Research On Laser Cladding Status Recognition And Source Localization Based On Acoustic Emission Detection

Laser cladding technology directly forms complex structural parts by melting and solidifying metal powder and matrix material with high energy laser beam.It has the characteristics of high forming efficiency and good mechanical properties.However,its technological parameters will directly affect the quality of forming parts.How to detect and identify the processing status online is of great significance for laser cladding manufacturing process.In this paper,laser cladding process status recognition and acoustic emission source localization are studied.The main work is as follows:1)The propagation characteristics of acoustic emission signals in stainless steel plates were studied,and the relationship between propagation distance and time-frequency domain amplitude and energy attenuation was obtained.Aiming at the noise interference of laser cladding AE signal,the influence of the selection of parameters such as basis function,decomposition layer,threshold function and threshold rule in wavelet packet on the noise reduction effect is analyzed.The experimental results show that Coif5 function,5-level decomposition and Visu Shrink rule have the best noise reduction effect.2)A multi-domain AE signal feature extraction method based on time domain,frequency domain and waveform parameters was constructed.A 14-dimensional feature matrix was determined for laser cladding processing state,including energy,ringing times,RMS,kurtosis,sparse factor,sample entropy,band energy entropy and singular value.On this basis,a feature selection method based on correlation analysis is designed,and a feature optimization method based on t-distribution stochastic domain embedding(t-SNE)algorithm is proposed,which realizes the de-redundancy optimization of the feature matrix and improves the effectiveness of subsequent laser cladding process status recognition.3)A least squares support vector machine(LSSVM)state recognition method based on Parameter Optimization of niche particle swarm optimization(NPSO)algorithm is proposed.The best combination of parameters optimized by NPSO algorithm is selected to improve the accuracy of classification and recognition.The results of laser cladding experiments under different process parameters verify the validity and practicability of t-SNE-NPSO-LSSVM model.4)Aiming at the problem that the traditional TDOA location method is difficult to apply in laser cladding acoustic emission signal,a multi-point location optimization method based on improved fruit fly optimization algorithm(FOA)is proposed.Acoustic emission events are extracted by the joint threshold method of kurtosis and sparse factor.The optimal objective function of multi-point location is improved by rotational symmetry.The dynamic threshold is optimized according to the attenuation characteristics of amplitude to calculate the time difference accurately.The FOA algorithm is improved from the aspects of search radius and target function.The experimental results show that the method can effectively improve the convergence speed and location accuracy.