A method for establishing a LIBS quantitative analysis model based on a novel neural network

Abstract

This invention relates to a method for establishing a LIBS quantitative analysis model based on a novel neural network, comprising the following steps: Step 1, constructing a quantitative analysis model based on a KAN network structure; standardizing the original spectral data; employing KANLinear transformation at each layer of the model structure, using a combination of basis functions and B-spline interpolation to perform nonlinear mapping on the input features, wherein the basis function part performs a standard linear transformation and is nonlinearly adjusted using the SiLU activation function; through progressive dimensionality reduction using multiple layers of KANLinear, the data sequentially passes through feature representations of different dimensions and is mapped to the output regression layer to achieve the prediction of quantitative indicators; using mean squared error as the loss function and employing the Adam optimization algorithm for gradient update; the LIBS quantitative analysis model constructed by this invention based on the Kolmogorov-Arnold Network (KAN) network structure and Bayesian optimization algorithm effectively overcomes the limitations of traditional LIBS quantitative analysis methods.

Images