Research On Application Of Bearing Fault Recognition Based On Improved Salp Swarm Algorithm

In recent years,with the continuous acceleration of the national industrial informatization process and the continuous evolution of international development in the industrial internet and other aspects,industrial big data technology and various fields have produced a relatively deep technical integration and application integration.Due to the increasing demand for various industrial equipment,the industrial scale is rapidly expanded,resulting in an increasing in the amount of data generated in various fields.In particular,in the process of bearing fault diagnosis,the complex structure of nonstationary signals has attracted special attention,and the performance error detection also has high requirements.At the same time,with the development of machine learning,many typical intelligent methods can quickly extract valuable information,which brings important significance to fault diagnosis research.This thesis proposes an effective and improved algorithm for the identification of rolling bearing faults based on the salp swarm algorithm.By studying and improving the salp swarm algorithm,the algorithm is implemented using the Spark parallelization platform.The main work of the paper is as follows.(1)The salp swarm algorithm(SSA)is a population-based optimization algorithm with excellent performance.However,due to the lack of inertial parameters,the algorithm lacks the ability to find a global search for potential solutions.This thesis converts the continuous version of SSA into a binary representation,introduces chaotic mapping to make the initial population uniformly distributed in the space.Finally,the dynamic weight factor is used to update the position,so that the population is not easy to fall into the local optimum.According to the experimental analysis,the improved salp swarm algorithm(IBSSA)can effectively solve multi-objective optimization problems and has a strong ability to search for optimal solutions.(2)Considering the huge amount of data in the actual scene,the performance of analyzing traditional algorithms for processing data sets is limited,and it is proposed to run on the Spark programming model,which can save a lot of computing time.Comparative experiments analyze the operating efficiency of the IBSSA algorithm and other optimization algorithms in stand-alone mode and parallel mode,as well as the speedup and scalability ratios of different node scales and different data scales.Experimental results prove that the Spark platform can process massive amounts of data.(3)The bearing data set collected by Case Western Reserve University contains a lot of noise.First,this thesis uses variational modal decomposition(VMD)to decompose the source signal into several intrinsic mode functions(IMFs)according to the actual situation,decomposes and reconstructs the bearing vibration signal,and iteratively searches for the optimal solution of the variational model,which determine the best modal function.Secondly,the wrapper method using feature selection can effectively reduce redundant feature vectors,so that it can accurately extract fault feature frequency signals,thereby extracting effective feature vectors.Finally,support vector machines are used to classify and identify fault types,and their advantages are verified by experiments.