Research On Composition Monitoring During Laser Additive Manufacturing Titanium Alloy Process

In recent years,laser additive manufactured functional graded parts has great application potential in aviation,medical and chemical engineering.However,fluctuation of powder feeding rate and elemental loss and other issues in laser additive manufacturing process lead to the deviation of deposited material’s composition from expected one,which further affects the microstructure and performance of products.Therefore,an accurate and stable composition monitoring for molten pool during manufacturing process is of great importance for product’s quality control.By employing laser-induced breakdown spectroscopy(LIBS)in this work,a non-destructive and real time composition monitoring is achieved for laser additive manufacturing process.And with the application of support vector regression(SVR)algorithm in composition analysis,traditional composition analyzing method’s shortage in poor accuracy and stability,limited input spectral signal,and calibration model can be affected by processing parameters.Detailed works are presented as follows:1.Based on the feature of laser additive manufacturing,LIBS is introduced and a real time spectrum collecting system is built in this work.And a spectral pre-processing method is proposed to improve the spectral signal-to-noise ratio as well as the efficiency and accuracy for spectral analysis.2.Due to traditional method’s poor accuracy and stability,SVR is applied to composition analysis.SVR’s kernel function projects the nonlinear relationship of line-intensity-ratio and elemental concentration in real space onto Hilbert space in which the relationship become linear regression.The result shows that SVR’s accuracy and stability for composition analysis are better than those of calibration curve algorithm,partial least square regression algorithm,and artificial neural networks algorithm.3.Since SVR’s analyzing parameter plays a key role in SVR’s performance,an iterative 2-fold cross validation method is proposed to optimize analyzing parameter.The result shows that the accuracy and stability for composition monitoring is improved by using the optimal value in training model.4.Since calibration curve method uses single spectral parameter to build its correlation with elemental concentration.both iine-iritensity-rario and integrated intensity are used as SVR’s input to build its correlation with elemental concentration.The result shows that the added dimension of integrated intensity makes different sample’s convoluted distribution become linearly discerns in new space,which improves SVR’s accuracy and stability for composition monitoring when compared with the situation using only line-intensity-ratio as input.5.Since SVR’s analyzing parameter can be affected by the variation of laser power,a laser-power-conditioned SVR algorithm is proposed,The result shows that by substituting the function of optimal analyzing parameter versus laser power into SVR algorithm,composition monitoring under variable laser power condition is achieved.