multiple local lower dimensional subspace in high dimensional space caneffectively supplement the deficiency of the manifold learning itself, so as to realize separationmodals, eventually get the modal parameters. Simulation test and actual project application testdemonstrates the reliability and validity of the proposed algorithms in this paper.