Research On Monitoring Cr Composition In FeCr Alloy Fabricated By Laser Additive Manufacturing

In recent years,gradient functional parts manufactured by laser additive manufacturing technology have been widely used in aerospace,medical equipment,industrial production and other fields.In the actual processing process,the element content in the formed sample deviates from the design value due to factors such as loss of elements and fluctuations in the amount of powder fed,which ultimately affects the microstructure and mechanical properties of the formed sample.Therefore,accurate and stable on-line monitoring of the composition during the manufacturing process is critical to controlling the quality of the shaped sample.In this experiment,Fe and Cr powder was used as material to produce alloy samples by metal laser additive manufacturing technology.The optimal process parameters were obtained by orthogonal experiment:laser power,laser spot diameter,scanning speed and powder feeding rate were 700 W,0.5 mm,5.0 mm/s and 4.6 g/min respectively.The spectral signal acquisition real-time system based on laser-induced plasma spectroscopy is used to collect the plasma spectral signals generated during the manufacturing process.The acquired spectral signals are processed by the pre-processing method proposed in this paper,which can improve the signal ratio of the signal and the accuracy and efficiency of the spectral analysis.By analyzing the relationship between spectral characteristics and elemental concentration,it is concluded that the intensity ratio and integral intensity of the spectral have a strong correlation with the concentration of Cr,and because the traditional calibration curve method has the disadvantages of different sensitivity of different element concentration ranges.Therefore,this paper uses chemometric analysis to quantitatively analyze and monitor elemental components.In this paper,the partial least squares method(PLSR),BP neural network and support vector regression(SVR)are used to construct the component monitoring model.The input vector is spectral line intensity ratios.The monitoring accuracy and stability of these models are compared and analyzed to obtain the best chemistry.According to the comparison,the element monitoring model build based on SVR algorithm has the best performance.The SVR algorithm uses the kernel function to map the nonlinear relationship between the line intensity ratio and the element concentration from the real space to the Hilbert space,so that the nonlinear relationship between the line intensity ratio and the element concentration is linear in the H space.Since the analysis parameters of the support vector regression model determine the performance of the model,this paper proposes to use the two-fold cross-validation method to optimize the analysis parameters.The results show that the accuracy and stability of the component monitoring model obtained by training with the optimal analysis parameters are improved.The calibration curve method uses a single spectral characteristic to establish a relationship with the element concentration,which affects the accuracy and stability of the model.In order to solve this problem,this paper uses line intensity ratio and integral intensity as the input vector of the chemometric analysis method.The results show that increasing the input makes the accuracy and stability of the component monitoring model improved.