Research On Prediction Method Of MLCC Electrode Structure Parameters And S-parameters Based On Machine Learning

With the arrival of the 5G era,the demand for Multi-layer Ceramic Capacitor(MLCC)is increasing and the requirements for their superior performance become higher in succession,so how to produce high-performance MLCC has become the research goal nowadays.In this research,the relationship between MLCC design parameters and electrode structure parameters is modeled by machine learning,and the design problem of MLCC electrode structure parameters with special functions is studied and solved;the relationship between preparation parameters and S-parameters is modeled,and the specific performance of MLCC before preparation is predicted by the model.This research introduces the MLCC parameters,machine learning algorithm,equivalent circuit model and MLCC sample parameter measurement method.On this basis,on the one hand,through linear regression algorithm,the design parameters of MLCC,such as rated voltage,rated capacity,test frequency,S-parameters of corresponding design and dielectric layer thickness,are taken as input,and the electrode structure parameters of MLCC,such as inner electrode layer number and inner electrode thickness,are modeled as output,and the model is used to predict MLCC electricity.The electrode structure parameters were compared with the actual MLCC Electrode structure parameters.On the other hand,through k-nearest neighbor algorithm,the preparation parameters of MLCC,such as rated voltage,rated capacitance,test frequency,thickness of dielectric layer,number of inner electrode layers and thickness of inner electrode,are taken as input.The electrode structure parameters of MLCC,such as the number of inner electrode layers and the thickness of inner electrode,are used as output to model.The model predicts the electrode structure parameters of MLCC and compares them with the actual MLCC Electrode structure parameters.On the other hand,through k-nearest neighbor algorithm,the preparation parameters of MLCC,such as rated voltage,rated capacitance,test frequency,thickness of dielectric layer,number of inner electrode layers and thickness of inner electrode,are taken as input,and S parameter is taken as output.The model is established,and the predicted S-parameter of MLCC and the actual measured s parameter are analyzed.At the same time,in order to deeply study the validity of S-parameters of machine learning prediction,combined with the series RLC model in low frequency circuit,full wave representation model in high frequency circuit and transmission line model commonly used in research and practical application,the MLCC equivalent circuit modeling system is proposed in ADS electromagnetic simulation software,through which the value of each component in equivalent circuit is extracted.The actual measured S-parameters and predicted S-parameters are converted into the values of the components in the equivalent circuit,and the values of the matched components are compared and analyzed to draw a conclusion.To sum up,the model established for MLCC by linear regression algorithm in this research basically matched the electrode structure parameters predicted with the actual measured electrode structure parameters,indicating the feasibility of predicting the electrode structure through the model and playing a guiding role in the production of MLCC with specific performance.Through the K-nearest neighbor algorithm for MLCC model,prediction of S-parameters and the actual measurement of S-parameters on frequency change trend is consistent,and through the MLCC equivalent circuit modeling system,the S-parameters for the circuit realization of each component in the value.After analysis,it is shown that the electrical performance of the corresponding MLCC can be accurately characterized by combining the preset rated capacitance with the predicted S-parameter.